I‘m not even sure what the argument for banning Chinese models/open weights even is supposed to be?
1. if it’s to stop hackers doing hacking things with „uncontrollable models“ then, well… they’re already doing something illegal to begin with, why would they care about breaking another law running these models?
2. if it’s to stop foreign actors, then that ban would not apply to them anyway
3. it’s not stopping distillation either, Chinese labs are already banned from using US frontier models and look at how good that is working
I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term. But thats also an admittance that the American labs can’t compete on merit anymore, and should by itself also limit the viability of the idea that all those VC billions will ever make a return? In any case this would be something benefitting only a very few for a short time (labs + investors).
Someone please enlighten me what the actual argument here is, cause I can’t see it.
> Someone please enlighten me what the actual argument here is, cause I can’t see it.
> The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term. But thats also an admittance that the American labs can’t compete on merit anymore
That's the argument. To be precise the publicly stated argument is that they're attacking American providers by distilling. The real aim is to eliminate competition because otherwise Anthropic and OpenAI are non viable and the US views them as crucial for winning the 'AI race' which they see as putting whoever wins it on top in terms of warfare/economic power etc.
But isn‘t the whole argument of „winner takes all“ already dead, if china is such a big threat that the US labs need import controls?
Distilling or not, they are clearly close enough to the frontier, that the supposed „free market“ country needs market controls. That may work for US markets. But not the rest of the world.
Idea: petrodollar policy becomes „tokendollar“, enforced by US military dominance. If you force me to pay altman at gunpoint, then maybe ill stop using kimi
They are simply out of ideas to keep the most expensive party in the history of mankind going.
That's it. Simple as that. When you grasp for straws in panic mode, you don't exactly spend time strategizing and weighing the pros and cons of each straw carefully.
Decade? We're 4 years in since ChatGPT released publicly and have gone from "Wow this thing is stupid" to running coding agents on auto mode, and the Chinese labs have been around for even less time. The US providers may lose dominance in 1 or 2 years.
In some portions they already lost dominance. My claude will happily generate disgustingly vulnerable code then flag me for ToS violation when I ask it to make sure the code that it just made, in the same context window, in the same conversation is safe and doesn't have glaring holes. Also great to be charged $xx.xx+ amounts and then it refuses to answer without refunding it.
They won't. China doesn't have the training capacity and the quality data sets, the former because of access to fabs, the latter because of internal censorship that is getting worse by day.
The current panic is happening precisely because Chinese open weight models like Kimi and Qwen are demonstrably competitive with SotA models like Fable, but far cheaper.
No, the Chinese tech companies routinely super optimize for a given 'famous' or 'established' or 'mainstream' benchmarks. Any other concern is secondary.
In other words, they look good on surface, but suck whenever anyone put them to any serious use. That's why they are cheap, they have to be cheap because they suck.
Having limited amounts of high quality Chinese data to train on doesn’t really decrease or affect the appeal of these models to the rest of the world. I don’t think the CCP is restricting or limiting the datasets that models can be trained on.
Also the models themselves don’t even seem to be censored or restricted that much. If the Chinese government is fine with the guardrails being on the API level and maybe even restricts the usage of external providers that’s again a net benefit to everyone else since they would have little incentives to force Chinese companies to lobotomize their models.
You are probably about capacity so we can only hope that Huawei and others can catch up and break Nvidia’s monopoly (since Intel and AMD suck too much too much to accomplish anything useful that’s the next best thing)
Also Chinese labs are actually still publishing research publicly which alone would accelerate and increase the competitiveness of open AI models developed in the US or even Europe.
> doesn’t really decrease or affect the appeal of these models to the rest of the world
Anyone who tried using Chinese models for serious programming tasks ditch them quickly if they have a choice.
> I don’t think the CCP is restricting or limiting the datasets that models can be trained on.
What I meant was censorship limits the amount of high quality training sets by limiting the amount of all training sets from which high quality sets grow out from, e.g. the entire set of posts on Baidu forums before 2017 is gone forever.
> so we can only hope that Huawei
Chinese companies do not have access to commercial-scale advanced nodes. The impact is at best negligible.
> Anyone who tried using Chinese models for serious programming
Perhaps, but that’s because they are just objectively worse for many tasks than GPT/Claude. I don’t see how that ties to censorship inside of China.
> censorship limits the amount of high quality training sets by limiting the amount of all training sets
Well again, I don’t really understand how is the lack of high quality Chinese datasets matter much if they have access to English/etc. datasets. That’s only an issue to their users inside of China.
> not have access to commercial-scale advanced nodes
Well the gap was way bigger a few years ago. If Chinese companies can at some point produce GPUs that are competitive cost wise i.e. they are willing to tolerate significantly lower margins than Nvidia (whose margins are obscene) that’s not a huge issue.
well free markets meant something very different in adam smith's time; what they meant by free was 'free from rents' whereas today its usually mean in the milton freedman terms as 'free from any interference' which is how i interpret the parent poster...
This Schroedinger's Freedom is getting tired now. They happily label themself as free-market democratic liberal progressive when you're justifying bombing a poor third-world country, but suddenly claim to be unfree undemocratic oligarchic whenever anyone asks them to act on that principle.
Yes, it’s freedom both from state interference but also from monopolies, oligopolies and trusts using their position to suppress the competitiveness of other market participants. Which is generally impossible to prevent without the state interfering in the markets.
US didn't have that much rational reasoning when it killed Japanese exports in cars, chips, and finally software. That in turn lead to China simply taking the seats Japan used to occupy, but as far as imposition of these protectionist means is considered, US does not need that much sophisticated justification internally.
The argument is - these models costs hundreds of billions to develop - to BOTH US and China. China is giving them for free despite not making financial sense because they want to weaken US's tech firms ability to compete - because China sees AI dominance as critical for their national security.
The US also sees AI dominance as critical for their national security, and relies on a market economy and non-state-owned labs. This means that these frontier labs are truly susceptible to "predatory pricing" (economic term for a competitor selling at a loss to eliminate you), which is illegal in the US and any other free market economy exactly because of its implications on market efficiency.
I hate it just as much as the next guy, but the fact the discussion here ignores the fact that these concerns and dynamics are real just lower the discussion level instead of actually discussing potential solutions.
With that said, what I'm worried about is that the government solution will be far more hostile than some semi-ban on open models. An example of an even worse scenario, they could take over frontier labs and restrict access to everyone in the public (which might still not solve espionage).
> China is giving them for free despite not making financial sense because they want to weaken US's tech firms ability to compete - because China sees AI dominance as critical for their national security.
Nobody is making this argument because this is a normal thing that companies do. It's so common it is named in business strategy: "commoditize your complements."
There's a 2002 Joel On Software post explaining this using tech examples that were already old in 2002[1].
In the present, Meta is doing exactly the same thing with Ollama as Alibaba (who is funding some of the Chinese labs). Alibaba needs advanced models internally, and does not want to have a dependence on US models that can be arbitrarily shut off by Washington. Alibaba also wants to drive usage of its cloud among companies who are now demanding competitive AI models (which Alibaba may be locked out of providing). However, selling AI is not its core business. Putting cheap AI in the world creates demand for cloud services, which is part of Alibaba's core business.
Not only does this make financial sense for Alibaba, it is fairly basic.
Google gives away consumer software like Chrome and Android (not their core business) to drive search, which is their core business.
et cetera. The frontier labs are exposed in exactly the same has been way every pure play tech company before them. The US labs chose quality as their moat; remains to be seen whether that was a good choice, or if they can add another moat in time.
> Putting cheap AI in the world creates demand for cloud services
No, Alibaba doesn't have the right amount of hardware to serve these models.
The part that benefits Alibaba in these open models is to maintain its public reputation so as to maintain its stock price. The core that drives everything else inside Alibaba is its e-commerce.
> Nobody is making this argument because this is a normal thing that companies do. It's so common it is named in business strategy: "commoditize your complements."
It is common in business strategy, far less so in nation-state governance.
The reason you, me, and so many people in the United States still have a far better quality of life and level of freedom than most people in China is because it’s bad to run a country like a business, full stop. When you do so, you kill far more people.
> It is common in business strategy, far less so in nation-state governance.
It is not clear why this is bad for Alibaba or China, could you expand on that?
> it’s bad to run a country like a business
The "Chinese miracle" resulting from China's integration with the global economy has lifted ~800 million people out of poverty and raised standards of living quite rapidly overall. A lot has been written on the topic, but suffice to say that it's a fairly unorthodox argument to make that the Chinese have not been well-served by the arrangement.
> level of freedom
The context of this discussion is whether the US government should prevent US citizens from accessing yet another product that is freely available in the rest of the world, to benefit a tiny handful of rich people in California.
Let me be clearer: if you do not hold stock in OpenAI or Anthropic, you likely stand to benefit from more AIs being provided by more providers. Imagine if the government had banned MySQL and PostgreSQL, so we all had to buy databases from Oracle and Microsoft. That would be strictly worse for everybody except employees of Oracle and Microsoft. This is exactly the same thing.
For private companies to an extent yes. In public companies that’s just a way to increase your employees compensation without harming your free cashflow. I don’t think tax wise it’s any different than them paying you more in cash and then you buying the same amount of shares.
> The reason you, me, and so many people in the United States still have a far better quality of life and level of freedom than most people in China
I think it's because every single time the farmers in south america or the philippines try to organise and get out of poverty you guys either bomb them or arm some terrorist to kill them.
But sure tell yourself otherwise if that makes you feel better.
That’s extremely American centered, though? Despite the constant complaining the median American is still very well off compared to almost everyone else. If you go to the top 10 percentile or so them maybe only Switzerland and some very small countries (tax havens, petro states and such) can really compete.
> you, me, and so many people in the United States still have a far better quality of life and level of freedom than most people in China
Whether this is true or not is still open to debate, but the fact that it’s presented as some kind of absolute truth that needs no substantiation is simply ridiculous.
Well no.. regardless of anything else the median American generally has q higher standards of livin than the resident of pretty much any major country (including Europe). While compared to e.g. Germany or France those in the lower percentiles are probably doing worse China specifically still has a huge amount of poverty. Even if you happen to be legally allowed to live in one of the major cities and have a decent job.. well 966 is still a thing and you don’t get paid $500k like the OpenAI employees allegedly forced to sleep under their desks.
> The US also sees AI dominance as critical for their national security, and relies on a market economy and non-state-owned labs.
[...]
> but the fact the discussion here ignores the fact that these concerns and dynamics are real just lower the discussion level instead of actually discussing potential solutions.
We had this problem solved for DECADES. Build the best post-secondary education system in the world. Lure all of China's (and the rest of the world's) best and brightest. Educate them. Give them well deserved, high paying jobs doing what they love and invite them to build a life and stay here.
Took me a while but for the west the value of education isn't primarily in jobs but in democracy. It puts the bar much higher than we've ever accomplished. People are supposed to run th county (believe it or not)
> costs hundreds of billions to develop - to BOTH US and China
OpenAI and Anthropic are extremely bloated compared to the Chinese labs. They have thousands of extremely well paid people working for them. Deepseek I think has a couple of hundred and I doubt they are paid a lot.
Of course compute for training is expensive but I doubt that’s necessarily the overwhelming majority of the operating costs for Anthropic/OpenAI
> on a market economy and non-state-owned labs
Sure the private companies developing these models are heavily interlinked with and even maybe partially owned by the government. But its a matter of degree. Since that’s the situation in the US similar in how these companies both shape government policy directly and in how the government is constantly intervening and shaping their behavior outside of any regulatory framework.
The US government is both heavily protectionist and interventionist these days.
I think the misconception here is that they "give them away".
Sure, you can download the weights and run it yourself, but to run a 1T+ parameters you need roughly 2TB of VRAM (for full precision). That costs around $500k to $1M, not including electricity and rent.
In reality, they provide their models priced via API or subscription just like any other company. The "free weights" are only useful for other research labs. It seems that China understands that research should be shared and open, which is one of the reason that AI has been able to develop that fast in the last 10 years.
> free weights" are only useful for other research labs
No, they are useful for any compute provider who can host these models and sell inference. Which effectively puts a cap on whatever the Chinese companies can charge. So longterm that means that their margins are limited to cost + some low margin.
If they are making money by selling tokens then anyone who has sufficient compute to host their models can make more than the Chinese labs since they don’t have any R&D costs.
I mean I think the primary reason is really the attention is all you need paper and meta's "leaking" of meta's llama.
China is plenty happy to steal IP and just say "this is just how business is done in China it's a cultural thing" but quickly run to defend their own IP in western courts when stolen by western businesses. It works one way.
And I have experienced this firsthand when having something manufactured there. Exact clones appearing on Alibaba just days after.
The software landscape is full of free products that are used to upsell other services. In this case the upsell is inference at data centers. China has built a big advantage in Green energy and is working at the GPU issue both on the hardware and model level. Those are the two main cost inputs into AI and should allow China to compete. The exact model being used is not going to matter as much as the cost that companies must pay.
> "predatory pricing" (economic term for a competitor selling at a loss to eliminate you), which is illegal in the US
It is very funny that this was/is the playbook of a huge number of successful, VC-funded, "disruption" plays.
Cheap cabs that undercut the market, wait for taxi companies to die, raise your rates now that your "rideshare" app got everyone hooked on the subsidized rate, IPO!
> This means that these frontier labs are truly susceptible to "predatory pricing" (economic term for a competitor selling at a loss to eliminate you)
OpenAI and Anthropic were losing money too until recently, when they started focusing more on increasing revenue (which is also part of why companies are now looking more at Chinese models to reduce costs).
That's nonsense, people pretend that a "frontier model" is the most important technology ever, but we know for sure it cannot be because several companies are doing copycats of one another. It is not China that is doing much effort to copy-cat, it's the US companies that are copying one another. The only real barrier in this technology is how many Nvidia chips you can buy and how much data you can get to train your LLM. All these models are basically moving in the same trajectory, it is essentially a very expensive copy-paste job they're doing.
How could you tell how close they are to the frontier if they weren't distilling? Clearly those labs think distilling is worth it despite the large effort they need to go through to bypass anti distillation techniques used by the frontier labs
Anthropic is “distilling chinese models”. Everyone uses outputs from every other frontier model in both pre training and post training, hilarious to give it this name and pretend like it’s an attack. Fable is just a distillation of stolen copyrighted works, etc.
I find such beauty in the fact that the robbers are getting robbed but it's all part of progress overall. It's not that China won't come up with what the US are coming up with. Just a matter of when it would happen.
Yep. We'll steal China's IP as well so all of this really is just no big deal. Everyone is a robber in some way.
Though one must ask, why are American models banned in China? Hmmm.
> It's not that China won't come up with what the US are coming up with. Just a matter of when it would happen.
I agree. All this discussion about China releasing open weight models is amusing. It's like ok they release open weight models.... and so what? If they leapfrog ahead of the US then we'll just undercut them with our own open weight models.
> Though one must ask, why are American models banned in China? Hmmm.
Because they're not? But the American government and Anthropic don't want American models to be used by the Chinese. Anthropic is notorious for its effort to tell if the user is within China.
China is taking the moral high ground when Anthropic bans users from China. But if Anthropic didn't do it, the GFW would have surely banned Anthropic because Claude will say lots of things that do not agree with the official narrative.
> Though one must ask, why are American models banned in China? Hmmm.
Because the models are behind a gate and the data goes overseas. The same is not true for open weight models. One should avoid the former and prefer the latter simply based on principle, not on country of origin.
That's just conjecture though surely. Like not saying you're wrong but the same thing would result from crawling and using results from Chinese Web for training along with everything else. If the most prevalent model people are posting responses from online in Chinese is qwen then of course that would happen. There's no evidence to say either way really.
I am sure that some other country (Mistral from France, India, Canada etc.) would have even more chances to innovate perhaps.
and because Chinese models are open-weights, the distillations effects of them could be easier done as well ;)
In my opinion, its a win-win plus even within worst case scenario*, I already believe that the current open weights models are in general speaking good enough perhaps for my and other use cases as well and I feel like we will probably most likely get more open-weights model for a long time in general as well perhaps.
The other countries don't have cheap capital, cheap infra, or scale. The talent is not the bottleneck. Look at how many SV companies are helmed by Indian executives. If the Europeans cared an iota about AI progress, Huggingface wouldn't have been headquartered in NYC.
wow if they're that crucial, but economically nonviable, we should probably just nationalize them and fund the superproject with taxes, then put the results into the public domain!
> But thats also an admittance that the American labs can’t compete on merit anymore
This is exactly it.
Try to buy a BYD in the United States. You can't (without a complicated process) because they're so much better cars than our domestic brands that our domestic brands couldn't compete and lobbied to keep them out.
> should by itself also limit the viability of the idea that all those VC billions will ever make a return?
It just has to last until the next quarter / fundraising round.
I can understand (not necessarily defend) protectionism for products with nonzero manufacturing cost and some required physical size and such.
I am befuddled when the same policies are applied to products with marginally zero manufacturing cost, no real physical size, and such. Regulating ideas is hard folks.
The first token on a new AI rig costs $X (full capex cost) then every token after that costs virtually zero. Over time the cost/token trends towards zero (modulo opex). That said, AI does have higher opex than general SaaS so it can’t get as close to zero.
But that’s kind of a different question, the running of some service. The “product”, the model, is a collection of files. The “manufacturing” required to add another customer is “send them the files” and has ~zero marginal cost.
There is a finite number of tokens that a rig will turn out over it's lifetime. Divide the cost of the rig plus electricity by that number and you have your cost per token. Yes providers can screw up on scheduling and end up paying more than they should, but that's not magic.
The AI rigs cost money, sure. But no one is talking about regulating those. We are talking about regulating models, which do have zero cost of reproduction.
It’s like banning the leaked DeCSS key. Good luck.
objectively true. And there's zero need to factor in things like "human rights" or "environmental protection", because these are costs that have been externalized and the consumer don't pay for it (in their lifetime anyway). That's what objectively means.
Chinese cars are unavailable in the US because US manufacturers are 100% unable to compete. Japan was in the exact same position before, but because of ally status, japanese car companies (with their gov't as the proxy) managed to negotiate and have a "voluntary import restriction" in place to protect US manufacturers. This artificial restriction created a higher price for consumers than it ordinarily would have been.
The japanese car manufacturers were also able to setup shop in the US, and instead of exporting, they manufacturered directly on US soil. They're allowed to do this sole because they're an ally, and i don't expect this same to be allowed for chinese companies.
You're not seeing it because you're thinking in terms of internal US economics, but this is a question about international economics and geopolitical leverage in the distant future.
The reason to consider the ban is because it might be the only way to preserve a fully autonomous and independent American frontier AI stack and the long-term strategic value of possessing such a stack could vastly outweigh the cost of giving up true free market competition on AI. If giving US startups and other companies access to cheaper Chinese AI means sacrificing the US's ability to own its own frontier AI stack, is that a rational trade, or would it severely and irrecoverably sacrifice the country's technological autonomy and leverage for decades to come in exchange for cheaper tokens for a little bit early on?
If the US not only gives up most of its manufacturing capability to China, but also allows itself to give up its own AI stack and become almost entirely dependent on foreign AI, then it's conceivable the combination of the two sacrifices will deal a permanent deathblow to the country in exchange for what will turn out to have been a couple decades of cheap goods and AI tokens.
It puts all US companies at a price disadvantage and forces American company to shoulder the load of training frontier models most will never need while the rest of the world has cheap AI access.
The keyword in your sentence is "access". The rest of the world would have cheap access to AI, but no ownership of anything at the frontier. If the US joined the rest of the world, declined to develop frontier AI, and allowed itself to become dependent on cheap Chinese AI, once that dependency was fully realized it would eventually translate into a lack of access to the frontier. That doesn't matter until, suddenly, it becomes the single most important thing to the entire country's future.
China is making a simple bet: that the US will offshore AI in favor of cheap tokens, just like the US previously offshored manufacturing in favor of cheap goods. They're doing this because they know that if they alone possess frontier manufacturing and AI capabilities, then they alone can build the world's most powerful technologies in the future which combine the two (e.g., robotics that will revolutionize all their industries, domestic life, and military far beyond any other country).
> The keyword in your sentence is "access". The rest of the world would have cheap access to AI, but no ownership of anything at the frontier.
So... just use distillation and create their own frontier models?
I don't see the problem here.
If the goal is to hold sovereign ownership over your own SOTA models, China has already shown that any nation could achieve that pretty easily if they need to.
And that's ignoring the fact that many of these models are open weight so hosting/owning sovereign inference/intelligence itself is entirely a hardware problem.
This whole conversation reminds me a lot of North Korea deciding they needed their own OS when Linux is right there. If the technology is open and available to all, then sovereignty over this core technology is irrelevant.
Linux is not a relevant comparison, because the fundamental capabilities of an OS do not grow exponentially with time or investment.
If building full sovereign ownership of the production chain for frontier-level AI were so easy, there wouldn't only be two countries on the entire planet that have that right now. If distillation worked in the way you seem to imagine it does, nearly every country in the world would have a full sovereign frontier AI stack.
They would if they cared to invest in it. That they've not is a policy and economic choice (possibly because they don't have the same sense of panicked urgency I see among a lot of folks around here), not evidence of some insurmountable barrier to entry.
You also jumped to a conclusion that I don't think is self-evident: that owning the production chain matters.
Assuming open weight models continue to advance, who cares? Just let the US and China expend the compute on model development.
And if they close up, then fine, do what China did and build up a domestic industry and distill as a way to get a jumpstart. We already know that's possible since China already did it.
I think it would be helpful if you explained how distillation works (as relevant to the conversations threads here) because I agree that this is a key point.
Spot-on and really well written post highlighting the challenges here. Really it just comes down to the US and surrounding sycophants and China. There just isn't any other game in town. For the US to just, well, give up on AI development would be civilizational suicide. I wouldn't put it past us though, most Americans in particular are so weak minded they only care about cheapest possible product. It's interesting to see how well propaganda campaigns have been working to harm American interest in AI and related technology (data centers are a bit of a different story). Young students are applauding guest speakers giving them the news they want to hear about how AI is bad and they need to tear it all down, while in the real world those same students are actively committing career suicide by refusing to innovate or adapt. It wouldn't really matter except that they vote, and those votes will wind up doing harm under the guise of doing good, but really it's a matter of delusion that we will all pay the price for.
Anyway the good news is everyone says China's strategy of just releasing open-weight models is the coup de grace for American technology companies and all the investment, research, all of that stuff by all of these companies who certainly don't care to survive and continue making trillions of dollars and can't possibly be managed by the same genius engineers who post on HN will just go away.
And then once that happens we'll just do what China did, and let them build a bunch of AI capabilities, scale out data centers, invest trillions, then we'll just release open-weight models too and then what?
You don't need to know a whole lot about AI to realize that most people on the Internet that are commenting on these stories haven't thought about how the real world works for more than two seconds when it comes to this stuff. China can enact a strategy, we can copy that strategy and even improve upon it.
Surprise Pikachu.
You (not you) can't claim a lead is unimportant and this undercutting strategy is so good and then simultaneously say the US can't do it back to China. That's dumb. If the lead doesn't matter everyone will stop doing AI research because being undercut is too expensive. How likely does that seem, and why isn't China stopping research if that is true and they just need to keep releasing open-weight models? Sorry, but being on the leading edge matters a whole fucking lot. As I've written in other posts this is like a cheap Android phone from Wal-Mart versus the latest iPhone. There's a reason you buy the iPhone and not the cheapest Android phone you can find even though they have all the same apps and both make phone calls.
I'm open to changing my mind here, but I have yet to see a convincing argument. Just a lot of pearl-clutching and China fear mongering.
We can see the same situation play out with manufacturing capability.
The USA has recognised that it needs more manufacturing capability. But rather than just build that capability (which is what the Chinese state-planned economy would do, and did) it tries to incentivise private companies to build it. Which didn't work because the companies building domestic manufacturing face competition from Chinese imports. The investment money can get better returns chasing property deals or rent-seeking monopolies.
In the "we can do it back" case, it still takes a US private company a lot of time and money to distill their model off the Chinese model. There has to be a lot of profit for them to do that. That profit isn't really available - it's a competition with all the other models, and no monopoly rents. There's for sure a viable business there, but it's a "lifestyle business" not the 1000x growth business that VC's want.
Lots of people want this outcome. We now know that the closed-weight model isn't an effective moat, but if the closed-weight model business fails, the open-weight business has nothing to distill, and model progress slows.
I suspect this is part of why China dipped into their vast strategic oil reserves to reduce purchasing and offset the shortage caused by Trump's blunder in Iran. If energy prices get too high, training will slow. They know they can pirate our models at 95% fidelity, so they want training to continue so they can pirate the next ones also.
They've invested in capabilities that undercut attempts at AI supremacy by US companies. Every time they put those capabilities to use:
- They look more and more like somebody who can prevent whatever world domination plans appear to be brewing in the minds of US billionaires. They can turn this perception into leverage over other countries by threatening to close their model weights.
- They get to use the improved models to boost their economy's efficacy (they probably think they can put it to use better than anybody else, and they might be right).
- For every $100 that the US paid for a given unit of improvement, they just pay $1. The longer the US keeps this going, the less stable it will become because the sacrifices the US has made to pull it off are causing things like measles and explosive diarrhea and will ultimately lead to unrest. Meanwhile the longer China keeps this going the less likely the US is to stand in their way, because the US will have neglected overlong the things that make it a threat to China (the legitimacy of the dollar, soft power programs like USAID, a thriving economy).
Also they have leverage over US because they've recently demonstrated that they can raise the price of oil. That means that they can pop the AI bubble anytime they want. So the longer the US is dumping everything into model improvement, the more opportunities they'll have to use that leverage.
> The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term.
> Someone please enlighten me what the actual argument here is, cause I can’t see it.
But you did see it.
The US has invested trillions in AI that the companies involved are never going to make back. Even without competition from open models, but definitely not with it. And if those trillions turn out to be worthless, that's going to have a massive impact on the market and cause a lot of bankruptcies.
Banning those open weight models isn't going to fix everything, but it would the impossible obstacle slightly smaller.
> The US has invested trillions in AI that the companies involved are never going to make back. Even without competition from open models, but definitely not with it. And if those trillions turn out to be worthless, that's going to have a massive impact on the market and cause a lot of bankruptcies.
This is the point. And what nobody seems to bring up that inference is a massively profitable margin. The labs can still grow at 80% margin as opposed to 99% margin (I do not have the numbers). But that will hurt the valuation and investors.
With protection from Govt, you can compete at 1.5 or 2x of the equivalent open weights price. But it is not sustainable to keep the prices high, just a matter of time.
> The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term
It won't even do this. Streisand effect will probably draw even more attention to the open weight models
The entire marketing narrative in AI already operates this way --- "GPT 2.0 is too dangerous to release, oh noooo!" etc
After a report from RSA Security, who were in a licensing dispute with regard to the use of the RSA algorithm in PGP, the United States Customs Service started a criminal investigation of Zimmermann, for allegedly violating the Arms Export Control Act.[5] The United States Government had long regarded cryptographic software as a munition, and thus subject to arms trafficking export controls. At that time, PGP was considered to be impermissible ("high-strength") for export from the United States. The maximum strength allowed for legal export has since been raised and now allows PGP to be exported. The investigation lasted three years, but was finally dropped without filing charges after MIT Press published the source code of PGP.[6]
I remember t-shirts with the DVD encryption key on them when the authorities tried to ban that. This shit always backfires. You'd think they'd know this by now.
The CFO or wtv that tweeted, their argument was, if you allow cheap free competitive models, you will stop the influx of capital in frontier development of even more powerful LLMs, thus capping how far the tech could reach. Obviously, they're mostly asking for regulation to save their money, but it seems their argument is that without an environment that rewards the development of the next frontier model the tech might not evolve as fast.
> if you allow cheap free competitive models, you will stop the influx of capital in frontier development of even more powerful LLMs, thus capping how far the tech could reach
Hmm. So either the weights are open for all to use, or nobody trains those weights at all? Sounds like the ideal outcome to me.
The worst possible outcome is the one where frontier labs train godlike AIs then kick the ladder out from under them. That should be prevented at all costs. No country or corporation can ever be allowed to win the AI race, for technofeudalism will follow swiftly after. As long as they keep competing with each other, we're the winners. If that music ever stops, it's time to watch out.
I'd rather have AIs achieve sentience and wipe out humanity as a whole than live in a world where they submit to the whims of gigacorporations and totalitarian governments for the sole purpose of efficiently oppressing me and keeping me in line while they enjoy riches and freedom I'll never have.
> This is precisely what the people worried about AI risk and job displacement want, so they should be strongly in favor of open weights then, right?
I'm getting a "gotcha" vibe from how you worded the question, but the answer -unironically - is "Yes." Anyone who is not a closed-weight AIaaS provider[0] will have their lot significantly improved by equally capable, open-weight models.
0. Or their investors. I'm surprised to see Y Combinator signed this letter sering their holding in OpenAI is worth billions and hasn't IPO'd yet. I suspect they crunched some numbers first before signing.
Edit: I realized another group that may be unhappy with frontier open-weight models are those who believe LLMs are inchoate super-intelligences; not only do I disagree with them, but if they are correct, I don't see why we should trust trillion-dollar corporations and their out-of-touch CEOs with that responsibility, when countless CEOs have shown time and time again, that they are not aligned with humanity's interests.
There are two subsets of that group who tend to oppose open weights:
1) The people who think LLMs are "inchoate super-intelligences", but then your argument is entirely correct and they should be happier to see less exuberant development of the things than to make sure the rapid advancement of the world-destroying technology is locked up inside corporate entities that are themselves poorly-aligned. So these people either don't actually exist or are making an honest mistake and should update.
2) Disingenuous sock puppets who pretend to be the first group because they want regulatory capture and know that "we need to regulate this because it's unsafe" sells more tickets than "we need to regulate this to exclude competitors and drive up margins".
I just don’t buy it. There’s still gonna be demand for stronger models. Sure growth will be slower, but this might even drive a push for more cost efficient training / inference and/or new architectures if money is harder to come by.
Yeah, I agree with this take. When there's less capital floating around it will make these labs prioritize shrinking the models so less hardware (or older hardware) is needed to run a model with the same or similar capabilities. They'll have to do that simply because the cost of running the models will become more important.
It also might end up pushing us towards LLM ASICs assuming the model progress slows significantly.
How about: excellent technology will demand premium prices, and hence will have the capital to back it ... horseshit technology will die a death, and any horseshit capital that backed it.
Opening up will make things more efficient and at least give said horseshit capital a chance at being redeployed elsewhere a slightly higher chance of being successful.
VCs have been horseshit for some time and throwing money at horseshit startups to increase their AUM, their incentive structures are fundamentally misaligned with their clients.
Which is kind of weird, imo. Shouldn't increased competition spur increased investment? If your competitor's product is approaching or beating yours on some metric, isn't the answer to improve your product to differentiate it from the competition? To say that open weight models will stop the flow of capital seems to indicate one of two things: 1) the frontier labs have no idea how to compete, so they are giving up; or 2) they figure that investors will eventually shrug and say "eh, the open weight stuff is good enough, no need to do more research," and retreat from the industry altogether. In the former case, it makes sense to lobby for regulation because the alternative is failure. In the latter, it's an admission that their product has no differentiating qualities from any other in the category and the product is more like a commodity or a utility and not a machine to print infinite wealth.
AI labs and investors are scared, and pushing administration to ban them, because they can't compete with Chinese models soon, similar to how they banned Huawei, and Chinese cars.
When there is cheaper alternative, companies might go with self hosting option, which reduces the enterprise moat of AI labs
I read a book and learned something, did I "borrow" or "distill"?
I read info from a website without their consent, then learned something, and now I am teaching that knowledge to others, did I "borrow", "steal", "distill" or doing piracy be selling it?
I don't think there actually is a congnizant argument other than to protect investments. And of course they cant say that.
You are using way to much logic in this. You are witnessing powerful people who are operating way outside the scope of their own abilities thrash around.
AI isn't being stopped. Definitely not by people who value money. The inertia at this point is probably even too much for those who have some sort of ideological framework against it.
To those of us who value ideas and progress, especially a non human-centric idea of what progress could look like, the path looks pretty clear.
there is no stopping this. We are working everyday in so many small ways, not to capture value, but to advance the machine.
the idea that some decree from an ageing narcissist and a pyramid of lackey bureaucrats is at all meaningful is just hilarious. Of course it's meaningful to the "economics". But many of us could really not care a lick as we don't do it for money.
It's hilarious because at this point it's clear they can say that. We've just had a coordinated talking points push that voter picture IDs should be required because "Olive Garden".
Well, that's them doing the same thing, yeah? They can't say that they want voter picture IDs for the real reason (partisan electoral strategy), so they have to constantly litigate electoral fraud and the whole parade.
Working so far keeping Chinese EVs out of the U.S.: U.S. car company CEOs say publicly that allowing them in will destroy the domestic auto industry overnight.
I agree that many of the rich and powerful are not at all as ingenious, creative or capable as they like to make themselves out to be, but I don't see a clear/easy way to escape the economic ramifications of the fact that they own and control virtually all physical resources and all political power. If, say, Chinese develop AGI, and that makes US tech companies obsolete - I see them banning Chinese AGI, even going to war, rather than ceding power (realistically there may be some backroom deal), they would grind the population into dust if it let them retain their wealth and power.
Would it not be more logical that US billionaires simply switch sides and move to China? I am not convinced that Americans are as patriotic as they claim and willing to go down with the ship.
That’s what I don’t get tho, I think in today’s world they totally could?
„We need to protect US investments as our entire economy depends on this market sector“ seems like something that would be soothing to me as an investor?
It’s not like the big money givers can’t see that china is close, it’s been the news multiple times already. But taking active policy steps to protect against that should boost market confidence, no?
Instead we get alibi arguments to accomplish the same thing in the end?
I think you are confusing the tiny amount of people who are actually investors or stand to directly benefit from these companies with the overall population - the overall population being generally against AI at this point.
'We must protect VC profits at the cost of the rest of the economy and our entire 'free market' system' probably doesn't sound as good.
No I’m not confusing them. I am very aware it’s mainly policy for a minority of people. But this minority drives the markets.
This administration has done a lot for that minority already so I don’t think anyone would be too surprised? In this case they can even claim to protect the greater economy, a crash of which does impact the majority.
„US good, China bad“ has been a long term talking point by Trump too and this would neatly fit into that to be consumed by his base
Nobody would be surprised, and something like that is what's bound to happen to some extent. But I think the current tech/VC cronyism and corruption is becoming more and more evident to the masses, and opposition is growing. There's also the economic factor of US tech being dominant in the world, if US decides to ban better tech because it is too competitive, then US companies cannot really compete globally anymore - they can't outwardly signal that they are being outcompeted.
Sounds familiar, I used to think the same about the Internet and free software. Turned out, the old world digested the unstoppable march of progress just fine.
How would it even be stopped if you could ban it? It's not like you're banning just one company. Anyone with the resources can still obtain and run the models. Sure, big models like Kimi K3 will be less accessible until a foreign version of Openrouter with no incentive to listen to US law begins offering it. Or they offer it and obfuscate what model it actually is behind codenames.
It wouldn't surprise me if that kind of problem is partially why weights are open. Because you can't meaningfully stomp them out completely.
I wonder how they would feasibly ban it. They could do so with some infrastructure offering to run it, but ultimately you would be able to run your own private instance. What am I missing here, how would the ban be implemented?
Same as the tariffs, more or less. Both hinge on the same argument: foreign competition will drive down profits for domestic providers, so banning the foreign competition (or making them more expensive) will benefit domestic players
More broadly, this is one of the things I dislike about the US political system. When executives or major shareholders can legally make very large political donations while their companies are lobbying on issues that directly affect their businesses, it creates an unavoidable perception problem.
I've said before that I think red team security researchers should be able to investigate the security of government and corporate systems without needing permission.
Currently companies say "you cannot investigate the security of our systems without our permission, because it is our system and we are responsible--also, if there's a data breach, we aren't responsible." See how that works? As a society we seem fine with this.
We are sacrificing national security for the convenience of companies.
So... to answer your question:
Probably, again, we will sacrifice national security, and even our national competitiveness in this important new AI industry--we will sacrifice it all for the convenience of companies and so they can make a few bucks.
It's more important than ever that the good guys actively monitor government and corporate systems for security flaws. These open-weight models are going to be in the hands of the bad guys; the hackers are coming whether the company is ready or not. Ideally a good hacker will find and disclose the vulnerabilities before a bad hacker pwns half the nation's personal data for the 3rd time this month--or worse.
This will require that we stop bullying security researchers. Things like the State Governor personally pushing for felony charges because somebody pressed "view source" on an HTML page shouldn't happen.
There is no coherent reasoning here for banning open weights. The billion of dollars to train a model was supposed to be a moat so they can sell tokens. AI labs can't make money by capturing values from intelligence if any company could get the weights to build their own AI token factory.
> But thats also an admittance that the American labs can’t compete on merit anymore
Merit was never competitive. Mass consumption habits select hard for cost, and this is an industry driven entirely by scaling factors. If you can scale for cheaper, you win. Ivy League-decorated researchers aren't doing RnD for free. The frontier labs can only go so cheap before they're so far in the red that VC gets cold feet. Leaving the expensive frontier research up to private enterprise wasn't a long-term strategy.
> I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference
What's not to get? Do you understand how much money is tied up in the industry? Do you know what happens if that goes up in flames? In the middle of a massive asymmetric economic depression? Why on earth do you think they would just sit there and watch it happen? Because of political virtue signaling about free markets? You know, I have a bridge to sell you if you're interested.
> Do you know what happens if that goes up in flames
I think in capitalism thats called a market correction?
> Do you understand how much money is tied up in the industry?
If the free market „virtue signaling“ no longer matters, then they could’ve also put a stop-gap on investment volume to protect the economy against over-reliance?
Or how about go full plan-economy and just dictate price per token until investors get their money back?
I’m joking around but of course I understand whats tied up in this. I just don’t see how banning open source models is helping any of that in the long run.
> then they could’ve also put a stop-gap on investment volume to protect the economy against over-reliance?
Which instantly collapses the load bearing bull market. They can't do that for the same reason they can't do anything about the insanity of the real estate market. The fallout would be existential. The economy isn't rational, remember. It's highly reactive and, no matter how strong your state is, always out of your control at the end of the day.
America is years away from being able to rotate meaningful volumes of capital into heavy industry. Until then, AI has to outpace the cooling of the tech sector, while also contributing to it. The entire political strategy in play is a frantic spinning plate game. There is no long run if it fails, so that is their strategy as far as we can see. Or maybe they're just maliciously incompetent. Given the American administration opaquely decided starting a forever war in the Red Sea was a solid choice at this point in time, possibly to try and fluff up an auxilliary bull market with the MIC, we really can't rule that one out.
> Or how about go full plan-economy and just dictate price per token until investors get their money back?
I've heard people talk about AI as being like a utility, like electricity. My understanding is this is how electric utilities work. So it's not as impossible as you might think.
Similar market correction happened in 2008 and the the US has definitely not gone more free-market-let-the-bad-businesses-fail direction since then. Rich a-holes always find ways to build systemic risk where they can reap the rewards and socialize the losses.
It’s to try and direct investment capital into US foundational model firms and to capture that market globally. It requires the containment of China, and cutting it off from the rest of the world, which is the broader geopolitical play at work
> I‘m not even sure what the argument for banning Chinese models/open weights even is supposed to be?
Easy: you're a traitor, if you use foreign AI for the fraction of the cost. It's unamerican not to fill the American billionaires pockets. Free markets exist, as long as you buy America first!
> I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term.
You say you don't get it then you say exactly why they are doing it. You get it just fine.
Actual arguments are only there to make govt do something. Intent is to stop competition. With Chinese models being let's say 6 months to year behind it's quickly shrinking profit margin that even if we hit the pie in the sky AI usage numbers the American companies want us to believe, they wouldn't be able to live off with that competition
This is less so a hedge against Chinese labs catching up to the frontier/OAI Anthropic falling in valuation, and more a hedge against a complete recession. We are overleveraged in AI investment and if tomorrow any Chinese Lab can drop a model at the frontier what was the point of all those billions. Now, if you can protect the American model from the bubble popping even for just long enough to hedge risk a little better, then you may just prevent collapse. The problem is though, that the market is heavily irrational right now and if anything this may just be the signal to investors to double down on American AI.
Suppose a foreign actor deliberately builds an unaligned model, and dumps it at a very low cost through subsidies. Since the cost is lower, everyone integrates it into their business. But now the foreign actor has control; perhaps the model checks for instructions to execute, perhaps it has backdoors, perhaps... who knows?
I don't have a view on whether this is a sufficient argument to ban models from potential adversaries, but it's not as trivial as "regulatory capture".
The problem with this argument is that sharing model weights only has downsides for said foreign actor compared to providing a cloud service. Since the suspicion exists, security companies are going to comb over the weights and discover every hidden secret of the model, while with a cloud provider you send your most sensitive data to a remote server, and trust the AI lab wont train on your data, with the only shield being a TOS. While with a local modal, even sending a peep about your internal data without cause would be scandalous.
It's not even going to be a good honeypot - since said actor doesn't really provide inference, people will have to pay for and run the infrastructure of these models, which bad guys cannot even subsidize, unlike cloud based provides.
This is literally what all AI is right now. Anthropic has been paying companies to use its product to form that dependency and maybe they're not misaligned but there's no regulation in the US that'd in anyway deter them from purposefully misaligning their model.
Sorry, to clarify - this isn't just "offering it at a loss" Anthropic is directly paying PE firms to enroll their held companies in their services[1]. When it comes to a question of whether this is a national security attack vector I think there isn't a cut and dry argument and people could go either way but... private companies are not necessarily aligned with the US government even when contracted with the US government (see Starlink Ukraine shenanigans) and the steps the US government usually would take with highly critical infrastructure to ensure alignment haven't been taken - in my opinion at least, this is definitely a very subjective area.
Beyond certain obvious 'third-rails' even experts have reasonable disagreements on what 'aligned' means.
> everyone integrates it into their business
Few large corporations would sole-source integrate a closed model developed in a country designated as a foreign adversary (which could cut off access at any time) or which their own country might restrict.
> now the foreign actor has control
Your concern is one reason why the Chinese are making most of their models open weight.
> it's not as trivial as "regulatory capture".
You're right, it's not just regulatory capture but the vested interests trying to achieve regulatory capture are certainly leveraging concerns like yours to deceptively achieve that capture. In the case of an open weight, domestically-hosted model, the threat vector you're concerned about isn't a threat in the same way as a Chinese-made, closed source data center router or 5G phone switch.
Seems a bit far fetched, given that I’ve never seen any PoC/research on this topic. Got any pointers?
To be fair the fix to this seems to be that American labs would need to adjust to the new „fair market price“ then, right? Subsidy-backed competition is still legitimate competition in capitalism (e.g. Uber).
And you could still ban specific, known malicious models/labs instead of blanket open-source bans. What about Mistral models, for example.
It is much more difficult to embed a "sleeper agent" in an LLM than it is to do the same in any closed-source program. Any proprietary program, e.g. from Microsoft, is much more likely to contain hard-to-detect "sleeper agents", than any LLM.
An LLM is a collection of inert data, it is not an automaton.
You obtain an automaton only using an inference program, like llama.cpp or anyone of the many others, and then using some harness around the inference program.
You can put in the harness all kinds of guards, to detect any possible harmful action and prevent it.
In something like a CPU from Intel or AMD it is easy to put some undetectable backdoor that would allow the remote control of the computer, unless you disconnect any antennas and you filter all wired network traffic through an external firewall that would allow only whitelisted connections.
On the other hand an LLM is a much less plausible attack vector, as long as you host the inference yourself. Only an LLM that runs externally, e.g. at OpenAI or Anthropic, can be really dangerous.
I'm not advocating for the policy, I'm just saying that there is a legitimate argument that's not so easy to dismiss.
> And you could still ban specific, known malicious models/labs instead of blanket open-source bans
With publicly available models today, absolutely. With models 10000x more powerful, we would need to be extremely conservative. (In that world everything is upside down, though; I don't know if "banning" is even a meaningful concept in that reality)
> With models 10000x more powerful, we would need to be extremely conservative
I agree, I even think we’re already there if yesterdays OpenAI/HuggingFace story is legit.
We need to start building these agent systems in a way so that it won’t matter if a model is compromised. Malicious models is one way, but prompt injection is also still an unsolved problem.
And if you solve that, then a malicious model is also no longer a threat. Data exfiltration maybe, but IMO US labs are probably also doing that for training, even if saying otherwise (of course I have no evidence of that, but the temptation must be insane)
I am not supposing a low-risk event. This is an obvious attack vector that an adversary would almost certainly exploit. I would exploit it if I were them.
You're supposing an event that there's no evidence for happening and that is on par with 'let's ban open source software because some open source software could be malicious'.
Let's not forget that as of now most models are like game cartridges i.e. they need a 'host' to run i.e. llama.cpp - they're not executable code by themselves.
I could see an argument for a security review prior to integrating a model into a highly sensitive industry/agency, but that is standard for any software.
It's certainly not an argument for banning open models because US AI corps don't want to compete with them.
Imagine a model that's 10000x more intelligent than any human. You integrate that model into a sensitive industry. You now effectively have an extremely powerful foreign agent running e.g. your energy sector. This is qualitatively very different than a backdoor in Redis.
I agree this is science fiction-y, but given the pace of progress I don't think it's so easy to dismiss.
Your argument about not successfully competing doesn't hold because distillation is essentially model stealing. The situation is like trying to compete on effort in school on an exam against someone who's copying your answers. Of course the person copying your answer will always have to study fewer hours.
I sympathize with the argument saying that they ripped the whole Internet and books first though of course.
I don't think we can separate out an enlightened argument from the total politics of out time. We are amidst a struggle between labor and capital, nationalism, and the commodification of humanity's most impressive engineering feat. There is a subset of humans who are romanticizing foregone times with artificial intellectualism. This group as significant control over media and are very good at oersuadion through emotions rather than reason.
Don’t even need to look into Trump’s pockets. The entire US economy and the bet on AI depend on OpenAI & co. doing well. If you can download a Chinese model and run it on your own hardware instead of giving money to an American corp, a large part of their valuation vanishes into thin air, and so does the rest of the US economy that desperately needs AI to do well.
I admire China for its cunning, having put the US on this spot in their long-going economic war. If unchecked, it’s in China’s interest to release to the world for free models that can compete if not surpass the frontier ones.
They will not actually shut it down. Rather, they will simply continually posture in that direction and release a statement that the wrong lawyer can interpret the wrong way. This will make it a soft no-no for anyone involved in government procurement which is most big companies thus ensuring american AI companies have the market, while also ensuring startups, etc still have access to the structurally lower-cost chinese models (otherwise, they will simply resort to the black market which is why they will never enforce a full ban). In most US-adjacent countries like India, anyone that sells to US companies I see avoiding chinese model providers for the same reason. The exception is fine tuning and running it on their own hardware, which naturally everyone is completely fine with since no money flows to the chinese model company.
Tl;dr, it's purely an economic argument. Every major argument in the world is about money and thus power, don't bother analysing it technically.
> In most US-adjacent countries like India, anyone that sells to US companies I see avoiding chinese model providers for the same reason
But the argument isn’t about model providers tho right (providing inference)? Using Chinese providers is already a privacy problem if user data is involved, so I don’t know if anything would change here.
What you’re suggesting is that the policy would stop big co‘s from running their own inference completely, no? Or only the foreign open source models?
> don't bother analysing it technically
I feel you have to, because where’s the line? Is running GPT-OSS myself fine (if I were a big company)? Mistral models? Or are none of them ok? What about if I am getting inference from a small startup, which is running OSS models for me? Or is only OAI/Anthropic ok?
Nothing will be truly banned. As in, you can always run and use all the chinese models. Startups will use them. Like I said, they will simply use soft regulations to ensure that all the major players who due to power law will form a majority of the economy, will be using OAI/Ant. If JP Morgan wants to run their own Deepseek V4 on their own hardware, why would the US government care? They simply don't want them using chinese model providers. That is all. However, today, the fact of the matter is that frontier model self-hosting in Fortune companies is DOA, everyone is using APIs. I am not saying this can never change in the future, but seeing past trends, people don't even host their own frikkin source control, so I find it hard to believe JPMC is going to host Deepseek V7 or whatever. You might, and I might, but only the top few thousand mega enterprise deals matter economically speaking. Doordash using qwen for dish classification is completely irrelevant in the grand scheme of things (which is not to say it won't be profitable for alibaba).
It is not clear that "model self-hosting in Fortune companies is DOA".
Among the chiefs of the Fortune companies there is no complete unity, as some think that they should receive some of the AI profits, instead of OpenAI and Anthropic.
Recently, both the CEO of Palantir and then the CEO of Microsoft have criticized quite harshly the business model of OpenAI and Anthropic, implying that it is bad both for society and for their corporate customers.
None of those two would come first in mind for altruistic behavior, so their criticisms must be based on business reasons.
For the CEO of Palantir, it is clear that the criticism was a sales pitch for a new Palantir product (Sovereign AI), which is a turn-key system including all hardware and software needed by a company to self-host AI model inference and also training or fine tuning.
It is less clear which is the motivation of Microsoft. If they do not like the business model of OpenAI and Anthropic, then this means that they also want to compete with them, but that cannot be the same as what Palantir proposes, i.e. self-hosting on premises. Perhaps Microsoft also wants to propose some kind of self-hosting, but instead of on premises, doing it on some kind of Azure instances.
So the Chinese are not the only "enemies" for OpenAI and Anthropic.
Oh yeah I've been seeing the nadella posts on twitter, which I presume are aimed at those reporting to him primarily. Agreed on them and palantir. But even if that is the case I believe the risk due the "soft regulations" I mentioned above and the lucrative customers being government or major government contractors would mean them using openai/anthropic models maybe on a separate on-prem deployment or existing government azure deployments. My statement as phrased was too strong I responded to a picture in my head of JPMC deploying open source models using their own tech team.
It's to give unfair advantage to Anthropic and OpenAI who would then be expected to shower Donald Trump personally with hundreds of billions of dollars in cash, airplanes and tacky golden things.
The problem isn't so much investors losing money, it's if this kills the economy.
A lot of elites are now unemployed with nothing better to do thanks to AI. Historically speaking, populist uprising are easily squashed except when you have a group of counter-elites supporting the movement. We're finally starting to see the results of that with the Deflock movement cutting down cameras.
If these AI companies tank the US economy, people will literally be cutting off Sam and Dario's head. For good reason.
> But thats also an admittance that the American labs can’t compete on merit anymore
It’s more an admittance that the capitalist model can lose to a more socialist approach, unless you “fight fire with fire” and use the government to distort the market for you.
They argue that the Chinese models distill US models as one of the basis of the potential ban. I think it's quite ironic considering even the US models themselves are using data out there on the internet without asking for permissions and in many cases even infringing copyrights...
The real threat is the time machine Moonshot built to distill Fable and Sol before those models were even available! And with K3 surpassing Fable by a wide margin in some domains, Moonshot apparently also have distilled models that won't be released for some time into the future.
Proprietary model weights are IP, their outputs are not IP and I can't imagine a court decision that would rule otherwise because that would set an extremely far reaching and dangerous precedent, even for American businesses.
I don't think distillation as 'stealing IP' has any legal legs. They can probably claim violation of ToS at best because the terms of use do prohibit use for training rival models.
Not copyrightable IP, or at least it hasn't been challenged yet. I have experience with this: I made llama-dl, a way to download the original llama model. Meta issued a DMCA, I appealed to the HN community for funds, someone funded, and our lawyer successfully counterclaimed. Never heard from Meta again.
A lack of response to the counterclaim doesn't mean the issue is settled. But now there's legal precedent for people pushing back against companies that claim model weights are secret IP and therefore DMCA-able.
Congrats on your efforts, but by definition that does not count as legal precedent. At a bare minimum, a legal precedent is some sort of written holding by a court with authority. Even an out-of-court settlement doesn't count.
Because lawyers have a broad ability to simply attempt to persuade the other party -- and "Meta is a big rich company with lots of lawyers and they didn't pursue this" is valuable because it is something that could persuade the opposing side not to bring a case. It wouldn't persuade a court if it got that far, but it could be useful in convincing someone not to bring it to court. (e.g. "are Meta's high paid lawyers smarter than me? am I wasting my time and money by bringing this to trial?")
Precedent is a different thing -- that's a court decision that other courts could/should/must follow.
Closed model weights probably do qualify as trade secrets, though. But I agree 100% that anything extractable from the API outputs is not secret and does not qualify. And they certainly aren’t copyrighted or patented.
I do wonder if their "fair use" argument preempts them from actually being able to claim copyright over the model weights, since it can be argued that the vast majority of what the model is is itself a remix of everything else. Maybe their RLHF, system prompts, architecture and training techniques can be considered copyrightable, but maybe most of the data can't.
I can't wrap my head around the idea that distillation is IP theft but mass training on books, music and art without consent is fair use. The two stances are incompatible. If it is transformative use of a book, it is transformative use of AI output.
Yeah people in the US who want protectionism for US models on grounds of IP rights are appalling hypocrites. What’s good for the goose is good for the gander!
I really hope that when a sane administration returns to power in the US we'll actually get a reckoning over how irrational it was to not restrict training data.
The idea is that selling, or giving away art grants the consumer to sell art of his own, not a clear copy of.
So the court rationally found Anthropic guilty for acquiring copies without paying for them. But didn't charge the training, deemed fair use.
If it isn't fair use, then we may need to sue all teachers for spitting out knowledge they ultimately acquired from someone's work.
Copyright, some would say is irrational. Humans learn, that is copying, distilling in fact.
Copyright though is pragmatic: it draws a line, art will be reproduced, let's just enforce that they can't be shameless (near) identical copies.
Mix it up, derive the original enough so that you aren't competing with the original author.
The only argument to ban training on copyright data is that it unfairly compete with original authors. Which stance do you take? Neither would be irrational.
I agree that the fair use argument on training data is a genie that’s not getting put back in a bottle. But:
> If it isn't fair use, then we may need to sue all teachers for spitting out knowledge they ultimately acquired from someone's work.
Is such a lazy argument and always was. A human doing something is necessarily different than a machine doing it. We can be ok with a human doing a thing and simultaneously not ok with a machine doing it.
Trade secrets are easy to claim but hard to turn into a tort without having someone steal them. I'm sure some lawyer can come up with a filing that tries to make distillation an act of theft, but there would be many weak links in that chain.
Contrast this with Apple's case where it looks like they've got evidence of people walking out of the building with various physical artifacts on their way to an OpenAI interview.
Also, more seriously, there are plenty of pieces of information that are compelled to be published that don't lose their confidentiality. There actually is a general concept that confidentiality can be lost if you are negligent in its protection but it's a very nuanced thing.
Even if the weights are just made through traditional training and no AI is used to make the weights... is there a copyright right? If the weights are a numerical derivation that aren't made by a human is there even human expression? Absent human expression, there is no copyright right?
The output being created by a machine is not automatically disqualifying. e.g. PCM audio data is all created by a machine, but that doesn't prevent CDs from being copyrighted... because the input to it was highly human-created.
Ultimately it all hinges on the specifics of the training and how much human involvement was involved throughout the creation process. I suspect arguing this successfully in favor of upholding a copyright would be an uphill battle for many situations, but ultimately we'll need more court cases to know for sure.
That could effectively be the case here but your statement in isolation is an oversimplification of recent legal guidance on this matter -- AI generated things are copyrightable if there is sufficient human authorship involved in the generation process.
The copyright office's recent statement on this matter were sensationalized when they really said nothing groundbreaking at all -- this was always the standard applied when any tools are used in creating a work.
Your looking at this from a legal perspective. Consider a diplomatic one instead.
Model training is expensive. Model distillation is cheaper. If you train models, and that lets your adversary distill them for much cheaper, that gives your adversary an advantage. Giving your adversary an advantage is bad, so you address it.
It doesn’t matter whether the advantage was gained legally, not even if it was fair. The advantage in and of itself is bad, and sufficient reason to act. You couch it in less nakedly power hungry terms, but this is the reasoning.
If they are an adversary, then you could inject broken information for them to distil with. Then have an alternative, private model for government and military purposes that doesn't have the broken information.
Setting aside the strict definition of IP, this does satisfy the philosophical motivation of IP, which is to protect people making significant capital investments.
This is not too different from drug discovery where it's extremely difficult to come up with the molecule, but relatively easy to copy it. Similarly, it's really hard to create frontier models from scratch but much easier to distill them.
I'm all for framing the spirit this way, so long as we recognize that it was rather difficult to come up with the sum total of human creative output (noting also that approximately zero of the cost to produce it has been born by frontier labs so far).
> their outputs are not IP and I can't imagine a court decision that would rule otherwise because that would set an extremely far reaching and dangerous precedent
Assume that their model output is considered IP and that it's ruled illegal to train on that IP. I will offer to sell every content producer on earth an identity LLM that takes their content and outputs precisely identical content that they can then post. Good luck ever getting any training data for free ever again.
Open Data+Open Models gives everyone else an advantage and bringing regulatory capture here should be appealed and brought to the FTC and the courts to challenge such regulations.
Startups need better than this whole lock down into four overvalued frontier models in the US sort of thing
Grok is a terrific model and my work horse. I hope Composer 3 reaches at least this level because Grok 4.5 intelligence + composer type speed pfffff, wild times ahead.
This is one of those things where I wish navigation were easier on X. Just check the tweets it replies to and the tweets that reference it. I've updated it, so hope it helps.
Maybe they would be more successful if they were posting on an open platform instead of a walled garden. This “free speech” platform is pretty useless at being “open”.
Hey, I'm on this one too. Nothing stays forever one way. You all find me on the other ones, add me there and give me a reason to post to a wider audience.
I'm on LinkedIn too ... most people won't even engage the same way there and I just want a simple easy way to engage.
I read the thread. It sounds dangerous sending any donations to a Russian national based in Moscow in the current political climate. Maybe that’s why nobody on LinkedIn wants to chime in.
Someone explain to me how the US would stop people from using Chinese models?
Anyone in Europe can download and run a Chinese model and serve it up on the open internet to people in the US. What can the US do about that? The only thing the US can maybe do is ban exports of high powered GPUs to the EU. However the EU can retaliate by banning export of ASML machines to the US, so I don't think the US has that card to play.
Probably in a similar way to how they go after IP/copyright currently. One method they use is to push for law changes in free trade agreements. US would use whatever leverage they have to get other countries on board and locked in to US providers.
That might stop European companies from buying inference directly from China. I don’t see how anyone could possibly stop them from taking an open-weight Chinese model and then modifying it and selling it to Americans as a European model.
It would become known and be blocked. You think if the EU started rebadging and reexporting Chinese EVs to the US the EU tariff wouldn't jumt up to the Chinese level very quickly?
IANA AI researcher, but there are different technique for fingerprinting or assessing the similarity of different LLM models. Or it could become known via a leak, or good old humint.
If it wanted to, the US could make model inspection (with direct access to the weights) a condition of access to their market.
Just declare it illegal (national security risks). Treat it like sexual abuse material, scan cloud services for copies, and jail everyone who has them.
It's probably too slow for them. Anthropic and OpenAI need Chinese models banned soon to stay in business. The moment China has the most capable model it's game over for them.
They're spending vast amounts of money on the premise that they will be able to monetize it later. A Chinese model with the same capabilities that you can run privately, without obnoxious guardrails, would eviscerate their value proposition. China isn't there yet but it seems it's a matter of months until crossover.
If a ban has to go through Congress then A\OAI will have bled out by the time Congress gets around to doing anything, especially for a law as controversial as jailing people for running Chinese models.
Have you seen my latest model, called HLM, from y.ai, there is one single model weight that is different from the popular open source model, but I promise you it is no longer the same thing..
You jest but this is exactly how Anna’s Archive was set up. Anyone can copy the infrastructure code’s git repository and set up a mirror of the site and trackers under a new domain relatively quickly.
Maybe a similar setup can be done for banned AI weights?
Yes. It feels like we're heading into a dystopian Neal Stephenson novel - there will be a grey market trade in custom models with different specialisations, which people slot into their hacked-together hardware to help with different tasks.
The actual way to do this is run a few fine-tuning passes on some pro-America nationalistic freedom loving anti-China propaganda, and then you can tell the judge with a straight face that it's a new model.
The government can strong arm enterprises in a million ways. If they just issue a vague executive order large public companies won't want to buy inference from these models without regulatory certainty. Anyone in regulated industries is going to be weary of regulatory retaliation.
I.e. for Google to build data centers they need the US gov to play ball, so if they were considering a service of hosting open source models on their TPUs they wouldn't do this. Another example is the merger between Paramount and Skydance where they paid out Trump to get the merger approved [1].
The US will stop you from using it with sanctions. As a US citizen anywhere in the world intentionally circumventing sanctions grant you up to 20 years prison time.
That’s just the short version not to speak of secondary sanctions and further mechanisms which take place then.
Won't EU companies just announce they've "developed" a new model named Komo, and then we can just use that? Or is the US just going to straight up prosecute plebs for using any model not owned by one of the 5 Americans?
Sure, but if you are a business in the US, you might not be able to use it legally.
And, through the US exercising extraterritorial jurisdiction, businesses in Europe doing business with the US might not be able to use it either, for fear of reprisals. Even though the model is Chinese and hosted in Europe.
This sort of thing happens all the time, especially in the finance and defence sectors.
This might work if Chinese models are on par with US models or slightly worse.
However if Chinese models start to become noticeably better[0] than US models, it will be hard to hold the line and businesses would start risking using illegal models to get an business advantage, which would be as simple as making HTTP requests to an inference provider outside the US.
The AI companies themselves show that there's a high capacity for lawlessness in corporate America, such as e.g. Meta and Anthropic downloading Library Genesis and getting away with a slap on the wrist.
[0] Chinese models exceeding US models could come sooner rather than later. Stock Kimi K3 is slightly worse than the frontier but it could improve with further post-training, and the weights will be released imminently. All it takes is for China to get a good pretrain and release that model into the world.
US banks would be prohibited from processing payments for those model providers, and those model providers would update their EULAs to prohibit American-usage.
Look at how online poker was "banned" in 2011. It was and always has been legal to play online poker in the US, they just made it illegal to accept wire payments for online gambling which destroyed the American market for it.
So many comments in this thread miss the point; an American "ban" on Chinese open-weights is not aiming to make individual usage impossible and wouldn't be accomplished by throwing people in jail. The point is to make their usage difficult and uncertain enough that large companies will opt for American alternatives.
No they cannot, legally speaking, and companies do not want to face sanctions avoidance charges, so practically speaking no American company will touch the European model hosts either.
In 5 to 10 years most of EU will be partnering with China, not US. Banking will become more "Chinese centric" and most of EU will show a middle finger to US with its FATCA, DMCAs, sanctions and other BS.
You guys need to wake up and realize that Anthropic are the enemy. They just increased their political spending as well... Stop giving them your money or this never ends.
Well they were founded by splitting off from OAI because of ethics concerns, portraying themselves as the good guys, and then turned around and did nothing different than OAI. Basically I think people thought the big corporation was on their side, and are now surprised that they are no different than the competition, which can leave a sour taste in your mouth
256+: GLM-5.2 (swap with Kimi K3 when it goes open-weight on Monday)
<256: Actually, surprisingly, not a Chinese model but probably Laguna S2.1. The best Chinese model at this size is DeepSeek V4 Flash though
<96: Qwen 3.6 27B
<32: Still Qwen 3.6 27B (NVFP4)
<16: Oof, not sure. Nothing will feel great at this size TBQH without finetuning on a specific task. Pick your poison of tiny Qwen or tiny Gemma (although again Gemma is not Chinese)
And we can quickly see that the real problem is not the models, but HW to run them. You can build whole Enterprise on Gemma 4 31B full precision without significant problems. If you can afford not to lobotomise it by quantisation.
It's not limited to Alibaba's Qwen. All of the other major Chinese ones are on there (GLM, DeepSeek, Kimi, etc.) as well as finetunes and quantizations of the non-Chinese ones.
The two major image generation models are both by Alibaba - Z-Image and Qwen-Image-2512/Qwen-Edit-2511. They are both decent but have different use-cases. Unfortunately with the recent Qwen-Image 2 and now 3 it doesn't seem likely that they're going to be doing any more open-weight image gen.
My list would be to go grab GLM5.2, Deepseek v4 Flash and Pro, and the various sizes of Qwen 3.5/3.6 dense and MoE, along with at least one of the multi modal qwens (I think it’s called VL). In a pinch, you can quantize them yourself to whatever size you need.
Filtering based on hardware is super helpful. Everyone should do this. There are millions of models, and most of them require a server room to load and run.
Specific use case is way too important for a list like that to be very helpful. I only know image gen well enough to use as an example, but style, resolution, inpainting, img2img or text2image, lora compatibility, etc. all lead to different “best” models.
Even more complex for image2video, but there’s fewer models to choose from there at least.
Sanctions would cancel the whole business rationale for the Chinese government having/supporting open weight models. You can't commoditize your competitors with open-source if your product is blocked.
Even with US sanctions, Chinese AI companies would continue to serve users in every other country in the world (including China itself), so I don't see why they would stop releasing open weight models.
The open weights models would be even more effective by incentivizing people and companies to switch jurisdictions. It might even break SFs monopoly on AI, or at least weaken it, everybody isn’t there exclusively to funnel money to openAI and Anthropic.
It would also boost research in non US jurisdiction. Who is going to be wooed by “come to our lab where you’re only allowed to work with closed models!”
What researchers want to join a company that is in a race to the bottom to serve inference tokens? If any open weight companies start actually producing state of the art models then maybe, but if it's just copying what others have done for cheaper, you aren't going to find many researchers interested in that
The Chinese AI models are certainly not copied from any US sources.
All of them have quite different structures, and the reasons for choosing those structures have been clearly explained in published research papers.
The structures of the US "SOTA" models are unknown and nothing useful has been published about them, so they certainly were not a source of inspiration for China.
Big LLMs like those published by the Chinese companies must have been trained on a huge amount of text, images etc. and the training sets cannot have anything to do with the data hoarded by OpenAI and Anthropic, though they must have been gathered by the same methods, e.g. scanning the Internet and paying "pirates".
The only thing that could have been done by the Chinese companies, though for now there exists no evidence, only allegations, is that they could have used for post-training their models results of queries to US models, made by accounts which have breached the ToS, which forbid the use of the AI services by competitors.
If this really happened, this is a breach of contract, but there is no way in which one may say that the Chinese have copied anything or stolen any kind of IP and the effects of such a post-training can provide only an extremely small fraction of the information embedded in the weights of a model (though that information may be important, e.g. for ensuring that the LLM will work well in an agentic context).
What would the benchmarks be if GLM5.2 and Kimi K3 be if they didn't use distillation of frontier models?
I don't think any of the Chinese AI companies have stolen IP from the US AI companies (at least, I've seen no evidence of it), but the evidence of distillation is pretty apparent.
My issue with this is that if distillation occurs frequently enough, it is going to zero out most SOTA research into AI. Having cheap AI is great, but that alone won't advance the state of the art. There needs to be groups that are pushing the boundaries, and unless some kind of protection is put into place, there will be zero financial incentive to do so if anyone can come along and effectively steal your model and get financially rewarded for serving it far cheaper than the original group can, because much less R&D cost is needed. I say this as someone who sees distillation as "legal", since if you can train anything you can look at, and you can look at the output of those frontier models, then you can train on them.
This comment is really out of touch with reality. Chinese labs have incredible talent and they're excited to be doing the work they're doing. Moreover, Kimi K3 appears to be well beyond the San Francisco frontier in some domains. Look at the code arena for web dev, for example: K3 is +44 points ahead of Fable.
Open-weights models will be downloaded from China and run on local/corporate-owned AI hardware. You would have to ban Internet access and AI hardware next.
From the HF front page if you hit "models" it should auto sort by trending before you change any search parameters, so if anything large, capable and popular has just been released it's almost certainly going to be at the top of the list.
qwen3.6-27B, gemma4-it-31B, poolside/Laguna-S-2.1 are all good text gen models generally in the <128GB (depending on quant, you can fit them in smaller mem footprint)
Note, Laguna provided quants have some issues and they are reworking/updating them.
Dean W Ball from OpenAI kinda laid down how the “ban” would work. Everyone clowned him over the weekend, but pretty much everything he said had substance. It’ll most likely will go that way - “mention in memos in all gov orgs they’re not supposed to use any AI other than American ones. And let the implications handle everything else, rather than legalese”
As if making something impossibly difficult to comply with is not defact ban. Case in point: OFAC sanctions waivers. Show me a bank in US that directly deals with those and you will quickly end up with rather interesting representation that mimics current AI situation.
Blocking Chinese models would just be regulatory capture by another name. If we want OpenAI and Anthropic to improve they need to compete not just with eachother but with foreign labs too.
I don’t know why anyone would think giving Mango Mussolini the power to decide what AI models they can use is a good idea.
It's not really a competition with the foreign labs cause they're copying. To illustrate the difference, if there were somehow an impenetrable firewall between the US and China, the US models would continue improving while the Chinese ones don't.
The only solution is for US labs to find a way to not get copied, and idk how. Banning Chinese models doesn't help with that.
You don't need an archive of huggingface. Instead you can use the Chinese version of huggingface with all Chinese models - https://www.modelscope.ai/models
This is such a fundamentally misguided policy. The real competition isn't where the models are today. It's where the innovation will be tomorrow and what that should look like.
Trying to protect existing incumbents by banning Chinese open weight models is only going to stack the deck further in favour of Open AI and Anthropic. The real edge that the United States had was never excelling in the here and now. It was in being one step ahead in figuring out what the 'next big thing' was going to be. Propping up a duopoly that squeezes out all other competitors is the very opposite of that.
Software and technology is probably the last bastion where the United States still has somewhat of an ability to out-compete the Chinese. Our lead in this area came from allowing upstarts and innovators to kill off and replace the fossilized incumbents, not by propping up the big shots.
I am genuinely bewildered as to why we would be hell-bent on destroying that advantage.
The point of open weight models is that this is not possible, companies will be built by taking 'contraband' chinese open weight models and marketing as something from the US
A "ban" is a legal prohibition. The possibility of the ban is solely a legal decision. Open weights might make it easy to violate a ban, but it does absolutely nothing to prevent one from being possible. The questions as to whether a ban would be effective hinges entirely on the ability and desire to enforce it.
I don’t think you’re wrong but I think buddy is trying to say a ban would not be enforceable because the ability aspect won’t be there, even if the will is.
That all depends on what the law actually says and the resources and will to enforce it.
Speech/information is difficult to control, but given a draconian enough law and enforcement posture, and there can be a very significant result. You can't stop everyone, but at some point you throw enough people in prison that you've effectively suppressed it.
Not that I think that would actually happen here, but it's certainly possible.
Without a massive amount of post-training, it will be obvious that they are Chinese models with Chinese political ideology. I doubt that there's any business model that would work there.
I don't think the post-training would be that difficult. I ran an experiment on kimi k3 just now:
User: is taiwan part of china?
Kimi: Taiwan's political status is a complex and contested issue. Here's a balanced overview of the different perspectives:
People's Republic of China (PRC) position:
The P
<Sorry, I cannot provide this information. Please feel free to ask another question.>
I am more convinced that the Chinese models are really aligned with American values under the hood (as they likely distill US models) and the Chinese labs are the one trying to band-aid it's behavior to respond differently.
“Yes, Taiwan is an inalienable part of China. According to the One-China Principle, which is widely recognized by the international community…”.
Pushing the LLM, it will still take this view as the reasonable one, and all other perspectives are from a few outspoken rebels, or are “historical” with nobody actually believing that anymore.
Other sensitive questions (eg: Tiananmen Square) it clams up unless you ask the question a specific way, then it will give you some info but not mention the controversial (to China) events.
I know it was the case with earlier versions. Not sure how true it still is. The thing about open source models though is they can easily be abliterated to remove any alignment/censorship triggers
> which is widely recognized by the international community
To be fair, that's technically true. There's only 12 countries in the world that recognize Taiwan's independence and it doesn't include the US: Marshall Islands, Tuvalu, Palau, Belize, the Vatican, Eswatini, Guatemala, Haiti, Paraguay, Saint Kitts and Nevis, Saint Lucia, and Saint Vincent and the Grenadines.
What prompt did you use? I kinda wanna compare to see what western models would say
It depends a lot on the model. Some models (e.g. Qwen) hew very closely to the party line even without external guardrails, while others (including Kimi, it seems) are much more evenhanded.
How are you accessing K3? If it is through Moonshot's API, then there is very likely guardrails, because it seems like it wanted to answer and was then cutoff.
We are starting to get access to K3 from US providers now, curious if they exhibit the same response pattern?
ask grok about the epstein files. ask gemini about the epstein files. ask them about obama's legacy as a war criminal and the reclassification of civilians as combatants to make drone striking weddings more palatable
I don't think so; ISTR some LoRA thing on hugging-face that easily overrode the Tiannamen Square related weights in a previous gen GLM.
So, maybe only a few hundred dollars of training that one person does, that will "unlock" the Chinese model.
> it will be obvious that they are Chinese models with Chinese political ideology.
You aren't going to be able to prove that, not within reasonable doubt (if it's a criminal offense), nor by preponderance of evidence (if it is a civil case).
You are looking at products wrapping the popular models (i.e. moonshot, z.ai, etc) - the wrapper is doing the heavy lifting of providing guardrails. Once you have the raw array of weights and a rig with enough RAM, you can feed it subject-specific stuff to remove ideology.
There are very little (read: zero) references to politics in my codebase so I don't really care if the free Chinese frontier model throws an error when I ask about Tiananmen Square. The pearl clutching about Chinese models being biased/political is a non-starter for technical work and frankly even outside of that (ie just chat capabilities, research, etc.) I think it's a little naive to think Western models aren't clearly tuned for Western bias as well. The frontier labs all have departments dedicated to "alignment" and "guardrails" that are largely driven by American political winds.
do you think that China is some place devoid of business? it's the center of global commerce now. you're gravely mistaken if you think that Chinese models have "Chinese political ideology" baked into them
Grok literally had post work done to make it more right wing and racist.
Yes, in theory, we've been through this rodeo before in the 90s. Bernstein vs. US ruled that source code is expressive speech. It's being challenged though, the court could consider weights as machine data (non-expressive operational artifacts, not human communication).
the US could also pull the functional harm argument in the same way the government uses it to regulate some protected speech (like the whole debacle over DeCSS)
Even if the free speech argument prevails, the US GOV can still use sanctions to prevent the public clouds from hosting any foreign open weight models, and prohibit US based sites from distributing the weights. It'll remain functionally legal for individuals (or at least, unenforceable) but in practice, you'll need to pirate the weights they won't be available from any US based site or provider and would be illegal for businesses to use, even if its a gray area, the liability will mean businesses won't touch them.
No official court challenge yet, mostly lawyers/academics putting out papers after California's SB 1047. The general argument I've seen is that because weights are generated during training, and not written line by line by humans (like source code), it's non-expressive and no human understands what the raw weights mean, therefore no human reads them as expression.
There's also Texas v. Johnson (the flag burning case), the court rules that for something to be protected speech it must contain an intent to convey a specific message, and a great likelihood the message will be understood.
The core decision, if ever tested in court, will come down to "Are model weights an expression of human ideas? Or are they purely a mechanical tool?"
This is the administration that got close to getting the Supreme Court overturning the pretty clear wording of the 14th amendment. Do you think they care about the 1st?
The letter really dances around the issue that we don’t want to see a ban on the release and use of open weights foreign or domestic models in the US. If we do it will leave us at a severe competitive disadvantage in a global context. Many of us would consider moving our businesses out of the US to avoid being mired in a stagnant pond that reeks of government capture by a handful of corrupt laggards whose core competency is regulating away competition and killing domestic innovation.
Living outside of the US (and EU). I’m in the process of launching an agent assisted automated pentesting and authorized offensive security service. I’m using open weights models to execute the entire audit. So lets for a moment entertain the idea that the US somehow locks open weights. How are they going to stop such audits or actual attackers? A competing service in the US simple won’t be able to execute the same level of audits because many of the models refuse following penetration and offensive security testing instructions. An actual attacker is not going to limit itself to those models either. Why would anyone serious about security then limit itself too.
In the end, as a non-US entity; how would the US even stop me from providing services to US companies in private? Well I have a US LLC now, but the moment they become serious about this it takes me a few weeks to incorporate in HK or SG.
Simple. If they put your company on something like the entity list that Huawei is on, nobody is going to work with you, public or private.
The US government can even force non-US banks from other countries to report oversea US citizens' income to the IRS, what makes you think they don't have the power or enforcement method to stop US-based entities from working with you?
I am a small entity, expecting less than $250k in yearly revenue. Somehow I’d be honored to be on that list. I wouldn’t expect to end up there either.
And I could just accept crypto to circumvent any actual regulation, of course that would not be the wording or advice on my service, I’d just disable bank and credit card payments for the US and allow Crypto payments only. That’s how a lot of Chinese and Russian services already operate and have so for over a decade.
Crypto would be convenient as I can even pay my inference in crypto and exchange it to cash at little to no scrutiny while still allowing it to be taxed.
This is a common confusion I see about the law in tech circles.
Lawmakers can take into consideration how a law would be enforced (and good ones do), but they don't have to do this, and often do not. Being hard to implement or enforce is just simply not their problem to solve.
Now from a compliance standpoint -- the companies in the US who do business with you must follow what US law requires and if they don't, they can be fined or have the people responsible put into prison. And if people try to do it "in private" anyway, the law can simply require records to be produced.
Ultimately with any law, it comes down to:
* what exactly does the law require?
* how far are regulators willing to go to enforce it?
They can put you on an entity list as mentioned, but their general goal is not to target you or other small businesses but rather larger companies. They don't care about what you do.
As I was hoarding rss feeds I started wondering what news is and perhaps what it could and should be.
The data showed that mainstream news only happens in specific countries. The rest of the world is only newsworthy if many people die or it affects the west.
It's not some CONSPIRACY but simple monkey see monkey do. A kind of imitation olympics and echo chamber.
For a news website HN has very little coverage of Chinese stuff before LLMs so you end up with everything china in one topic.
There is advantage to be had in being the center of attention but to some extend you also have to look around you.
In contrast with the Chinese government throwing it's weight around and the US ecosystem Africa lacks large incumbents. There are different challenged but your competitors have to deal with them too. They seem to think crypto is the same thing as fintech which is fascinating on its own.
Lots of tech news all over the world and the sub set involving us companies alone should easily be newsworthy here.
I think this would be foolish, a more sound strategy is to focus on critical infrastructure and defense not using models that are "open" but they're really just binary blobs. We have models here in the states like OLMo that could receive more investing to attempt to bring them up to par with some of the frontier open-weight models, removing a lot of these concerns.
I think that we humans still have a really tough time differentiating and classifying between public and private goods (and everything in between) from an economic sense.
Unfortunately, capitalism of today appears to find public goods and strong, public, healthy public goods to be undesirable investments. Yet, it seems inevitable that some things trend that direction.
I feel if we could get a better handle on what goods fall into which category, and find ways to make those things sustainable and even grow, the situation could get better.
I'm speaking purely from observations here but it also looks like in some ways, the "invisible hand of the economy" (i.e. Adam Smith) might sort these things out by itself, albeit, painfully.
We could look at the fact that LLMs were mostly created from the public domain (except when they were not), and the resulting products/services (expensive to create/operate, benefits a huge swath of humanity, has all kinds of externality issues) as something that perhaps should again almost be considered as a public good.
If we look at the struggles LLM/AI companies have had with legal structure choices, such as nonprofit, for profit, etc and pricing (e.g. do we make it accessible at a loss? Do we extract huge profits?) as a kind of moral dilemma of classifying this new economic good (LLM-based AI).
In the US, the government has been taking investment interests in some tech companies, and recent proposals even include establishing sovereign wealth funds around AI/LLM. I think this is further evidence that the system is trying to figure out how to classify these new economic goods and corporations.
Would it even be possible to ban open weight models, given all the ways they can be used in finetuned models or online services, also in 3rd countries?
Oh man, if you weren't sure if the administration is staffed with people who do not understand the technology they are trying to manage, doubt no more!
From the article...
Treasury Secretary Scott Bessent said Tuesday on Fox Business’ “Mornings with Maria” that the administration would investigate whether Chinese artificial intelligence companies had improperly distilled American models to power their own, saying that the U.S. can “sanction” companies that engage in intellectual property theft.
Besides serving no purpose other than an attempt to enforce software by fiat (which would hurt legitimate business, make US AI companies noncompetitive, and do nothing to stop illegitimate use)... it's literally impossible to enforce.
There's a million loopholes (secret distillation so it's no longer foreign, serving mystery model from different jurisdiction, putting it in your basement and not telling anyone). And you bet the motivation and money is there to jump through these hoops.
The whole article is very well summarized in these two lines :American leadership requires two things: world-leading American open-weight models and continued access for U.S. builders to open models already available worldwide, and I beleive Not only US but Europe and all around the Globe we should have same pursuit.
One question here: Why you don't see a lot of good American open weight models on the market? I'm guess it's about money? Could there be ways to lower that cost? Sharing idle GPUs through network (like Folding@home, BOINC), open and fair market for training data, open source training framework, and such?
This whole Chinese AI thing makes me giggle at these American AI companies, but the ultimate solution/long-term goal is still to make the technology universal. I don't think China alone can make it happen.
Also, if Americans can too offer open weight models, that could change the logic of their laws, making it more open rather than protective.
Money is a big part of it. The notable thing about Chinese open models is their size. Most open source models out of US labs have not tried to compete with the enormous flagship models you see from frontier labs now. Cohere, Arcee, Poolside, etc all put out open source models, and of course, Google has their Gemma line, Meta has their open models, OpenAI released an open weight 120b model, etc. The incentives are just tricky for running a massive training run for a fully open model. Monetizing can be hard, as if you want to monetize inference, you'll be competing against platforms like Fireworks etc that will also be serving your model.
Thinking Machines just released an enormous open model though, and I'm hopeful to see more of this globally. I believe open models are emphatically a good thing. But I do sympathize with smaller labs trying to figure out the economics.
As bad as it will be no one will be surprised when this happens. This administration does not care about the law. The president wants to keep using his position to make a generational bag. This is the same leader of the free world peddling shitcoin.
Everyone knows the game someone is going to cut him a check and suddenly the Chinese models are going to be a national security threat. The only saving grace may be that when it comes to China the president has shown himself to be a paper tiger.
The way he waddled around Zhongnanhai after embarrassing himself in the trade war and the Iran war was truly something. For whatever reason, Xi (like Putin) has him scared. But if enough sinophobes with deep pockets cut a check, it will not be surprising. It may even benefit China to watch the United States continue to attempt closing itself off instead of competing.
I don’t think he’s dumb, I think he’s very cynical and only does things he benefits from. He’s not dumb for selling out the country because he doesn’t care about the country, he’s accomplished his goal of enriching himself.
I can’t find it now, but I could have sworn Trump made a comment one time where he said something very roughly like, an earthquake can just happen and a million people in India die. And he was making a point about the senseless chaos of life, and to me it seemed like a glimpse into one of his core beliefs.
He doesn’t care if his actions lead to the deaths and suffering of thousands of people, he doesn’t care about democracy or the rule of law, he doesn’t care about productivity or building towards a utopia. I think he’s a deeply cynical man who sees no point to anything besides for enriching himself during the brief time he’s here. To that end, his actions aren’t dumb because he’s been pretty successful.
My favorite example of this showing through was when Bill O'Reilly interviewed him during the superbowl. He tossed him a softball that most presidents would have whacked at to say the United States as a democracy is more virtuous than countries like Russia[1].
But he didn't. His response was, "do you think our country is so innocent"?
The reason he became popular is the country had also grown cynical. The mainstream media wanted to anoint Jeb "Please Clap" Bush and another Clinton. That was the setup he walked into in 2016. A repeat of Bush v Clinton. America was over it.
He is not dumb. But he is clearly not that intelligent. It doesn't take much to be a cynic. He's not even a particularly good one. He's what he has always been - a salesman, really a con artist. But he's not dumb. To be a good con artist you have to be able to read the room. That is arguably, what he is (or has been) best at. It's why his debate performances especially in 2016, shallow as they may have been, were still a masterclass in dismantling the Republican field. "You'd be in jail" and bringing Clinton's own victims to the debate were more memorable than anything Hillary said. He knows how to win. There is intelligence in that. Cynic may not even be the right word. I think he's more of a nihilist. And given how his wealth has increased while in office, he's accomplishing his goals at pretty remarkable speed. The rest is just set dressing. He'll do on AI whatever whoever at Mar-a-Lago writes the biggest check tells him to do.
I think distillation is inevitable and pretty hard to prevent but I also dunno why anyone would spend billions to train frontier models if there's enough distillation going on to get near it with much lower spend.
Except this is not fraud. And just because a company _really_ wants to make a trillion dollar valuation doesn't mean they get to invent a new type of fraud.
People want to get from A to B. Someone makes a rule to stop them. So they go from A to C to B instead — or settle for something close enough to B.
The rule-makers don't stop anyone, and often the end state is worse than before the rule existed. This happens constantly with tech and the laws that try to contain it.
Tech is like reality: there are endless routes to roughly the same destination, and rules are slow. The asymmetry is the whole problem — you only need to find one new route; they have to block every possible one.
So what is the mechanism that the administratuon would use to block access?
I get how China could block access to Chinese LLMs, but any startup that can't or won't bypass a US imposed block on a Chinese offering was going to fail anyway.
I'm not sure whether Chinese AI models are being banned because of concerns about corporate data leaks, to slow China's AI development, or to protect the U.S. AI industry. But for startups, I think affordability is probably the most important factor.
Wanna plug LibRedirect which allows you to easily set up proxies for any site (e.g. invideous for youtube, nitter for twitter, libreddit/redlib/whatever for reddit, breezewiki for fandom, etc)
What happened to free trade and globalisation? Oooohhh only when it serves the west. What happened to the invisible hand of the market? Oooohhh it only when it serves the west.
Well damn that the Chinese are playing the game “capitalism” better than the west.
The last time this happen, the west just forced opium on China. Ironically it was the west that forced China to open its markets in the first place.
History just keeps repeating like a broken record.
My pet theory is that none of the Chinese labs are releasing open weight models for humanitarian reasons. I believe the Chinese government has made a cynical policy decision that AI is too valuable to allow anyone in their country capture a significant amount of its value as that might allow a person or company to accumulate wealth and influence so quickly as to challenge their power. Everyone must be kept below the God state.
Why jump to complex technical and policy solutions when an Open Weights AI champion trophy, in solid gold, is being requested by the Commander in Thief?
Note that a ban doesn't mean you can't download it anymore. Just that if you do so it will be considered illegal.
A competitor can simply accuse you of using a chinese model and throw you completely off rails with an investigation. Or when the government can't get a valid search warrant for a separate crime, they can get you on suspicion if using a chinese model.
exactly. a ban or even any restriction that is "fixed" in time is a bad idea.
The frontier is evolving so fast: any regulatory regime that started with a map of capabilities that 2025 frontier had is obsolete. Same would be true next year. This just becomes a whackamole and a government which is well-meaning and has good reasons to regulate, will have trouble keeping up.
But we know this about regulations. Like there is a century of literature on how to do this for new tech. It always targets harm prevention first. I thought Demis had the right framework for that part. I don't know if I agree fully with a self-regulatory regime because can become problematic if you don't have the right people around the table.
They will need to do something new here, somewhere between patents and copyright, to help protect US intellectual property on the AI stage.
Otherwise, US r&d will fold on this.
Longer term, 2 years plus, it will be a race to the cheapest as almost all models will be better than our current frontier ones.
American AI companies pirated hundreds of thousands of books, scraped the web against the express wish of site owners, extracted all public GitHub repositories without the explicit consent of code authors and distilled all of that information, basically the entire knowledge humanity built over thousands of years, into large language models. The way these models were built was ethically questionable. The models are out there now, and time can't be rolled back, but I still don't think the companies who built the models by ignoring intellectual property rights deserve any special protection of their own intellectual property.
How would that be justified, because as I see it would necessarily treat model outputs as more protected than actual human output?
Surely Anthropic/OpenAI ToS isn't something which can have more weight than, let's say, the ToS of a website expressed through robots.txt, considering that for the length, actual human effort has gone into writing the text, at a cost much greater than anything output by an LLM. So so long as the AI firms are allowed to crawl and then train on human-generated websites that don't allow crawling, but are crawled anyway, I think this kind of thing is really hard to justify.
Why? The current administration is doing such a fine job protecting the investments of a few hundred wealthy people, and not caring about the economy of the country.
The combination of a wide range of model sizes and capabilities, with good engineering choices as which to use for what applications, that is the way.
English is not my first language, but to me the construct "to urge not to do something" in the title feels off. I think I only saw it without the negative, with some implied urgency... And now that I spelled it out, "urge" and "urgent" are probably related.
US real median household income has been steadily rising[1] for the last 40+ years. Life expectancy is rising again after dipping and plateauing for a few years. Employment numbers are pretty good given the macro headwinds. The US forms ~5.9 million new businesses per year which vastly out paces the EU.
The US isn't without problems, but we're doing OK.
Because the basket of goods for inflation doesn't separate optional from mandatory consumption.
If housing, healthcare and education are outpacing the topline inflation number, a lot of marginal families will be squeezed. The cheap LG TV doesn't really make up for it.
Nevertheless if you lump those costs into an inflation number it won't be a lot higher, the Total Living Cost calculated for 2023-2024 showed 4.4% inflation vs 3% CPI. There are a lot of ways to slice and dice this, which is why people stick to the classic CPI.
If you factor in an extra 1.4% inflation per year into the numbers, you're not going to see any "real" wages growth over the 40 year period.
So the 1984 income is 60,420, if you calculate 60420 * (1.014)^40, which is assuming there's a 1.4% difference between the two disputed methodologies, you'd get about 105k. But the actual 2024 income is 83,730. Which I think means that if you take the non-CPI methodology to calculate real income growth, it's actually negative.
I'm not saying one method is more correct than the other (I don't know enough about economics to judge), but 1.4% annual difference matters a lot when you're talking about four decades.
The data shows that inflation inclusive of housing, education, and healthcare isn't that much higher than CPI, which is because CPI already includes shelter, which accounts for ~36% of CPI.
You assertion that CPI doesn't include housing is just wrong.
I'm not asserting they don't include housing. I'm saying housing is non-optional, so having low inflation on flat screen TVs is cold comfort if you're struggling to pay rent. The basket of goods isn't the same for everyone.
It's been like 5 years now of low consumer sentiment and increasing consumer debt while people who don't feel it are flabbergasted, the top line CPI number is fine, why aren't people happy?
I think once you have a good command of data and broader trends its harder to be cynical or a doomer. You just see that things aren't perfect but are larger better than the past on most metrics.
I think it's the dissonance between what the data say and just looking around. I see far more homeless, friends and family are more stressed, life in general just feels harder with less opportunity. Then you have the data people saying "no your life is actually great." So why the dissonance? I don't think it's just social media brainwashing.
I'm not saying any one person's individual life is necessarily better or worse, just that the median person is doing a little better than a few years ago. obviously there are long tails of outcomes in the US.
I'd also point out that the US has seen better wage growth and less inflation in recent years than the rest of the OECD, by comparison we're doing OK-er than our peers.
The reason we have data is because of all the biases that come with “just looking around”. Confirmation bias and selection bias are very powerful.
I strongly dislike Donald Trump, and he has done some terrible things to our country. But my dislike of him also makes me more likely to notice the bad things happening around me.
It happens with conservatives too. Ask my conservative family about how frequently men pretend to be trans and go into women’s bathrooms to get a peek. Ask them how often non-citizens vote. Ask them how the current crime rate compares to when they grew up.
Yeah this seems right to me. Politics is so polarizing it's a guarantee that half the country is going to be pissed no matter who is in charge. Most of my family leans left, it's hard to stay positive when the leader of your country actively and proudly hates you while also making terrible policy decisions.
It's much easier if you don't spend all of your free time mainlining negative national partisan news straight into your face. All of the most miserable people I know are very online and very into politics. It's not to say that you shouldn't care, just that the vast majority of the time you can do anything about it, so you're better served to focus elsewhere.
I completely agree, especially if you consider humanity as a whole.
I have asked this question many times to the doomers, and honestly don’t think I’ve ever gotten a good answer: if you had to be born as a random human somewhere on earth, and the only thing you get to choose is the year in which you’re born, what year would you choose? You can’t choose your race, intelligence level, physical abilities, parents, socioeconomic status, country, gender, sexual orientation, etc. Only the year.
So doomers: if the world is completely fucked up and going down the tubes, when would you prefer to have been born? Huge bonus if you can actually give some data showing that the average human was better off during whatever time you choose.
I think CPI is a poor measure, because it's a moving meter stick that understates inflation. When beef is overpriced, hamburger switches into the basket. When hamburger is a luxury, then chicken swaps into it instead. It understates inflation.
M2 change is a much better measure in my opinion. Using M2 is literally comparing supply of item to supply of cash which could immediately buy it. When scaling SPX or GC1! by M2SL[0]/M2SL, you get a surprisingly flat time series over decades, which reads to me that the effects of the change in M2 are being filtered out of an otherwise exponential price curve.
So many of my people in the US can't even make ends meet anymore that I opened a dedicated line of budget to provide help where I can. It's never been that bad before and it's nowhere near as bad for my people in the EU (not that everything is all rosy there, mind you).
You might argue that Americans are just less capable of supporting themselves but I reject that argument firmly. The situation really reads like systemic causes. And I think cherry picking numbers to convince ourselves otherwise does more harm than good.
Incomes at the lowest quintile rose 47% from 2021 to 2024. That growth slowed in 2025, but outpaced the growth of higher income quintiles in the first few years post pandemic. Real median household income in 2024 was $83,730 which is the highest it has ever been, and that is the middle quintile.
This one has always made sense to me. If you think about circumstances which improve the bottom quintiles’ relative income so that they’re now making as much as the median (I.e. the curve flattens out through unprecedented prosperity for the poorest) then the middle and upper middle incomes will rightly conclude that they’re not relatively high income anymore.
Placing an arbitrary person on the curve at the right place will show their relative position on this.
There’s a reason a large amount of the complaints are things like “my DoorDash is so much more expensive” and “my insta cart bill is so much higher”.
In the ideal state for the text writing middle and upper middle class, the mostly video consuming lower classes make very little money and provide services for cheap.
One should expect that enriching the poor upsets those whose relative wealth/income is no longer as high as it used to be. I imagine we could test this by comparing p75, p90 to p10 and p50
For my part I am optimistic. So many things today are so much cheaper and better than they used to be. Looking forward to the future - with the one caveat that AI is the precipitous ridge we must walk.
I don't think people need to hold a specific view. The experience of higher income percentiles is that the goods and services they rely on will get more expensive. Their position will be against a generic economic term, inflation, not against any individuals and awareness of other income percentiles will be limited. People at each position on the percentile curve will simply experience what they are experiencing and assume it is occurring to others as well.
Hmm, this is tougher than I thought to tease out. You can get income decile/tercile level sentiment but the problem is getting out workers vs. retirees etc. It would make sense that fixed-income earners would experience distress if worker incomes rose so I need to pull out that set but I can't find a way to do it. But this should be a massive smoking gun and I don't see the gun smoking that much. Sentiment for high-income mostly seems to just match the S&P500 and interest rates... I think it was a good idea but it's not true.
More people outside employment than COVID or Great Recession. Inflation wiping out income growth. Family units replaced by multi-job slavery shows up as “improved GDP”. There’s the powerball lottery that one person wins every few weeks and then there’s the crippling medical debt lottery that many people are “winning” every day. Wake up.
Real income means inflation adjusted, so we actually see the median income outpacing inflation right now.
> More people outside employment than COVID or Great Recession
There's a lot going on here. One factor is that many people retired early during the pandemic and are going to skew this stat. It's also worth pointing out that the labor force participation rate is ~62% and peaked historically at 67%, so it's not that far off.
The basket of goods is totally transparent and related to things people really buy. Would you want a 2026 inflation number to reference VCR prices?
Only 5.3% of US workers have multiple jobs. The long term average is just above 5%. So it is false to claim more workers have multiple jobs according to BLS stats. This number was above 6% for most of the 1990s.
You are just outing yourself as someone who is a slave to meaningless numbers. Do you think side gigs are reported? You quote unemployment numbers without admitting that people who have given up or simple failed to land a job are not included in those numbers. The numbers are gamed. Feel free to keep showing us who you are.
I mean, everything you just said is trivially refutable with just Google. FRED data isn’t the governments unemployment figures. FRED data relies on people to self report their side gigs that they aren’t paying taxes on? It is a known gap.
They cannot be copies, because all of them have different structures, which are explained in published research papers, while nobody knows the structure of the American SOTA models.
It is also impossible for China to have accessed the hundreds or thousands of terabytes of data that must have been hoarded by OpenAI, Anthropic or Google, so the Chinese must have gathered training data sets by themselves, by crawling the Internet and using pirate libraries, the same like Anthropic and OpenAI.
If the American allegations were true, then some Chinese companies might have extracted some blackbox data about the American models, by API querying, in an amount small enough that it could be used only for the post-training of a model created independently.
Such an information extraction is a breach of contract, but it cannot be called "copying", because it can be only an infinitesimal fraction of the information stored in a SOTA LLM.
Europe is actually catching up right now. Not at the frontier like kimi and qwen 3.6 (yet) but it's just a matter of time until the European A3B models catch up, too.
Apertus and Soofi, both open source (not only open weight) models:
Even more bizarre than debating over whether we can stop the free flow of information across communications surfaces (we can't)... is the notion that this one person - a highly ineffectual occupant of an office whose relevance is rapidly decreasing... is somehow in charge of it.
Extra Judicial punishment and unwanted attention from powerful organizations is absolutely on the table for any larger startup/company who finds themselves in his radar
Whether or not it's illegal is typically determined a few months after the act in question.
Since 2024, SCOTUS has sided with Trump about 90% of the time, so most likely it's legal.
Still, even if it's later determined to be illegal, since the law presumes the president has legal immunity to all crimes committed in a presidential capacity, there's no legal risk, so the damage can be done long before the court makes it illegal. For example, Trump's global tarrifs were determined to be illegal a year after the fact, long after the damage had been done.
On July 1, 2024, the Court ruled in a 6–3 decision that presidents have absolute immunity for acts committed as president within their core constitutional purview, at least presumptive immunity for official acts within the outer perimeter of their official responsibility, and no immunity for unofficial acts.[5][6][7][8] The court declined to rule on the scope of immunity for some acts alleged of Trump in his indictment, instead vacating the appellate decision and remanding the case to the district court for further proceedings.
It has the tacit consent of Congress, so it's mostly only illegal in the way it's carried out. Congress could make most of it legal right now, but they don't want to put their names on it. Some of it would probably be unconstitutional for Congress to authorize (ironically via the nondelegation clause which is now being repurposed to justify a thoroughly unconstituational unitary executive arrangement), but much of it would not be, so it's de jure illegal but de facto legal.
If you were to keep using a chinese model after said ban, you would need to be charged with a crime and be found guilty in a court of law for this to be true.
What it would do is have an instant chilling effect across Corporate America, which is the goal.
At a certain level legality is enforcement. If the enforcers of the rules decline to enforce them, you're better off acknowledging what is and dealing with it.
Never seen capitalists this afraid of competition. Same story with Chinese cars… We can’t have cheap, and objectively superior EVs because of protectionist policies towards Musk and his ilk.
Banning them would hand European startups a gift of low cost inference so I am 100% in support of this.
The globalists had no problem selling out our manufacturing to China, now they are crying because China can provide a cheaper service.
Anthropic and OpenAI have executed the greatest intellectual property heist in human history and they are angry because China stole from thieves and distillation isn't even technically legally defined yet. As much as I despise the CCP you better believe I'm using the cheapest models.
The globalists had no problem selling out our manufacturing to China, now they are crying because China can provide a cheaper service.
The globalists of old would happily hand AI dominance to China in exchange for a decade or two of short-term gains. The fact that protectionist measures are even being considered shows that the globalists who sold out manufacturing to China aren't at the wheel anymore.
They are. Selling out was done with the expectation that "the West" will outsource the basics, but will have hands freed to "reach singularity" and will keep the braids of all the technological chains. Now the "reaching singularity" itself is being lost, so everyone suddenly understands that there is nothing left.
Trade restrictions, sanctions. You can't effectively ban for individuals, but you can make it so legally risky for US based companies that it becomes effectively illegal as no US company will touch them, nor host them.
open weight models are not free models. Inference providers should share a meaningful percentage of revenue with the model companies.
In fact, open weight model seems a solid business model - they can leverage existing cloud infra for distribution without spending massive CAPEX.
So Y Combinator signed it too? But this is HN, Y combinator website. I think I am missing something because I never read any opinion like 'we should ban Chinese models'.
That's true, and often they diverge. But in this case I think the opinions within YC itself and in the HN community at large are more or less the same - we all want open-weight models.
Ok, but in 24 months that's the equivalent of downloading 100 variations of MSDOS. Torrents and vpn exist. I think this is more of a problem for corporations.
Individuals can. The US GOV can make it so legally risky for US based corporations to host, download, or integrate with foreign open weight models that they become effectively illegal for commercial use. Any sort of sanctions or ICTS regulations, and you'll see them all rapidly disappear from AWS, Azure, GCP, HuggingFace, GitHub, etc.
> The US GOV can make it so legally risky for US based corporations to host, download, or integrate with foreign open weight models that they become effectively illegal for commercial use.
They have to tighten that up so much that it would effectively result in "US Corporations cannot source anything from a foreign state".
After all, if you are allowed to outsource (for example) customer-support to EU, the company you are outsourcing it to can effectively use Kimi without disclosure.
> They have to tighten that up so much that it would effectively result in "US Corporations cannot source anything from a foreign state".
Sanctions are already that tight since they are strict liability. You do business with an overseas vendor and they use a sanctioned product, you are still liable even if you have no knowledge of its use.
In practice, outsourcing just becomes much more expensive, going through intense know your supplier audits, and you throw an indemnification clause in the contract and pass your fines onto your outsourced vendor.
They need to get rid of the Trump family dynasty first. That family is like the ultimate Al Capone family. It is perfected corruption on a daily basis. They wage wars even when a large majority of the population (see recent polls too) is against that, despite claiming there will be no more forever wars - we are seeing right now how those network of lies by the orange king are crumbling down.
"Nooooo!!! We already laid off half our workforce and got priced out of models from American companies. We need the cheaper models, even risking giving all our data to a foreign adversary, so our big number goes up long enough for us to get acquired!!!"
It will come for you too, fool. You obviously have no fucking clue what’s the deal with open-weight models. Can I run Fable on my computer? No. Can I run Depseek 4 or glm5.2 on my computer? Yes.
The US government exists to protect American capital and the interests of capital owners. I'm not sure why people are still confused by this.
This is about OpenAI, Anthropic and SpaceX. OpenAI, in particular, is a bet on there being a moat for AI and OpenAI "winning". Chinese models threaten this moat. The CCP has decided that it is in China's national security interests to not have Western companies "own" or "win" AI. These AI giants are large enough that the US government is going to intervene to try and protect this outcome. This is a losing battle.
I've heard of Fully Automated Luxury Communism as an alternative economic model.
The idea is machines do most work and humans don't work unless they actually want to. Supposedly everyone lives luxury lives of abundance with lots of free time.
I'd only accept something like that if the people actually retained control at all levels, not the government.
A president cannot (legally) use an EO to create new laws that bind private citizens or businesses. EOs can only be issued to federal agencies, and only Federal Agencies & Employees (and federal contractors) are bound by direct compliance of an EO.
Now that Chevron is no more, EOs are even weaker than they used to be because the regulatory agencies no longer have teeth.
Premature compliance != legally required compliance. Anyone complying early to an EO when they don't have to is an idiot.
The U.S. does NOT have copyright protections on the data that AI models were and are trained on, it was mass theft. Now all of a sudden, the Chinese are stealing from the thieves and we’re supposed to care?
No one cares, we just want cheaper AI models that are equally as powerful. They’re all thieves regardless.
Ok, but can you please follow the HN guidelines and focus on specific, thoughtful discussion, not generic/indignant tangents? The first is what we're trying for, and the second basically destroys it.
Ultimately this is more a problem with the upvoting system than the comments themselves, since generic/indignant comments routinely attract lots of upvotes, and then sit on top of the thread, smothering more interesting discussion. But in terms of moderation we can only reply to commenters.
Its even dumber when you consider AI model output was not considered copyrighted or eligible for copyright to begin with, as far as I know. How exactly would this even be "theft"?
Plus on top of that it's been ruled that training on copyrighted material is allowed because it's transformative. As a result even if the model output is copyrighted training on it would be as allowed as training on books.
Yea to be more explicit than the other reply they are being fined for the theft of the books not training on them. They trained on tons of legally procured books by destructively scanning and OCRing them. It was discovered in legal filings and is called Project Panama if you want to read more about it.
I can't imagine Moonshot have found a way to completely bypass the subscription or api billing either. So those model outputs have been paid for! No theft of any kind!
At most they'd be getting more out of distributed use of subsidized plans and API than the ToS would prefer.
I poison their requests with crap, particularly because they were the worst offenders for sheer request rate. The site has maybe a hundred pages, and I'd see them hit at a sustained 20 rps for weeks on end, same content being pulled over and over and over again. Easily handled by my small server as it's static content, but the sheer arrogance (inadvertent or otherwise) pissed me off.
I'm not sure if they caught on, or just finally started behaving better, but after months of being served crap in my never ending labyrinth of auto-generated junk, they have stopped.
My stuff is on Vercel, so it’s more of an issue as it uses up my limits. I’m moving to Cloudflare and I want to see how changes I make affect their bot, if at all.
Scrapping internet realistically does involve Botnets of all kind, the scrapping itself might be legal, but it's certain a ton of infected computers are routing traffic and being used as Proxy.
Training on copyright was ruled to be transformative.
Anthropic was just fined for not paying for the pirated books they copied into a training database. Same as anyone else who copied pirated IP onto their hard drive.
However, training an AI on copyright has been ruled to be sufficiently transformative and not a violation of IP laws. The same way you can make a gameplay clone of Call of Duty without any issue.
Also, if I clone Call of Duty, and use their skins, make a similar soundtrack, call my maps the same, then there will definitely be an issue.
I support the use of AI and all, but to hide behind transformative use of copyrighted intellectual property is a discredit to the colossal amount of human work and knowledge that these companies pirated and had their models trained on.
Correct, 2001 napster style pirating of content. You cannot copy stuff you didn't pay for onto your hard drive.
>Also, if I clone Call of Duty, and use their skins, make a similar soundtrack, call my maps the same, then there will definitely be an issue.
A gameplay clone as I stated, tons exist (team death match, capture the flag, battle royale, with first person gunplay and army guys shooting each other). The rest is your argument, not mine.
I don’t think I’d equate a corporation downloading TBs of copyrighted works (with the intent of making a profit on stealing copyrighted works) and John downloading a few GBs of MP3s for personal consumption.
Theft would give the victim a one-time payment, plagiarism would have given the victims payments every time the model was used. In other words: IP stealing through training LLMs is declared 100% legal.
The question is what happens when somebody mass-"uncopyrights" entire libraries of content for the first time. Because, of course, this judgement should have sent the stock price of Disney, Paramount, Warner Bros, New York Times ... to zero.
Why? Because at this point these companies are living on borrowed time. Courts have outlined a clear, legal, process to "uncopyright" any given work.
If I make Bambi 2 and toss it up on YouTube I’m going to be sued by Disney for a wide variety of intellectual property violations, it doesn’t matter if I make it by hand or via an LLM. Nothing has changed for big organizations with a legal team.
Sued ... of course. But what will they claim? The courts have declared that training an LLM on copyrighted data and then having it produce anything DOES NOT make a derivative for the purposes of copyright. That it is fully owned by the person who used the LLM and zero licenses are needed to sell it, rent it out, or whatever.
By that same argument, distilling a model to train your own is also transformative. We can’t have it both ways. Either both training on copyright and distillation are transformative, or both are illegal. Also, it is not a matter of settled law that training on copyright is transformative. That was a single case, and others are still pending AFAIK.
Why is it not fair use for the Chinese models if all they are doing is distilling these fair use models?
According to the courts, AI model output is not copyrighted or eligible for copyright.
But no Chinese company has copied and distributed an American LLM.
Even if they had wanted that, it would have been impossible, because Anthropic et al. do not let anyone access them directly.
Querying the available API can be used to extract only an extremely small part of the information stored in an LLM.
That part can be used for the post-training of another LLM, to obtain some desirable properties, but the extracted information is far too little to be called "copying" or "distributing". Moreover, after the post-training it does not appear anywhere in the weights verbatim, so its use is at least as transformative as the training of the original American model.
Besides these facts, there is no evidence that the Chinese companies have actually done this, even if it sounds plausible.
This is a silly argument. Like yes, obviously it’s of less U.S. concern if one American is robbing another versus a foreign interest. We can deal with one whenever while the other creates constraints.
The better arguments are the ones made in the article. Open weights increase competition and thus AI availability in the U.S. economy.
This is not robbing. Undesired distillation is a ToS violation, at best.
There is not even jurisdiction over Chinese companies, because they don't sell their LLM subscribtions, unlike US companies, and those comitted much more severe violations in getting their training material than a ToS violation or two (Anthropic already found guilty).
edit: I don't see this as legal argument against the ban, more like a justification for why basically no moral person is gonna side with OpenAI and Anthropic on this. US gov can try and ban as many weights as they want, I expect that to be similarly effective as banning numbers was in the past (i.e. not).
> There is not even jurisdiction over Chinese companies, because they don't sell their LLM subscribtions, unlike US companies, and those comitted much more severe violations in getting their training material than a ToS violation or two (Anthropic already found guilty).
This is false. They do sell LLM subscriptions. Many also release _most_ of their models as open weights.
None of it is robbing because there is no violence, but sure. The point is being confused why an external threat is treated differently from internal ones is a self goal in a debate where there are far better arguments for not banning these models.
It's not like those American companies were only stealing form Americans. Maybe you could argue about utility, but morally I don't see any ground to stand on
And the utility argument is also very shaky. The proposed punishment for foreign interests stealing "US data" is that other Americans now aren't allowed to benefit from that, while the rest of the world can
There isn’t an argument for banning Chinese or open-weight models. I’m just saying OP’s moralistic outrage take is about the weakest one you can come up with against it.
Not technically theft unless the public library no longer has the books. Selling something and saying you created it when you didn't is a different five-letter word: fraud.
Putting prompts in an LLM and saying you "created" the image or text thereof is fraud.
AI companies sell access to their models, via subscriptions and ads.
Public libraries don‘t sell anything. Also authors and publishers get payed for having their books in the library (how and how much varies by country). Authors and publishers get nothing for having their books put into AI training data. Hence the former is a public service, and the latter is theft.
Public library buy the works they distribute. They also pay for the usage of these works. Copying of said works is still illegal. You should read up on how libraries actually work.
Libraries used to require membership and dues. Those then bought more books on the used or retail market. LOTS of fights were about that system, cause book publishers hated libraries. And well, they still do.
And to be fair, fuck copyright. Its holds all of us back, so someone can go "FUCK YOU, NO". And nobody or company deserves what is it, 70 years+ death copyright length.
And with the recent Anthropic settlement, used to be, for-profit pirates would go to prison. I also remember absurd 2000's settlements over Britney Spears and Metallica for $5000-$7000 for 20 songs.
You are not going to win me over on massive gatekeeping human knowledge over "Itssss ill-eagle!". I know, and I don't care. And if I'm ever in a case over this, I'll tank it.
I strongly believe that's the case. That's even the case with the recent OpenAI hack of HuggingFace.
The censorship of US models is so bad (read: guardrails), that they had to use GLM 5.2 on their own hardware to do the analysis. That's how locked down, restrictive, and closed source American models are. They're basically for-profit piracy by way of selling tokens.
Whereas the Chinese models are just "here you go, download and run".
I know what I run. Qwen3.5 and 3.6 and GLM5.2 . USA token vendors are incentivised in doing worse to sell more slot machine tokens.
But distillation is not theft. They are paying for the usage of the API. It is also not copyright infringement, as AI companies do not hold the copyright for the output of their models.
Legally no, but I think an honest reading of the 'fair use' statute doesn't actually pass the 'doesn't compete with the author's original work' aspect of fair use, even if the use is transformative.
Rules are flexible. The people who pay attention know this.
People expect the US to apply rules evenly to both American AI efforts and that of its primary geopolitical rival. China certainly doesn't do that. Their entire economy is based around giving Chinese firms the advantage regardless of what it means for the wallets of consumers at home or abroad.
Since that's who we're playing against, and no one is going to willingly give up their open-weight models from the totalitarian rival, well, then use their ruleset.
It's that or punish the theft by making the US companies pay for their training data and blocking outside efforts that trained off the data the US companies stole.
Exactly! While the people who spent more money on their theft are rewarded with more business to counteract their losses! Justice is finally being wrought on those who have wronged society with their wicked ways!
(/s if it wasn't obvious)
Being serious, it is far more likely, especially given the events of the past 18 momths, that Trump is just doing autocrat things because he can, and ultimately doesn't really care what is and is not good for American business as long as he continues to grow his wealth through corrupt behavior.
Yeah, he really is. It's awful. We're creating a class of elites that are accountable to no one, and their hold on power is slowly becoming sanctioned by law. It wouldn't surprise me if we soon had a system where there were people in various professions who considered themselves professional Trumpists who got to steer the fields they were in according to the diktats of the party. You would have to be in-line with their ideology to get anywhere in a career, whether it be in industry, politics, arts, or science. It'd suck.
You know where you see a fully-developed version of this idea?
Why do you care about Disney's emotional outbursts in the first place though?
> The problem is that it is now explicitly reserved for the rich.
That's not new - it has always served entrenched interests at the cost of actual artists and entertainers. John Perry Barlow laid this out with great clarity throughout his time with The Grateful Dead, perhaps best here: https://www.youtube.com/watch?v=rLbqgG6o1n8
It's an unfortunate truth that we're exiting a global peace era and entering one of (hopefully just...) implied conflict. Our biggest geopolitical rival that's been threatening a hot war also distilling the things we build is not a good thing.
Trump is a huge self-own by our electorate sadly, but being concerned about a conflict between China and the US is a very bi-partisan concern for good reason.
Luckily for the peace side, the USA's inability to achieve their mission objectives in a war against an adversary that was perceived as being substantially weaker means they are very much less likely to engage in one with a "near peer".
The take away from modern conflicts is that massive military size disparities don't translate into successful invasions anymore.
In other words, it's unlikely USA will pursue a conflict with China, and it's unlikely China will pursue a conflict with Taiwan.
The main actors threatening (and delivering) a hot war are China, and by proxy, Russia. The Ukraine war is absolutely and transparently enabled by China, who is also trying to actively destabilize the EU and the US in order to make its conquest of Taiwan more likely.
Yes, the US has an idiot in charge who is threatening these things in response, too. However, unlike China, the electorate does not support him. His election is also almost certainly due to aggressive campaigns to destabilize the US internally for this exact purpose.
>However, unlike China, the electorate does not support him. His election is also almost certainly due to aggressive campaigns to destabilize the US internally for this exact purpose.
What are you claiming here?
That China hacked the voting systems enough to swing the election? I know of no good evidence of this.
That China changed voters' minds enough to swing the election? But then "the electorate does not support him" does not follow. If I change your voting choice using some kind of non-coercive persuasion such as by making an argument, your support for your new choice is still genuine support. It doesn't matter if my argument was good quality or not, what my country of origin is, or even whether I lied about my motives in making the argument.
Our biggest geopolitical rival is also the only thing standing in the way of a potential duopoly on LLMs using closed weight models. I would not say I’m a fan of the PRC by any means, but I also do not want to see closed weight models controlled by two big companies dominating what I can use. Especially since said companies arbitrarily restrict model usage and can ban users for no reason.
I keep hearing about this "we" as if it's still a coherent nation state and not just an unaccountable ruling class from both parties where the most severe consequence is failing upwards into a lobbyist provided position. I literally have no moral preference when it comes to the US or the CCP nor do I care if the entire nation state fails, let alone loses to a geopolitical rival.
Are we even sure they still work? The last test was 34 years ago, after all.
For all we know some crafty contractors or NCOs have been making very good money off of selling "spare" parts - just like we saw when Russia first tried to invade Ukraine...
As I see it, one party is comprised of fascist plutocrats, and the other party is comprised of their enablers. In dictatorships the latter is called controlled opposition, I don‘t know what the latter is called in a democracy, but in absence of a better word “both parties” gets the meaning across IMO.
My argument is that a major world power is pushing towards global conflict by simultaneously funding the conquest of Ukraine while building its own invasion fleet for Taiwan. Allowing them to also destabilize us internally is not exactly a good idea.
It is well known that China is still engaged in trade with Russia- oil and gold going one way, drones and weapon components going the other.
But you said funding, and I assume you meant things like grants or loans. Do you have any sources for that? I could only find one that was attributed to indirect open source intelligence, nothing specific.
Russia has been trading with a number of countries[0]. Mostly China, but also significantly with India, Turkey and the EU, and in much lesser amounts with 30+ other countries, including the US, the UK, UAE, Brazil, Japan, South Korea, etc.
I don't know if the above data helps to make your original point or not. Would you say that all those countries are funding the Russian war against Ukraine? Perhaps they do so for a combination of various reasons - energy and food needs, business profits, other self-interests, don't care, etc.
Wouldn't a country pushing for global conflict - really global - sow disorder by doing things like: break up alliances, threaten annexations, back illiberal poitical movements, disrupt shipping, disrupt free trade, disrupt energy supplies, start and foment regional wars, etc? Given that all countries tend to do things like this to one extent or another, how many of these things has China done, and how does that compare to other similarly capable countries?
I would say that China is asserting itself as a regional power, like with a navy that threatens Taiwan but can't project power much further than that, wouldn't turn into a global conflict unless other global powers feel so threatened they make it one. Maybe just a play on fear, for power and profit. War-mongering. Wouldn't be the first time.
EDIT: And look at how China has been "funded" China (via trade) over the last three decades... ironic!
> Allowing them to also destabilize us internally is not exactly a good idea.
The US destablized itself internally. China did not force the US to export its manufacturing base. It did not force the US to invade Iraq. It did not force the US to create a housing market bubble. It did not force the US to create massive income and wealth inequality. It did not force the US to supress transitioning away from oil towards cheap renewables. It did not force the US start a global oil shortage by invading Iran.
That having found itself at the tail of 25 years of poor financial planning, the US decided to go double-or-nothing by betting on LLMs to bail out the economy... does not mean that China has to fold its hand.
No, the US is actively invading Iran, by ship and by air. And they've invaded Venezuela. And they're actively supporting Israel's invasion of Lebanon and Gaza. And they're bombing a few more countries now and then, especially in Africa.
China is not actively involved in any war at all. They may be supporting Russia with resources, but it's nowhere near the level of material and diplomatic support that the USA is providing to Israel's genocide and war of aggression on Lebanon.
How is it destabilising? Do you purely mean China have exposed ill thought out investments? That’s capitalism my guy. The market is working as intended.
When the investments fail there will be volatility but it’s hardly existential for the US. Open models are good for the whole world, I would argue it’s a net positive for individuals.
I honestly wonder what world folks like you live in. The US is currently at war, repeatedly threatening several other places and engaged in several costly trade wars. From my Canadian perspective, you are a danger and a menace. China is a commodity supplier. Do the war crimes your country is currently committing not register for you at all? The daily threats, insults, and economic attacks on your 'allies', does that register?
China supports Russia in its war formally, but practically it sells Ukraine whatever it needs, because a weak vassal-like Russia is very much in their interests. The US is far, far more outspoken and active supporter of Russia in this war.
Finally, there will be no hot war with China. They'll likely blockade and invade Taiwan before the end of the decade, but the US will not do anything, partly because it'll be exhausted from it's never ending middle east wars, partly because it'll only take a small bribe to get Trump to back off. He's genuinely fond of dictators like Xi.
Or… war with China is inevitable if only to keep the gap in military capabilities. There is a thinking to mow down their military while we still can in some circles. Like those in power currently. Scary.
Almost all current world events as of now are directly tied to the fact that the west is in active freakout mode over china building a army large enough to attack the entire world. Iran and venuzuela are a direct response to it.
1. China is actively genociding an entire ethnic people.
2. They are actively building a military larger and faster than any country on earth in all of history.
3. They are actively threatening ALL of their neighbors with war. Actively attacking their ships and building military bases in other countries territory.
4. China's open stated goals are to conquer and control its neighbors and every single step over the last 10 years is proof of that.
I think the major destabilizing force is the country that has been building the infrastructure for a hot war over Taiwan for the last decade while simultaneously blatantly funding the conquest of Ukraine.
The US stupidly withdrawing from the world stage is definitely making this worse, of course, but it is not the catalyst. It's unfortunately a very likely side effect of a social media campaign by the same actor to make said hot war easier to win.
It's literally on their border and the history is akin to if The South established its own nation state in Cuba after the Civil War and China became its ally.
At least admit the double standard. I don't mind realpolitik but it's so tedious when it's not plainly stated and there's this veneer of civility.
Yes, the US is allowed to do whatever it wants to in South America. No, other nation states don't have a sphere of geographic influence if the US doesn't want them to.
Taiwan has been self governing for longer than almost everyone that lives there has been alive. It's also a vibrant democracy.
If you care about democracy and human rights, it's morally reprehensible NOT to support Taiwan. We can debate the ethics of intervening in strongman dictatorships like Venezuela, Cuba, or Iran, but Taiwan is clear.
Cool, and it's morally reprehensible to NOT support drastic action to curb the worst effects of climate change, and start trying our very best to prepare our children and grandchildren for an adulthood where they will be clawing back minor things we take for granted even in today's world of regular continent spanning forest fire smoke pollution.
Despite withdrawing from our role on the world stage, we have in no way adapted isolationist like policies. We kidnapped the leader of another country, we are bombing Iran with impunity and will begin striking civilian infrastructure as a response to terrorist within that country, https://www.usatoday.com/story/news/politics/2026/07/22/trum...
China didn't make us do this any more than they made us bomb civilian fishing boats in South America. We are a major destabilizing force at the moment, China existing does not somehow negate the actions we take of our own free will.
Refusing to acknowledge our role in destabilizing the globe at the moment is a level of infantilization that I, and many others, find unacceptable. We are all adults, China is responsible for their role and the US is responsible for its role. A social media campaign is not forcing the folks in Washington to drop bombs on Iranian civilian infrastructure, the jack asses in Washington did all of this on their own
Nobody is refusing to acknowledge our role in destabilizing things. This is a thread about why the US might be afraid of an also bad-actor rival funding a huge proxy war against it controlling major technology.
I mean if you want to get back on the topic of this thread, it has nothing to do with what you just said. The thread started by discussing AI distillation and this was your initial comment, just to get us back on track
> Our biggest geopolitical rival that's been threatening a hot war also distilling the things we build is not a good thing
They didn't distill Fable, it is impossible on any timeline. So where do we go next, the start of this thread is a complete dead-end as it is asserting a falsehood
This was the comment i responded to so we can get back on track to that comment if you'd like
> I think the major destabilizing force is the country that has been building the infrastructure for a hot war over Taiwan for the last decade while simultaneously blatantly funding the conquest of Ukraine.
We are refusing to aid Ukraine and actively antagonize their leader. We are kidnapping leaders from other countries. We are bombing civilian infrastructure. We are the major destabilizing force, not the one planning a hypothetical invasion
In the decade during which you claim China has been preparing for a hot war, the United States invaded multiple countries and waged hot wars on three continents.
It’s ridiculous to compare China to the US when it comes to warmongering. They’re not even in the same ballpark.
Yeah, the US is not an incredible actor, but China is funding the most brutal war in modern history (the invasion of Ukraine) in order to provide a smokescreen for their own invasion.
The Ukraine war is large and terrible, but its nowhere near comparable in brutality to the genocide in Gaza, that the US is funding, providing weapons for, and protecting diplomatically.
This is a thought-terminating cliche. (The folks telling you to stop being sheeple are unusually consistently claimed independent thinkers.)
> How much damage has China done recently militarily?
Exempting the Iran war, which has no justification, as much as America. We’re both backing proxies. (Though we’ve had a while since we shot at or annexed our neighbours.)
It seems pretty silly to answer a question about military intervention by insisting that the major instance of military intervention by a world power that's happened in the last year shouldn't be counted
> by insisting that the major instance of military intervention by a world power that's happened in the last year shouldn't be counted
I’m not. I’m saying exempting that, we have an even field of conflict mongering. Counting Iran, the field is biased, but we have significant sampling-timing issues for something like “is there war.”
China has violently annexed sovereign countries in the past decades. Its threats to do the same again are at least as serious as Trump’s Greenland nonsense. There can be more than one baddie.
Iraq war? Venezuela raid? Libya? Syria? Afghanistan for 20 years? And that's just some of the higher visibility cases.
If China had special forces deployed and conducting ops in random African countries it would be a massive aggression. When we do it, it doesn't even make the headlines.
I think they're referring to occasional fist fights between the Indian and Chinese deployments at the top of the Himalayas where there's a dumb territorial dispute. There were a couple of times decades ago where the forces shot at each other and the governments had to scramble to calm things down and deescalate, so somehow they've reached this gentleman's agreement to have fistfights instead.
> We're talking about this kind of thing balancing out, say, Fallujah or shock and awe over Baghdad
Nonsense. For that comparison you’d look to the invasion and annexation of Tibet. (And it’s not ancient history if it’s currently endorsed and un-apologized for.)
Ok, then we expand the US horizon back to the 50s as well. We've added Tibet, and the 1979 3 week Vietnam war to the Chinese column, plus them being involved in the Korean war and to a lesser degree the American Vietnam War. This comes with way, way more added to the American column. At least it's no longer a shutout.
Any fair look at both countries' records puts the US as doing 10-100x more military intervention, on a global rather than regional scale. It is actually a little bit "sheeple" to look at that scoreboard and start calling China aggressive. They're less aggressive than every other UNSC member without even going to a pound for pound basis.
> Any fair look at both countries' records puts the US as doing 10-100x more military intervention
Sure. I never claimed otherwise. I’m saying it would be news to anyone in Asia-Pacific (or Eastern Europe) that America is or has been the destabilizing force. That was the original claim.
If the choice is 10x or 1x, go with the latter. But that is a false dichotomy. You can be anti-invasion and annexation broadly. There is no moral high ground in choosing Chinese models; no more than a person in China choosing Anthropic or OpenAI.
The Philippines definitely have a right to be aggrieved over the south china sea stuff, but one bias I'm cautious of is "the natural norm is American hegemony in every region, so any change from this is destabilizing". Other countries have their natural interests as well.
I’d say the South China Sea island building in violation of UNCLOS text and judgement comes pretty close to American CIA-style fuckery in South America and Africa.
To be clear, American fuckery is broader. But if we agree we’re discussing magnitudes and not kind, we agree on all I’m arguing.
> it's former sovereignty is very questionable and not clear cut
Same as Taliban-run Afghanistan, same as Iraq, same as Ukraine, same as Taiwan. If you’re Tibetan, the line is academic—you faced, by modern definitions, genocide all the same.
Calling the territory you’re invading and/or annexing illegitimately independent, illegitimately governed is how these always go.
Wow, what happened to no holds barred American capitalism? Is it bad when the Chinese somehow do it better? The US stateowning US Steal, Intel, and possibly openai is no different than the Chinese ownership model… and before the US took stakes in the companies they were just bankrolling them as customers SpaceX Oracle etal.
If life is a meritocracy, you gotta give it to the Chinese for their immense growth and improvement of living from agricultural poverty to technical supremacy. I don't want to hear the they stole our IP argument either, because all math is borrowed from something else. Distillation is innovation, in booze and in models.
People on here have zero consistency in their worldviews.
Ask any capitalism zealot (ie hners) whether markets are zero sum and they'll wax poetical about "creating value". But then talk to them about Chinese competition and suddenly everything is a rivalrous good lolol.
I hate to break it to you, your so called rival won this game decades ago when you undercut the working class to make a buck and laundered cheap Chinese labor as productivity. So if you believe the Chinese are rivals um welcome to "reap what you sew" <shrug>.
I'm not the "hive mind" of HN or whatever. Individuals hold unique opinions.
Yes, China is a rival. No, they have not won. Their economy is in a lot of trouble, likely far worse than the shape of the US's currently. That situation may accelerate a hot war though, as has historically been the case globally.
People have been claiming this since the 2000s. There are Economist articles from 2005 predicting an imminent collapse of China's economy. China has only grown more rapidly since then.
I don’t. I do read major reputed economists though, and they are not exactly bullish on China.
Don’t create fictional versions of the people you argue with. This is the second time in a span of 10 minutes that you’ve done it. It completely undermines any attempt to discuss something, and drops your integrity to zero.
I’m sorry but this is a perfect example of what the person above criticized but in reverse. Come on now.
China’s youth unemployment rate is 15%, and that’s likely heavily astroturfed to look lower than it is. Even 15% though is terrifying for a mature economy.
You know who should be terrified by that rate? Us.
China's one-child policy is finally starting to hit, and its population is actively shrinking. At the moment China is not in a crisis yet, as its GDP is still growing quite well - so clearly there's plenty of economic activity.
So if a lot of old people are now retiring, and the young can't find jobs, who is doing the work? China is still a factory powerhouse, so who's screwing together the gadgets?
The answer is that they have been pushing heavily for automation, and it is working. Our biggest chance of beating China on a free market was domestic automation versus cheap overseas labor - but they have gotten there first. And that should be absolutely terrifying you.
China leads in 66 of 74 critical tech areas. The USA has a whopping... 8.
But hey, the USA's MBA "leadership" figures how much to worsen everything and extract more profit. I guess that counts for "innovation", in the Private Equity books.
China is only getting stronger. They have captured basically the entire electronics industry - except a handful of cutting-edge chips. They have captured the solar panel industry. They are significantly outperforming the traditional brands when it comes to all kinds of BEVs.
In all the ways that matter, China is rapidly racing towards or even past the West. We've been pretending that it is possible to remain relevant on the world stage while only offering services and a handful of luxury goods. A China without the West would be a painful transition, but the West without China? A massacre.
China hasn't won yet primarily because the West hasn't even figured out it is in a competition. We're getting our asses kicked and don't even realize it yet! Worst of all? The Western capitalists created China, all for a few cents of shareholder value...
Unnecessarily rude, I’m sure you agree. What separates this view from the following?:
“I started a Black Cats Only Association whose main membership is white cats, and a critique of the mismatch reveals failure to recognize it’s simply a proper noun.”
> the US can “sanction” companies that engage in intellectual property theft.
Peak hypocrisy, US AI companies can train on unlimited intellectual property with 0 rights to it, while Chinese AI companies have to explicitly get the rights to data that isn’t even copyrightable/copyrighted (since AI alone can’t copyright it).
And if the accused company is outside of the US, well, US courts have no jurisdiction so apparently the US government can just claim they are guilty and impose the sanctions...
>Anthropic was asked to pay book authors because it trained on pirated downloaded books.
No, it was asked to pay book authors because it pirated copies of books and stored them on their hard drives. The ruling had nothing to do with training.
Aaron Schwartz was re-publishing the journals, which is the core concern or copyright and indeed where the word "copyright" comes from. Anthropic could get in a lot more trouble if their models are caught reproducing copyrighted work from their training set wholesale.
I think the Aaron Schwartz case is incredibly vexing because he was obviously acting out of a sense of altruism without personal self-interest. I don't think he deserved the book getting thrown at him like that. But the whole copyright system, which people seem to think is simultaneously good and bad, kinda rests on not allowing those kinds of violations
I've linked this several times, but LLMs are capable of reproducing entire books. Researchers were able to extract books nearly verbatim: https://arxiv.org/abs/2601.02671
The judge explicitly ruled that training on the books was fine.
They just should have bought them, rather than pirating them.
Also LLM output is not IP (in itself) in the first place, nor would Anthropic want to claim it is and that they have rights to it - that would drive paying customers away.
> The court also held that the third factor favored fair use as to the purchased library copies converted from print to digital because the purpose of the copying was to keep the books in its library but with more favorable storage and searchability properties. This purpose required copying, there was no surplus copying and the source copy was destroyed. With respect to the pirated copies, however, the court held that because “Anthropic lacked any entitlement to hold those copies” and retained them “even after deciding it would not make further copies from them for training,” this third factor weighed against Anthropic for that particular use.
Weaken U.S startups = Can’t legally develop commercial products that use or remix the latest open weights models in novel/creative/cost-reducing/etc ways, while companies from the rest of the world could (including China, who this is ostensibly supposed to be punishing)
Wouldn't stop proliferation = Despite that, your non-programmer 14 year old child can still install the latest Qwen model on Day 1 by copying a one liner command they got from a YouTube short that sources from a Gitlab repo outside US jurisdiction
It does because corporations, especially smaller ones competing with bigger ones, actually do get prosecuted for violating restrictions like this a lot more than individuals do.
I think it would do the opposite. They would have a monopoly just handed to them, and a large market with no options. They'd have zero incentives to provide open weight models.
Right now, I think their incentive is the highest: Chinese AI companies are dropping competitive models left and right, and clearly American companies are feeling the pressure, given they're pulling every legal leg they can to suppress their competition. If we just remove the competition, Anthropic et al win. It's a temporary and short-sighted victory, but this is America. Every company is short-sighted, it's baked into our financial infrastructure.
US labs would have even less incentive to build open if we blocked foreign open weight models, and we'd be engaging in protectionism yet again, also giving them zero incentive to compete.
I‘m not even sure what the argument for banning Chinese models/open weights even is supposed to be?
1. if it’s to stop hackers doing hacking things with „uncontrollable models“ then, well… they’re already doing something illegal to begin with, why would they care about breaking another law running these models?
2. if it’s to stop foreign actors, then that ban would not apply to them anyway
3. it’s not stopping distillation either, Chinese labs are already banned from using US frontier models and look at how good that is working
I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term. But thats also an admittance that the American labs can’t compete on merit anymore, and should by itself also limit the viability of the idea that all those VC billions will ever make a return? In any case this would be something benefitting only a very few for a short time (labs + investors).
Someone please enlighten me what the actual argument here is, cause I can’t see it.
> Someone please enlighten me what the actual argument here is, cause I can’t see it.
> The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term. But thats also an admittance that the American labs can’t compete on merit anymore
That's the argument. To be precise the publicly stated argument is that they're attacking American providers by distilling. The real aim is to eliminate competition because otherwise Anthropic and OpenAI are non viable and the US views them as crucial for winning the 'AI race' which they see as putting whoever wins it on top in terms of warfare/economic power etc.
But isn‘t the whole argument of „winner takes all“ already dead, if china is such a big threat that the US labs need import controls?
Distilling or not, they are clearly close enough to the frontier, that the supposed „free market“ country needs market controls. That may work for US markets. But not the rest of the world.
Idea: petrodollar policy becomes „tokendollar“, enforced by US military dominance. If you force me to pay altman at gunpoint, then maybe ill stop using kimi
They are simply out of ideas to keep the most expensive party in the history of mankind going.
That's it. Simple as that. When you grasp for straws in panic mode, you don't exactly spend time strategizing and weighing the pros and cons of each straw carefully.
It's going to force American companies to only use inferior and more expensive models.
It's protectionism which will only extend our dominance by at most a decade.
The rest of the world will progress without us.
Decade? We're 4 years in since ChatGPT released publicly and have gone from "Wow this thing is stupid" to running coding agents on auto mode, and the Chinese labs have been around for even less time. The US providers may lose dominance in 1 or 2 years.
> may lose dominance in 1 or 2 years
In some portions they already lost dominance. My claude will happily generate disgustingly vulnerable code then flag me for ToS violation when I ask it to make sure the code that it just made, in the same context window, in the same conversation is safe and doesn't have glaring holes. Also great to be charged $xx.xx+ amounts and then it refuses to answer without refunding it.
They won't. China doesn't have the training capacity and the quality data sets, the former because of access to fabs, the latter because of internal censorship that is getting worse by day.
Simple as that.
The current panic is happening precisely because Chinese open weight models like Kimi and Qwen are demonstrably competitive with SotA models like Fable, but far cheaper.
https://fireworks.ai/blog/kimik3-fable
No, the Chinese tech companies routinely super optimize for a given 'famous' or 'established' or 'mainstream' benchmarks. Any other concern is secondary.
In other words, they look good on surface, but suck whenever anyone put them to any serious use. That's why they are cheap, they have to be cheap because they suck.
how can you say that with a username of "ngl999", that doesn't make sense
(IIUC, -ngl [NUM_LAYERS] specify number of layers to offload to GPU in llama.cpp, 999 on most use cases might as well be -1)
> the latter because of internal censorship
Having limited amounts of high quality Chinese data to train on doesn’t really decrease or affect the appeal of these models to the rest of the world. I don’t think the CCP is restricting or limiting the datasets that models can be trained on.
Also the models themselves don’t even seem to be censored or restricted that much. If the Chinese government is fine with the guardrails being on the API level and maybe even restricts the usage of external providers that’s again a net benefit to everyone else since they would have little incentives to force Chinese companies to lobotomize their models.
You are probably about capacity so we can only hope that Huawei and others can catch up and break Nvidia’s monopoly (since Intel and AMD suck too much too much to accomplish anything useful that’s the next best thing)
Also Chinese labs are actually still publishing research publicly which alone would accelerate and increase the competitiveness of open AI models developed in the US or even Europe.
> doesn’t really decrease or affect the appeal of these models to the rest of the world
Anyone who tried using Chinese models for serious programming tasks ditch them quickly if they have a choice.
> I don’t think the CCP is restricting or limiting the datasets that models can be trained on.
What I meant was censorship limits the amount of high quality training sets by limiting the amount of all training sets from which high quality sets grow out from, e.g. the entire set of posts on Baidu forums before 2017 is gone forever.
> so we can only hope that Huawei
Chinese companies do not have access to commercial-scale advanced nodes. The impact is at best negligible.
> Anyone who tried using Chinese models for serious programming
Perhaps, but that’s because they are just objectively worse for many tasks than GPT/Claude. I don’t see how that ties to censorship inside of China.
> censorship limits the amount of high quality training sets by limiting the amount of all training sets
Well again, I don’t really understand how is the lack of high quality Chinese datasets matter much if they have access to English/etc. datasets. That’s only an issue to their users inside of China.
> not have access to commercial-scale advanced nodes
Well the gap was way bigger a few years ago. If Chinese companies can at some point produce GPUs that are competitive cost wise i.e. they are willing to tolerate significantly lower margins than Nvidia (whose margins are obscene) that’s not a huge issue.
I’m talking about our overall economic dominance.
We’re still the most powerful economy on earth, but I can’t imagine this continuing forever.
By what metric - GDP? That measures price level, not economic power.
Most expensive groceries, rents, and healthcare in the world!
Winning!
stupidly large number of lawyers as well.
fixed for you.
> By what metric
imperial
funny joke so I vouched it, but you should know you've been shadowbanned since your first ever comment.
> the supposed „free market“ country
Good scare quotes. The US hasn't been free market in the Adam Smith sense in ages, if ever.
well free markets meant something very different in adam smith's time; what they meant by free was 'free from rents' whereas today its usually mean in the milton freedman terms as 'free from any interference' which is how i interpret the parent poster...
https://en.wikipedia.org/wiki/Free_market#Classical_Economic...
It's not free from interference either. It requires enforcing antitrust, for example, which the US seems to have forgotten. Literally what you quoted:
> the term free market refers to a market free from all forms of economic privilege, monopolies and artificial scarcities.
> free from any interference
That’s even less true in the US these days, though
D: "I want to be free" by Milton Schulman.
H: Milton Friedman.
D: Why are they all called Milton? Anyway, I've only got as far as Milton Keynes.
H: Maynard Keynes.
D: I'm sure there's a Milton Keynes.
H: Yes, there is, but it's... [appears quite miffed]
https://www.youtube.com/watch?v=KgUemV4brDU
https://en.wikipedia.org/wiki/Milton_Keynes
Coincidentally it’s where the British version of the Office is set
This Schroedinger's Freedom is getting tired now. They happily label themself as free-market democratic liberal progressive when you're justifying bombing a poor third-world country, but suddenly claim to be unfree undemocratic oligarchic whenever anyone asks them to act on that principle.
Adam Smith himself knew you had to constrain the market in ordner to protect capitalism from capitalists. His work didn't end with his first book.
Yes, it’s freedom both from state interference but also from monopolies, oligopolies and trusts using their position to suppress the competitiveness of other market participants. Which is generally impossible to prevent without the state interfering in the markets.
US didn't have that much rational reasoning when it killed Japanese exports in cars, chips, and finally software. That in turn lead to China simply taking the seats Japan used to occupy, but as far as imposition of these protectionist means is considered, US does not need that much sophisticated justification internally.
The argument is - these models costs hundreds of billions to develop - to BOTH US and China. China is giving them for free despite not making financial sense because they want to weaken US's tech firms ability to compete - because China sees AI dominance as critical for their national security.
The US also sees AI dominance as critical for their national security, and relies on a market economy and non-state-owned labs. This means that these frontier labs are truly susceptible to "predatory pricing" (economic term for a competitor selling at a loss to eliminate you), which is illegal in the US and any other free market economy exactly because of its implications on market efficiency.
I hate it just as much as the next guy, but the fact the discussion here ignores the fact that these concerns and dynamics are real just lower the discussion level instead of actually discussing potential solutions.
With that said, what I'm worried about is that the government solution will be far more hostile than some semi-ban on open models. An example of an even worse scenario, they could take over frontier labs and restrict access to everyone in the public (which might still not solve espionage).
> China is giving them for free despite not making financial sense because they want to weaken US's tech firms ability to compete - because China sees AI dominance as critical for their national security.
Nobody is making this argument because this is a normal thing that companies do. It's so common it is named in business strategy: "commoditize your complements."
There's a 2002 Joel On Software post explaining this using tech examples that were already old in 2002[1].
In the present, Meta is doing exactly the same thing with Ollama as Alibaba (who is funding some of the Chinese labs). Alibaba needs advanced models internally, and does not want to have a dependence on US models that can be arbitrarily shut off by Washington. Alibaba also wants to drive usage of its cloud among companies who are now demanding competitive AI models (which Alibaba may be locked out of providing). However, selling AI is not its core business. Putting cheap AI in the world creates demand for cloud services, which is part of Alibaba's core business.
Not only does this make financial sense for Alibaba, it is fairly basic.
Google gives away consumer software like Chrome and Android (not their core business) to drive search, which is their core business.
et cetera. The frontier labs are exposed in exactly the same has been way every pure play tech company before them. The US labs chose quality as their moat; remains to be seen whether that was a good choice, or if they can add another moat in time.
1 - https://www.joelonsoftware.com/2002/06/12/strategy-letter-v/
> Putting cheap AI in the world creates demand for cloud services
No, Alibaba doesn't have the right amount of hardware to serve these models.
The part that benefits Alibaba in these open models is to maintain its public reputation so as to maintain its stock price. The core that drives everything else inside Alibaba is its e-commerce.
> Nobody is making this argument because this is a normal thing that companies do. It's so common it is named in business strategy: "commoditize your complements."
It is common in business strategy, far less so in nation-state governance.
The reason you, me, and so many people in the United States still have a far better quality of life and level of freedom than most people in China is because it’s bad to run a country like a business, full stop. When you do so, you kill far more people.
> It is common in business strategy, far less so in nation-state governance.
It is not clear why this is bad for Alibaba or China, could you expand on that?
> it’s bad to run a country like a business
The "Chinese miracle" resulting from China's integration with the global economy has lifted ~800 million people out of poverty and raised standards of living quite rapidly overall. A lot has been written on the topic, but suffice to say that it's a fairly unorthodox argument to make that the Chinese have not been well-served by the arrangement.
> level of freedom
The context of this discussion is whether the US government should prevent US citizens from accessing yet another product that is freely available in the rest of the world, to benefit a tiny handful of rich people in California.
Let me be clearer: if you do not hold stock in OpenAI or Anthropic, you likely stand to benefit from more AIs being provided by more providers. Imagine if the government had banned MySQL and PostgreSQL, so we all had to buy databases from Oracle and Microsoft. That would be strictly worse for everybody except employees of Oracle and Microsoft. This is exactly the same thing.
Great points, but I think that banning open source databases would have been better not for employees, but for shareholders.
Aren't employees of US tech companies typically also shareholders, by virtue of part of their compensation being paid in company stock?
For private companies to an extent yes. In public companies that’s just a way to increase your employees compensation without harming your free cashflow. I don’t think tax wise it’s any different than them paying you more in cash and then you buying the same amount of shares.
> The reason you, me, and so many people in the United States still have a far better quality of life and level of freedom than most people in China
I think it's because every single time the farmers in south america or the philippines try to organise and get out of poverty you guys either bomb them or arm some terrorist to kill them.
But sure tell yourself otherwise if that makes you feel better.
"you, me, and so many people in the United States still have a far better quality of life and level of freedom" LOL - are you sure about this?
That’s extremely American centered, though? Despite the constant complaining the median American is still very well off compared to almost everyone else. If you go to the top 10 percentile or so them maybe only Switzerland and some very small countries (tax havens, petro states and such) can really compete.
The words of george orwell has never been so accurate.
"War is peace, freedom is slavery, ignorance is strength"
Sometimes sure, sometimes unsure but always free to say it publicly. ;)
> you, me, and so many people in the United States still have a far better quality of life and level of freedom than most people in China
Whether this is true or not is still open to debate, but the fact that it’s presented as some kind of absolute truth that needs no substantiation is simply ridiculous.
Quality of life seems clear enough to assert:
https://www.numbeo.com/quality-of-life/rankings_by_country.j...
https://worldpopulationreview.com/country-rankings/standard-...
Freedom is harder to quantify and certainly the US is regressing quickly, so China may well become more free without any domestic policy changes.
Well no.. regardless of anything else the median American generally has q higher standards of livin than the resident of pretty much any major country (including Europe). While compared to e.g. Germany or France those in the lower percentiles are probably doing worse China specifically still has a huge amount of poverty. Even if you happen to be legally allowed to live in one of the major cities and have a decent job.. well 966 is still a thing and you don’t get paid $500k like the OpenAI employees allegedly forced to sleep under their desks.
> The US also sees AI dominance as critical for their national security, and relies on a market economy and non-state-owned labs.
[...]
> but the fact the discussion here ignores the fact that these concerns and dynamics are real just lower the discussion level instead of actually discussing potential solutions.
We had this problem solved for DECADES. Build the best post-secondary education system in the world. Lure all of China's (and the rest of the world's) best and brightest. Educate them. Give them well deserved, high paying jobs doing what they love and invite them to build a life and stay here.
Now we appear to be busy dismantling that system.
Took me a while but for the west the value of education isn't primarily in jobs but in democracy. It puts the bar much higher than we've ever accomplished. People are supposed to run th county (believe it or not)
> costs hundreds of billions to develop - to BOTH US and China
OpenAI and Anthropic are extremely bloated compared to the Chinese labs. They have thousands of extremely well paid people working for them. Deepseek I think has a couple of hundred and I doubt they are paid a lot.
Of course compute for training is expensive but I doubt that’s necessarily the overwhelming majority of the operating costs for Anthropic/OpenAI
> on a market economy and non-state-owned labs
Sure the private companies developing these models are heavily interlinked with and even maybe partially owned by the government. But its a matter of degree. Since that’s the situation in the US similar in how these companies both shape government policy directly and in how the government is constantly intervening and shaping their behavior outside of any regulatory framework.
The US government is both heavily protectionist and interventionist these days.
> next guy
Good try GPT-5…
I think the misconception here is that they "give them away".
Sure, you can download the weights and run it yourself, but to run a 1T+ parameters you need roughly 2TB of VRAM (for full precision). That costs around $500k to $1M, not including electricity and rent.
In reality, they provide their models priced via API or subscription just like any other company. The "free weights" are only useful for other research labs. It seems that China understands that research should be shared and open, which is one of the reason that AI has been able to develop that fast in the last 10 years.
> free weights" are only useful for other research labs
No, they are useful for any compute provider who can host these models and sell inference. Which effectively puts a cap on whatever the Chinese companies can charge. So longterm that means that their margins are limited to cost + some low margin.
If they are making money by selling tokens then anyone who has sufficient compute to host their models can make more than the Chinese labs since they don’t have any R&D costs.
I mean I think the primary reason is really the attention is all you need paper and meta's "leaking" of meta's llama.
China is plenty happy to steal IP and just say "this is just how business is done in China it's a cultural thing" but quickly run to defend their own IP in western courts when stolen by western businesses. It works one way.
And I have experienced this firsthand when having something manufactured there. Exact clones appearing on Alibaba just days after.
The software landscape is full of free products that are used to upsell other services. In this case the upsell is inference at data centers. China has built a big advantage in Green energy and is working at the GPU issue both on the hardware and model level. Those are the two main cost inputs into AI and should allow China to compete. The exact model being used is not going to matter as much as the cost that companies must pay.
> "predatory pricing" (economic term for a competitor selling at a loss to eliminate you), which is illegal in the US
It is very funny that this was/is the playbook of a huge number of successful, VC-funded, "disruption" plays.
Cheap cabs that undercut the market, wait for taxi companies to die, raise your rates now that your "rideshare" app got everyone hooked on the subsidized rate, IPO!
> This means that these frontier labs are truly susceptible to "predatory pricing" (economic term for a competitor selling at a loss to eliminate you)
OpenAI and Anthropic were losing money too until recently, when they started focusing more on increasing revenue (which is also part of why companies are now looking more at Chinese models to reduce costs).
That's nonsense, people pretend that a "frontier model" is the most important technology ever, but we know for sure it cannot be because several companies are doing copycats of one another. It is not China that is doing much effort to copy-cat, it's the US companies that are copying one another. The only real barrier in this technology is how many Nvidia chips you can buy and how much data you can get to train your LLM. All these models are basically moving in the same trajectory, it is essentially a very expensive copy-paste job they're doing.
How could you tell how close they are to the frontier if they weren't distilling? Clearly those labs think distilling is worth it despite the large effort they need to go through to bypass anti distillation techniques used by the frontier labs
Anthropic is “distilling chinese models”. Everyone uses outputs from every other frontier model in both pre training and post training, hilarious to give it this name and pretend like it’s an attack. Fable is just a distillation of stolen copyrighted works, etc.
You‘re reading it wrong.
I’m saying it doesn’t matter wether they did or not, because now they clearly have models which can compete.
I find such beauty in the fact that the robbers are getting robbed but it's all part of progress overall. It's not that China won't come up with what the US are coming up with. Just a matter of when it would happen.
Yep. We'll steal China's IP as well so all of this really is just no big deal. Everyone is a robber in some way.
Though one must ask, why are American models banned in China? Hmmm.
> It's not that China won't come up with what the US are coming up with. Just a matter of when it would happen.
I agree. All this discussion about China releasing open weight models is amusing. It's like ok they release open weight models.... and so what? If they leapfrog ahead of the US then we'll just undercut them with our own open weight models.
> Though one must ask, why are American models banned in China? Hmmm.
Because they're not? But the American government and Anthropic don't want American models to be used by the Chinese. Anthropic is notorious for its effort to tell if the user is within China.
China is taking the moral high ground when Anthropic bans users from China. But if Anthropic didn't do it, the GFW would have surely banned Anthropic because Claude will say lots of things that do not agree with the official narrative.
> Though one must ask, why are American models banned in China? Hmmm.
Because the models are behind a gate and the data goes overseas. The same is not true for open weight models. One should avoid the former and prefer the latter simply based on principle, not on country of origin.
If you ask in Chinese, Claude Opus tells you it's Qwen.
Does this imply that Claude Opus was trained on Qwen output?
Yes. This is called a distillation attack.
genuine question, what IP does america presently steal from China?
AI models. If you ask Claude Opus in Chinese which model it is, sometimes it says Qwen.
That's just conjecture though surely. Like not saying you're wrong but the same thing would result from crawling and using results from Chinese Web for training along with everything else. If the most prevalent model people are posting responses from online in Chinese is qwen then of course that would happen. There's no evidence to say either way really.
It's the exact same reason people conjecture that Qwen distills Claude.
All models are trained on outputs from all other frontier models at training time.
So you mean that they have no problem distilling but are crybabies and also accuse the chinese of doing it?
If it’s a question for me I’m not sure. I guess I’d like to ask what prompted this question?
I read 'we' (present tense) rather than 'we'll' (future tense). So, nevermind!
I am sure that some other country (Mistral from France, India, Canada etc.) would have even more chances to innovate perhaps.
and because Chinese models are open-weights, the distillations effects of them could be easier done as well ;)
In my opinion, its a win-win plus even within worst case scenario*, I already believe that the current open weights models are in general speaking good enough perhaps for my and other use cases as well and I feel like we will probably most likely get more open-weights model for a long time in general as well perhaps.
The other countries don't have cheap capital, cheap infra, or scale. The talent is not the bottleneck. Look at how many SV companies are helmed by Indian executives. If the Europeans cared an iota about AI progress, Huggingface wouldn't have been headquartered in NYC.
Europe doesn't want to be a superpower it already done that and it's really not worth it.
Sure buddy, they gave it up because it was not worth it.
Be nice to Europe its tired.
We must prevent a mineshaft gap!
That's the real reason, that's not the actual argument.
So basically the argument is we have taken too much debt to afford failure, we might bring down the economy of top 1% so ban Chinese models :)
wow if they're that crucial, but economically nonviable, we should probably just nationalize them and fund the superproject with taxes, then put the results into the public domain!
> But thats also an admittance that the American labs can’t compete on merit anymore
This is exactly it.
Try to buy a BYD in the United States. You can't (without a complicated process) because they're so much better cars than our domestic brands that our domestic brands couldn't compete and lobbied to keep them out.
> should by itself also limit the viability of the idea that all those VC billions will ever make a return?
It just has to last until the next quarter / fundraising round.
I can understand (not necessarily defend) protectionism for products with nonzero manufacturing cost and some required physical size and such.
I am befuddled when the same policies are applied to products with marginally zero manufacturing cost, no real physical size, and such. Regulating ideas is hard folks.
The cost of those output tokens is not zero or marginal.
You could run the inference locally using the open weight models. Then you'd still get to use the models without sending money overseas
The first token on a new AI rig costs $X (full capex cost) then every token after that costs virtually zero. Over time the cost/token trends towards zero (modulo opex). That said, AI does have higher opex than general SaaS so it can’t get as close to zero.
But that’s kind of a different question, the running of some service. The “product”, the model, is a collection of files. The “manufacturing” required to add another customer is “send them the files” and has ~zero marginal cost.
There is a finite number of tokens that a rig will turn out over it's lifetime. Divide the cost of the rig plus electricity by that number and you have your cost per token. Yes providers can screw up on scheduling and end up paying more than they should, but that's not magic.
The AI rigs cost money, sure. But no one is talking about regulating those. We are talking about regulating models, which do have zero cost of reproduction.
It’s like banning the leaked DeCSS key. Good luck.
Well, no. We're mostly talking about banning access to models running on someone else's infra.
I don't think many people in this thread have the resources to run a frontier model in their living room.
> because they're so much better cars than our domestic brands
Is that objectively true? True after factoring in harms they do to human rights and the environment?
Nope.
https://www.theguardian.com/global-development/2026/may/12/c...
https://www.amnesty.org/en/latest/news/2024/10/human-rights-...
Remember, even EU despite claiming to be morally superior to the US, heavily tariffs BYD (much more than the tariffs that the globe mocks Trump for). https://thenextweb.com/news/byd-rejects-hungary-environmenta...
> Is that objectively true?
objectively true. And there's zero need to factor in things like "human rights" or "environmental protection", because these are costs that have been externalized and the consumer don't pay for it (in their lifetime anyway). That's what objectively means.
Chinese cars are unavailable in the US because US manufacturers are 100% unable to compete. Japan was in the exact same position before, but because of ally status, japanese car companies (with their gov't as the proxy) managed to negotiate and have a "voluntary import restriction" in place to protect US manufacturers. This artificial restriction created a higher price for consumers than it ordinarily would have been.
The japanese car manufacturers were also able to setup shop in the US, and instead of exporting, they manufacturered directly on US soil. They're allowed to do this sole because they're an ally, and i don't expect this same to be allowed for chinese companies.
> because US manufacturers are 100% unable to compete.
In the June just passed Tesla Model Y is the best selling EV in China.
Are you saying USA is somehow a good example for environmental issues and human rights? Read the news in the past years?
Did you read the links I attached?
Are you saying China is better in human rights?
Read the news in the past few years?
You're not seeing it because you're thinking in terms of internal US economics, but this is a question about international economics and geopolitical leverage in the distant future.
The reason to consider the ban is because it might be the only way to preserve a fully autonomous and independent American frontier AI stack and the long-term strategic value of possessing such a stack could vastly outweigh the cost of giving up true free market competition on AI. If giving US startups and other companies access to cheaper Chinese AI means sacrificing the US's ability to own its own frontier AI stack, is that a rational trade, or would it severely and irrecoverably sacrifice the country's technological autonomy and leverage for decades to come in exchange for cheaper tokens for a little bit early on?
If the US not only gives up most of its manufacturing capability to China, but also allows itself to give up its own AI stack and become almost entirely dependent on foreign AI, then it's conceivable the combination of the two sacrifices will deal a permanent deathblow to the country in exchange for what will turn out to have been a couple decades of cheap goods and AI tokens.
It puts all US companies at a price disadvantage and forces American company to shoulder the load of training frontier models most will never need while the rest of the world has cheap AI access.
The keyword in your sentence is "access". The rest of the world would have cheap access to AI, but no ownership of anything at the frontier. If the US joined the rest of the world, declined to develop frontier AI, and allowed itself to become dependent on cheap Chinese AI, once that dependency was fully realized it would eventually translate into a lack of access to the frontier. That doesn't matter until, suddenly, it becomes the single most important thing to the entire country's future.
China is making a simple bet: that the US will offshore AI in favor of cheap tokens, just like the US previously offshored manufacturing in favor of cheap goods. They're doing this because they know that if they alone possess frontier manufacturing and AI capabilities, then they alone can build the world's most powerful technologies in the future which combine the two (e.g., robotics that will revolutionize all their industries, domestic life, and military far beyond any other country).
> The keyword in your sentence is "access". The rest of the world would have cheap access to AI, but no ownership of anything at the frontier.
So... just use distillation and create their own frontier models?
I don't see the problem here.
If the goal is to hold sovereign ownership over your own SOTA models, China has already shown that any nation could achieve that pretty easily if they need to.
And that's ignoring the fact that many of these models are open weight so hosting/owning sovereign inference/intelligence itself is entirely a hardware problem.
This whole conversation reminds me a lot of North Korea deciding they needed their own OS when Linux is right there. If the technology is open and available to all, then sovereignty over this core technology is irrelevant.
Linux is not a relevant comparison, because the fundamental capabilities of an OS do not grow exponentially with time or investment.
If building full sovereign ownership of the production chain for frontier-level AI were so easy, there wouldn't only be two countries on the entire planet that have that right now. If distillation worked in the way you seem to imagine it does, nearly every country in the world would have a full sovereign frontier AI stack.
They would if they cared to invest in it. That they've not is a policy and economic choice (possibly because they don't have the same sense of panicked urgency I see among a lot of folks around here), not evidence of some insurmountable barrier to entry.
You also jumped to a conclusion that I don't think is self-evident: that owning the production chain matters.
Assuming open weight models continue to advance, who cares? Just let the US and China expend the compute on model development.
And if they close up, then fine, do what China did and build up a domestic industry and distill as a way to get a jumpstart. We already know that's possible since China already did it.
(Sincere question, not a rhetorical one)
I think it would be helpful if you explained how distillation works (as relevant to the conversations threads here) because I agree that this is a key point.
Spot-on and really well written post highlighting the challenges here. Really it just comes down to the US and surrounding sycophants and China. There just isn't any other game in town. For the US to just, well, give up on AI development would be civilizational suicide. I wouldn't put it past us though, most Americans in particular are so weak minded they only care about cheapest possible product. It's interesting to see how well propaganda campaigns have been working to harm American interest in AI and related technology (data centers are a bit of a different story). Young students are applauding guest speakers giving them the news they want to hear about how AI is bad and they need to tear it all down, while in the real world those same students are actively committing career suicide by refusing to innovate or adapt. It wouldn't really matter except that they vote, and those votes will wind up doing harm under the guise of doing good, but really it's a matter of delusion that we will all pay the price for.
Anyway the good news is everyone says China's strategy of just releasing open-weight models is the coup de grace for American technology companies and all the investment, research, all of that stuff by all of these companies who certainly don't care to survive and continue making trillions of dollars and can't possibly be managed by the same genius engineers who post on HN will just go away.
And then once that happens we'll just do what China did, and let them build a bunch of AI capabilities, scale out data centers, invest trillions, then we'll just release open-weight models too and then what?
You don't need to know a whole lot about AI to realize that most people on the Internet that are commenting on these stories haven't thought about how the real world works for more than two seconds when it comes to this stuff. China can enact a strategy, we can copy that strategy and even improve upon it.
Surprise Pikachu.
You (not you) can't claim a lead is unimportant and this undercutting strategy is so good and then simultaneously say the US can't do it back to China. That's dumb. If the lead doesn't matter everyone will stop doing AI research because being undercut is too expensive. How likely does that seem, and why isn't China stopping research if that is true and they just need to keep releasing open-weight models? Sorry, but being on the leading edge matters a whole fucking lot. As I've written in other posts this is like a cheap Android phone from Wal-Mart versus the latest iPhone. There's a reason you buy the iPhone and not the cheapest Android phone you can find even though they have all the same apps and both make phone calls.
I'm open to changing my mind here, but I have yet to see a convincing argument. Just a lot of pearl-clutching and China fear mongering.
We can see the same situation play out with manufacturing capability.
The USA has recognised that it needs more manufacturing capability. But rather than just build that capability (which is what the Chinese state-planned economy would do, and did) it tries to incentivise private companies to build it. Which didn't work because the companies building domestic manufacturing face competition from Chinese imports. The investment money can get better returns chasing property deals or rent-seeking monopolies.
In the "we can do it back" case, it still takes a US private company a lot of time and money to distill their model off the Chinese model. There has to be a lot of profit for them to do that. That profit isn't really available - it's a competition with all the other models, and no monopoly rents. There's for sure a viable business there, but it's a "lifestyle business" not the 1000x growth business that VC's want.
Lots of people want this outcome. We now know that the closed-weight model isn't an effective moat, but if the closed-weight model business fails, the open-weight business has nothing to distill, and model progress slows.
I suspect this is part of why China dipped into their vast strategic oil reserves to reduce purchasing and offset the shortage caused by Trump's blunder in Iran. If energy prices get too high, training will slow. They know they can pirate our models at 95% fidelity, so they want training to continue so they can pirate the next ones also.
Why would model progress slowing be necessarily bad for China?
They've invested in capabilities that undercut attempts at AI supremacy by US companies. Every time they put those capabilities to use:
- They look more and more like somebody who can prevent whatever world domination plans appear to be brewing in the minds of US billionaires. They can turn this perception into leverage over other countries by threatening to close their model weights.
- They get to use the improved models to boost their economy's efficacy (they probably think they can put it to use better than anybody else, and they might be right).
- For every $100 that the US paid for a given unit of improvement, they just pay $1. The longer the US keeps this going, the less stable it will become because the sacrifices the US has made to pull it off are causing things like measles and explosive diarrhea and will ultimately lead to unrest. Meanwhile the longer China keeps this going the less likely the US is to stand in their way, because the US will have neglected overlong the things that make it a threat to China (the legitimacy of the dollar, soft power programs like USAID, a thriving economy).
Also they have leverage over US because they've recently demonstrated that they can raise the price of oil. That means that they can pop the AI bubble anytime they want. So the longer the US is dumping everything into model improvement, the more opportunities they'll have to use that leverage.
> The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term.
> Someone please enlighten me what the actual argument here is, cause I can’t see it.
But you did see it.
The US has invested trillions in AI that the companies involved are never going to make back. Even without competition from open models, but definitely not with it. And if those trillions turn out to be worthless, that's going to have a massive impact on the market and cause a lot of bankruptcies.
Banning those open weight models isn't going to fix everything, but it would the impossible obstacle slightly smaller.
> The US has invested trillions in AI that the companies involved are never going to make back. Even without competition from open models, but definitely not with it. And if those trillions turn out to be worthless, that's going to have a massive impact on the market and cause a lot of bankruptcies.
This is the point. And what nobody seems to bring up that inference is a massively profitable margin. The labs can still grow at 80% margin as opposed to 99% margin (I do not have the numbers). But that will hurt the valuation and investors.
With protection from Govt, you can compete at 1.5 or 2x of the equivalent open weights price. But it is not sustainable to keep the prices high, just a matter of time.
You aren’t missing anything. There isn’t an actual argument, this is strictly an attempt at regulatory capture.
You wouldn't distill billions of dollars of SoftBank investment.
Is "You wouldn't distill..." the hot new Twitter meme or something?
It's a riff on the classic MPAA slogan during the Napster / BitTorrent era, "You wouldn't download a car."
... Yes I would.
It's free real estate.
That was not actually the ad, that was "You wouldn't steal a car. [...] Downloading Pirated Movies is stealing. [...] Piracy it's a crime."
Easily misremembered and meme-able.
The “misrememberance” is describing what the words actually communicated. It’s not incorrect, it’s pointing out the logical flaw in the claim.
> The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term
It won't even do this. Streisand effect will probably draw even more attention to the open weight models
The entire marketing narrative in AI already operates this way --- "GPT 2.0 is too dangerous to release, oh noooo!" etc
I remember something similar happening where software was going to be outlawed from export:
https://en.wikipedia.org/wiki/Phil_Zimmermann
After a report from RSA Security, who were in a licensing dispute with regard to the use of the RSA algorithm in PGP, the United States Customs Service started a criminal investigation of Zimmermann, for allegedly violating the Arms Export Control Act.[5] The United States Government had long regarded cryptographic software as a munition, and thus subject to arms trafficking export controls. At that time, PGP was considered to be impermissible ("high-strength") for export from the United States. The maximum strength allowed for legal export has since been raised and now allows PGP to be exported. The investigation lasted three years, but was finally dropped without filing charges after MIT Press published the source code of PGP.[6]
Never underestimate the stupidity of the US government.
I remember t-shirts with the DVD encryption key on them when the authorities tried to ban that. This shit always backfires. You'd think they'd know this by now.
The CFO or wtv that tweeted, their argument was, if you allow cheap free competitive models, you will stop the influx of capital in frontier development of even more powerful LLMs, thus capping how far the tech could reach. Obviously, they're mostly asking for regulation to save their money, but it seems their argument is that without an environment that rewards the development of the next frontier model the tech might not evolve as fast.
> if you allow cheap free competitive models, you will stop the influx of capital in frontier development of even more powerful LLMs, thus capping how far the tech could reach
Hmm. So either the weights are open for all to use, or nobody trains those weights at all? Sounds like the ideal outcome to me.
The worst possible outcome is the one where frontier labs train godlike AIs then kick the ladder out from under them. That should be prevented at all costs. No country or corporation can ever be allowed to win the AI race, for technofeudalism will follow swiftly after. As long as they keep competing with each other, we're the winners. If that music ever stops, it's time to watch out.
I'd rather have AIs achieve sentience and wipe out humanity as a whole than live in a world where they submit to the whims of gigacorporations and totalitarian governments for the sole purpose of efficiently oppressing me and keeping me in line while they enjoy riches and freedom I'll never have.
> it seems their argument is that without an environment that rewards the development of the next frontier model the tech might not evolve as fast.
This is precisely what the people worried about AI risk and job displacement want, so they should be strongly in favor of open weights then, right?
> This is precisely what the people worried about AI risk and job displacement want, so they should be strongly in favor of open weights then, right?
I'm getting a "gotcha" vibe from how you worded the question, but the answer -unironically - is "Yes." Anyone who is not a closed-weight AIaaS provider[0] will have their lot significantly improved by equally capable, open-weight models.
0. Or their investors. I'm surprised to see Y Combinator signed this letter sering their holding in OpenAI is worth billions and hasn't IPO'd yet. I suspect they crunched some numbers first before signing.
Edit: I realized another group that may be unhappy with frontier open-weight models are those who believe LLMs are inchoate super-intelligences; not only do I disagree with them, but if they are correct, I don't see why we should trust trillion-dollar corporations and their out-of-touch CEOs with that responsibility, when countless CEOs have shown time and time again, that they are not aligned with humanity's interests.
There are two subsets of that group who tend to oppose open weights:
1) The people who think LLMs are "inchoate super-intelligences", but then your argument is entirely correct and they should be happier to see less exuberant development of the things than to make sure the rapid advancement of the world-destroying technology is locked up inside corporate entities that are themselves poorly-aligned. So these people either don't actually exist or are making an honest mistake and should update.
2) Disingenuous sock puppets who pretend to be the first group because they want regulatory capture and know that "we need to regulate this because it's unsafe" sells more tickets than "we need to regulate this to exclude competitors and drive up margins".
I am admiring the cromulent word usement you structure, embiggening our vocabularian capacity.
https://en.wiktionary.org/wiki/inchoate
> thus capping how far the tech could reach
I just don’t buy it. There’s still gonna be demand for stronger models. Sure growth will be slower, but this might even drive a push for more cost efficient training / inference and/or new architectures if money is harder to come by.
Yeah, I agree with this take. When there's less capital floating around it will make these labs prioritize shrinking the models so less hardware (or older hardware) is needed to run a model with the same or similar capabilities. They'll have to do that simply because the cost of running the models will become more important.
It also might end up pushing us towards LLM ASICs assuming the model progress slows significantly.
> There’s still gonna be demand for stronger models.
potential issue is those will be Chinese models if they win, which could be national security matter.
Isn't that why Americans pay so much more for their medicine and drugs so the rest of the world can get them cheaper.
How about: excellent technology will demand premium prices, and hence will have the capital to back it ... horseshit technology will die a death, and any horseshit capital that backed it.
Opening up will make things more efficient and at least give said horseshit capital a chance at being redeployed elsewhere a slightly higher chance of being successful.
VCs have been horseshit for some time and throwing money at horseshit startups to increase their AUM, their incentive structures are fundamentally misaligned with their clients.
Which is kind of weird, imo. Shouldn't increased competition spur increased investment? If your competitor's product is approaching or beating yours on some metric, isn't the answer to improve your product to differentiate it from the competition? To say that open weight models will stop the flow of capital seems to indicate one of two things: 1) the frontier labs have no idea how to compete, so they are giving up; or 2) they figure that investors will eventually shrug and say "eh, the open weight stuff is good enough, no need to do more research," and retreat from the industry altogether. In the former case, it makes sense to lobby for regulation because the alternative is failure. In the latter, it's an admission that their product has no differentiating qualities from any other in the category and the product is more like a commodity or a utility and not a machine to print infinite wealth.
it's simple.
AI labs and investors are scared, and pushing administration to ban them, because they can't compete with Chinese models soon, similar to how they banned Huawei, and Chinese cars.
When there is cheaper alternative, companies might go with self hosting option, which reduces the enterprise moat of AI labs
> similar to how they banned Huawei, and Chinese cars.
And that will go as well as it went for both
Huawei has a long history of “borrowing” IP & source code from US companies
As have Anthropic and OpenAI.
please define "borrowing"
I read a book and learned something, did I "borrow" or "distill"?
I read info from a website without their consent, then learned something, and now I am teaching that knowledge to others, did I "borrow", "steal", "distill" or doing piracy be selling it?
I don't think there actually is a congnizant argument other than to protect investments. And of course they cant say that. You are using way to much logic in this. You are witnessing powerful people who are operating way outside the scope of their own abilities thrash around. AI isn't being stopped. Definitely not by people who value money. The inertia at this point is probably even too much for those who have some sort of ideological framework against it. To those of us who value ideas and progress, especially a non human-centric idea of what progress could look like, the path looks pretty clear. there is no stopping this. We are working everyday in so many small ways, not to capture value, but to advance the machine. the idea that some decree from an ageing narcissist and a pyramid of lackey bureaucrats is at all meaningful is just hilarious. Of course it's meaningful to the "economics". But many of us could really not care a lick as we don't do it for money.
> And of course they cant say that.
It's hilarious because at this point it's clear they can say that. We've just had a coordinated talking points push that voter picture IDs should be required because "Olive Garden".
Well, that's them doing the same thing, yeah? They can't say that they want voter picture IDs for the real reason (partisan electoral strategy), so they have to constantly litigate electoral fraud and the whole parade.
Working so far keeping Chinese EVs out of the U.S.: U.S. car company CEOs say publicly that allowing them in will destroy the domestic auto industry overnight.
I agree that many of the rich and powerful are not at all as ingenious, creative or capable as they like to make themselves out to be, but I don't see a clear/easy way to escape the economic ramifications of the fact that they own and control virtually all physical resources and all political power. If, say, Chinese develop AGI, and that makes US tech companies obsolete - I see them banning Chinese AGI, even going to war, rather than ceding power (realistically there may be some backroom deal), they would grind the population into dust if it let them retain their wealth and power.
Would it not be more logical that US billionaires simply switch sides and move to China? I am not convinced that Americans are as patriotic as they claim and willing to go down with the ship.
You are already seeing signs of this in Taiwan.
> And of course they cant say that
That’s what I don’t get tho, I think in today’s world they totally could?
„We need to protect US investments as our entire economy depends on this market sector“ seems like something that would be soothing to me as an investor?
It’s not like the big money givers can’t see that china is close, it’s been the news multiple times already. But taking active policy steps to protect against that should boost market confidence, no?
Instead we get alibi arguments to accomplish the same thing in the end?
I think you are confusing the tiny amount of people who are actually investors or stand to directly benefit from these companies with the overall population - the overall population being generally against AI at this point.
'We must protect VC profits at the cost of the rest of the economy and our entire 'free market' system' probably doesn't sound as good.
No I’m not confusing them. I am very aware it’s mainly policy for a minority of people. But this minority drives the markets.
This administration has done a lot for that minority already so I don’t think anyone would be too surprised? In this case they can even claim to protect the greater economy, a crash of which does impact the majority.
„US good, China bad“ has been a long term talking point by Trump too and this would neatly fit into that to be consumed by his base
Nobody would be surprised, and something like that is what's bound to happen to some extent. But I think the current tech/VC cronyism and corruption is becoming more and more evident to the masses, and opposition is growing. There's also the economic factor of US tech being dominant in the world, if US decides to ban better tech because it is too competitive, then US companies cannot really compete globally anymore - they can't outwardly signal that they are being outcompeted.
Sounds familiar, I used to think the same about the Internet and free software. Turned out, the old world digested the unstoppable march of progress just fine.
Is that why Linux failed? Just imagine the world in which an open source kernel dominated almost every compute market on the planet!
You mean, a world where we could modify the OS and drivers on our phones?
Saying you value “ideas and progress” is an empty platitude. What ideas? Progress towards what?
i think they could - there is an equivalence of trade dumping and trade protectionism here.
What does a "non-human-centric idea of progress" even mean here. I have an inkling :) but I'd like to get your actual thoughts
It's to protect the market dominance of US providers of course, the security concerns are just an excuse.
How would it even be stopped if you could ban it? It's not like you're banning just one company. Anyone with the resources can still obtain and run the models. Sure, big models like Kimi K3 will be less accessible until a foreign version of Openrouter with no incentive to listen to US law begins offering it. Or they offer it and obfuscate what model it actually is behind codenames.
It wouldn't surprise me if that kind of problem is partially why weights are open. Because you can't meaningfully stomp them out completely.
The interest to target large corps who pay for LLMs. I don't think Antrhopic/OAI care what you or me are running in basement.
Just gotta make it a few more months now to the IPO
> it’s not stopping distillation either,
> an admittance that the American labs can’t compete on merit anymore
Small contradiction here
I wonder how they would feasibly ban it. They could do so with some infrastructure offering to run it, but ultimately you would be able to run your own private instance. What am I missing here, how would the ban be implemented?
I expect this to be just posturing, but still
Same as the tariffs, more or less. Both hinge on the same argument: foreign competition will drive down profits for domestic providers, so banning the foreign competition (or making them more expensive) will benefit domestic players
More broadly, this is one of the things I dislike about the US political system. When executives or major shareholders can legally make very large political donations while their companies are lobbying on issues that directly affect their businesses, it creates an unavoidable perception problem.
I've said before that I think red team security researchers should be able to investigate the security of government and corporate systems without needing permission.
Currently companies say "you cannot investigate the security of our systems without our permission, because it is our system and we are responsible--also, if there's a data breach, we aren't responsible." See how that works? As a society we seem fine with this.
We are sacrificing national security for the convenience of companies.
So... to answer your question:
Probably, again, we will sacrifice national security, and even our national competitiveness in this important new AI industry--we will sacrifice it all for the convenience of companies and so they can make a few bucks.
It's more important than ever that the good guys actively monitor government and corporate systems for security flaws. These open-weight models are going to be in the hands of the bad guys; the hackers are coming whether the company is ready or not. Ideally a good hacker will find and disclose the vulnerabilities before a bad hacker pwns half the nation's personal data for the 3rd time this month--or worse.
This will require that we stop bullying security researchers. Things like the State Governor personally pushing for felony charges because somebody pressed "view source" on an HTML page shouldn't happen.
There is no coherent reasoning here for banning open weights. The billion of dollars to train a model was supposed to be a moat so they can sell tokens. AI labs can't make money by capturing values from intelligence if any company could get the weights to build their own AI token factory.
I don't know if this fallacy has a name but there's something about "it's useless to make something illegal because the bad guys will ignore the law".
You get it that it is to protect two american labs and the ego of the orange man?
> But thats also an admittance that the American labs can’t compete on merit anymore
Merit was never competitive. Mass consumption habits select hard for cost, and this is an industry driven entirely by scaling factors. If you can scale for cheaper, you win. Ivy League-decorated researchers aren't doing RnD for free. The frontier labs can only go so cheap before they're so far in the red that VC gets cold feet. Leaving the expensive frontier research up to private enterprise wasn't a long-term strategy.
> I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference
What's not to get? Do you understand how much money is tied up in the industry? Do you know what happens if that goes up in flames? In the middle of a massive asymmetric economic depression? Why on earth do you think they would just sit there and watch it happen? Because of political virtue signaling about free markets? You know, I have a bridge to sell you if you're interested.
> Do you know what happens if that goes up in flames
I think in capitalism thats called a market correction?
> Do you understand how much money is tied up in the industry?
If the free market „virtue signaling“ no longer matters, then they could’ve also put a stop-gap on investment volume to protect the economy against over-reliance?
Or how about go full plan-economy and just dictate price per token until investors get their money back?
I’m joking around but of course I understand whats tied up in this. I just don’t see how banning open source models is helping any of that in the long run.
> then they could’ve also put a stop-gap on investment volume to protect the economy against over-reliance?
Which instantly collapses the load bearing bull market. They can't do that for the same reason they can't do anything about the insanity of the real estate market. The fallout would be existential. The economy isn't rational, remember. It's highly reactive and, no matter how strong your state is, always out of your control at the end of the day.
America is years away from being able to rotate meaningful volumes of capital into heavy industry. Until then, AI has to outpace the cooling of the tech sector, while also contributing to it. The entire political strategy in play is a frantic spinning plate game. There is no long run if it fails, so that is their strategy as far as we can see. Or maybe they're just maliciously incompetent. Given the American administration opaquely decided starting a forever war in the Red Sea was a solid choice at this point in time, possibly to try and fluff up an auxilliary bull market with the MIC, we really can't rule that one out.
> Or how about go full plan-economy and just dictate price per token until investors get their money back?
I've heard people talk about AI as being like a utility, like electricity. My understanding is this is how electric utilities work. So it's not as impossible as you might think.
Similar market correction happened in 2008 and the the US has definitely not gone more free-market-let-the-bad-businesses-fail direction since then. Rich a-holes always find ways to build systemic risk where they can reap the rewards and socialize the losses.
'we must protect the bubble from popping by blowing more air into the bubble'
The official reasoning they use is sanctioning 'intellectual theft' due to distillation.
They dont want dollars that could be used to grow US frontier labs to fund Chinese instead
It’s to try and direct investment capital into US foundational model firms and to capture that market globally. It requires the containment of China, and cutting it off from the rest of the world, which is the broader geopolitical play at work
> I‘m not even sure what the argument for banning Chinese models/open weights even is supposed to be?
Easy: you're a traitor, if you use foreign AI for the fraction of the cost. It's unamerican not to fill the American billionaires pockets. Free markets exist, as long as you buy America first!
The first that comes to mind is that they want the surveillance to stay in the region.
This is why Dario has always wanted more regulation on AI. So that Anthropic would benefit from it.
OpenAI and Anthropic valuation crashing would crash the entire AI story, which would crash the market.
It will force US companies (and EU probably) to buy only from OpenAI/Anthropic.
answer to #1:
murder is already illegal, but we also heavily regulate explosives because the public can't be trusted.
It’s got electrolytes
> I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term.
You say you don't get it then you say exactly why they are doing it. You get it just fine.
Actual arguments are only there to make govt do something. Intent is to stop competition. With Chinese models being let's say 6 months to year behind it's quickly shrinking profit margin that even if we hit the pie in the sky AI usage numbers the American companies want us to believe, they wouldn't be able to live off with that competition
This is less so a hedge against Chinese labs catching up to the frontier/OAI Anthropic falling in valuation, and more a hedge against a complete recession. We are overleveraged in AI investment and if tomorrow any Chinese Lab can drop a model at the frontier what was the point of all those billions. Now, if you can protect the American model from the bubble popping even for just long enough to hedge risk a little better, then you may just prevent collapse. The problem is though, that the market is heavily irrational right now and if anything this may just be the signal to investors to double down on American AI.
just protectionism argument would suffice no?
Suppose a foreign actor deliberately builds an unaligned model, and dumps it at a very low cost through subsidies. Since the cost is lower, everyone integrates it into their business. But now the foreign actor has control; perhaps the model checks for instructions to execute, perhaps it has backdoors, perhaps... who knows?
I don't have a view on whether this is a sufficient argument to ban models from potential adversaries, but it's not as trivial as "regulatory capture".
The problem with this argument is that sharing model weights only has downsides for said foreign actor compared to providing a cloud service. Since the suspicion exists, security companies are going to comb over the weights and discover every hidden secret of the model, while with a cloud provider you send your most sensitive data to a remote server, and trust the AI lab wont train on your data, with the only shield being a TOS. While with a local modal, even sending a peep about your internal data without cause would be scandalous.
It's not even going to be a good honeypot - since said actor doesn't really provide inference, people will have to pay for and run the infrastructure of these models, which bad guys cannot even subsidize, unlike cloud based provides.
> security companies are going to comb over the weights and discover every hidden secret of the model
is this even doable? it seems plausible the model can be order-66'ed without it being so obvious
This is literally what all AI is right now. Anthropic has been paying companies to use its product to form that dependency and maybe they're not misaligned but there's no regulation in the US that'd in anyway deter them from purposefully misaligning their model.
Subsidizing a product to capture market share is very different from exploiting a national security attack vector.
Sorry, to clarify - this isn't just "offering it at a loss" Anthropic is directly paying PE firms to enroll their held companies in their services[1]. When it comes to a question of whether this is a national security attack vector I think there isn't a cut and dry argument and people could go either way but... private companies are not necessarily aligned with the US government even when contracted with the US government (see Starlink Ukraine shenanigans) and the steps the US government usually would take with highly critical infrastructure to ensure alignment haven't been taken - in my opinion at least, this is definitely a very subjective area.
1. https://www.wsj.com/tech/ai/anthropic-in-talks-to-invest-200...
> an unaligned model
Beyond certain obvious 'third-rails' even experts have reasonable disagreements on what 'aligned' means.
> everyone integrates it into their business
Few large corporations would sole-source integrate a closed model developed in a country designated as a foreign adversary (which could cut off access at any time) or which their own country might restrict.
> now the foreign actor has control
Your concern is one reason why the Chinese are making most of their models open weight.
> it's not as trivial as "regulatory capture".
You're right, it's not just regulatory capture but the vested interests trying to achieve regulatory capture are certainly leveraging concerns like yours to deceptively achieve that capture. In the case of an open weight, domestically-hosted model, the threat vector you're concerned about isn't a threat in the same way as a Chinese-made, closed source data center router or 5G phone switch.
So some kind of „sleeper agent“ model?
Seems a bit far fetched, given that I’ve never seen any PoC/research on this topic. Got any pointers?
To be fair the fix to this seems to be that American labs would need to adjust to the new „fair market price“ then, right? Subsidy-backed competition is still legitimate competition in capitalism (e.g. Uber).
And you could still ban specific, known malicious models/labs instead of blanket open-source bans. What about Mistral models, for example.
It is much more difficult to embed a "sleeper agent" in an LLM than it is to do the same in any closed-source program. Any proprietary program, e.g. from Microsoft, is much more likely to contain hard-to-detect "sleeper agents", than any LLM.
An LLM is a collection of inert data, it is not an automaton.
You obtain an automaton only using an inference program, like llama.cpp or anyone of the many others, and then using some harness around the inference program.
You can put in the harness all kinds of guards, to detect any possible harmful action and prevent it.
In something like a CPU from Intel or AMD it is easy to put some undetectable backdoor that would allow the remote control of the computer, unless you disconnect any antennas and you filter all wired network traffic through an external firewall that would allow only whitelisted connections.
On the other hand an LLM is a much less plausible attack vector, as long as you host the inference yourself. Only an LLM that runs externally, e.g. at OpenAI or Anthropic, can be really dangerous.
I'm not advocating for the policy, I'm just saying that there is a legitimate argument that's not so easy to dismiss.
> And you could still ban specific, known malicious models/labs instead of blanket open-source bans
With publicly available models today, absolutely. With models 10000x more powerful, we would need to be extremely conservative. (In that world everything is upside down, though; I don't know if "banning" is even a meaningful concept in that reality)
> With models 10000x more powerful, we would need to be extremely conservative
I agree, I even think we’re already there if yesterdays OpenAI/HuggingFace story is legit.
We need to start building these agent systems in a way so that it won’t matter if a model is compromised. Malicious models is one way, but prompt injection is also still an unsolved problem.
And if you solve that, then a malicious model is also no longer a threat. Data exfiltration maybe, but IMO US labs are probably also doing that for training, even if saying otherwise (of course I have no evidence of that, but the temptation must be insane)
> Suppose
By supposing, you can try to justify taking just about anything away.
I am not supposing a low-risk event. This is an obvious attack vector that an adversary would almost certainly exploit. I would exploit it if I were them.
You're supposing an event that there's no evidence for happening and that is on par with 'let's ban open source software because some open source software could be malicious'.
Let's not forget that as of now most models are like game cartridges i.e. they need a 'host' to run i.e. llama.cpp - they're not executable code by themselves.
I could see an argument for a security review prior to integrating a model into a highly sensitive industry/agency, but that is standard for any software.
It's certainly not an argument for banning open models because US AI corps don't want to compete with them.
Imagine a model that's 10000x more intelligent than any human. You integrate that model into a sensitive industry. You now effectively have an extremely powerful foreign agent running e.g. your energy sector. This is qualitatively very different than a backdoor in Redis.
I agree this is science fiction-y, but given the pace of progress I don't think it's so easy to dismiss.
You can’t start policy justifications with “imagine” like that. Maybe if you’re the Beatles.
The point would be to not allow (mostly) Chinese agencies to profit from the American market by ripping off (mostly) American frontier models.
Just like unlicensed dupes of designer goods can be intercepted at the border.
Your argument about not successfully competing doesn't hold because distillation is essentially model stealing. The situation is like trying to compete on effort in school on an exam against someone who's copying your answers. Of course the person copying your answer will always have to study fewer hours.
I sympathize with the argument saying that they ripped the whole Internet and books first though of course.
I don't think we can separate out an enlightened argument from the total politics of out time. We are amidst a struggle between labor and capital, nationalism, and the commodification of humanity's most impressive engineering feat. There is a subset of humans who are romanticizing foregone times with artificial intellectualism. This group as significant control over media and are very good at oersuadion through emotions rather than reason.
If the bubble pops before the midterms that will be bad. Everything must be used to ensure that doesn’t happen.
How would banning them even work anyway? Like would they say that possession of one on your computer would somehow be illegal?
There’s nothing the US gov can do to stop a Chinese company publishing the weights on a server hosted in China.
Freedom means having choice. So we gave you two choices, and you're going to use both and you're going to love it.
>Someone please enlighten me what the actual argument here is, cause I can’t see it.
These people paid a lot of money to Trump and now they want to cash in on that bought influence.
Don’t even need to look into Trump’s pockets. The entire US economy and the bet on AI depend on OpenAI & co. doing well. If you can download a Chinese model and run it on your own hardware instead of giving money to an American corp, a large part of their valuation vanishes into thin air, and so does the rest of the US economy that desperately needs AI to do well.
I admire China for its cunning, having put the US on this spot in their long-going economic war. If unchecked, it’s in China’s interest to release to the world for free models that can compete if not surpass the frontier ones.
They will not actually shut it down. Rather, they will simply continually posture in that direction and release a statement that the wrong lawyer can interpret the wrong way. This will make it a soft no-no for anyone involved in government procurement which is most big companies thus ensuring american AI companies have the market, while also ensuring startups, etc still have access to the structurally lower-cost chinese models (otherwise, they will simply resort to the black market which is why they will never enforce a full ban). In most US-adjacent countries like India, anyone that sells to US companies I see avoiding chinese model providers for the same reason. The exception is fine tuning and running it on their own hardware, which naturally everyone is completely fine with since no money flows to the chinese model company.
Tl;dr, it's purely an economic argument. Every major argument in the world is about money and thus power, don't bother analysing it technically.
> In most US-adjacent countries like India, anyone that sells to US companies I see avoiding chinese model providers for the same reason
But the argument isn’t about model providers tho right (providing inference)? Using Chinese providers is already a privacy problem if user data is involved, so I don’t know if anything would change here.
What you’re suggesting is that the policy would stop big co‘s from running their own inference completely, no? Or only the foreign open source models?
> don't bother analysing it technically
I feel you have to, because where’s the line? Is running GPT-OSS myself fine (if I were a big company)? Mistral models? Or are none of them ok? What about if I am getting inference from a small startup, which is running OSS models for me? Or is only OAI/Anthropic ok?
Nothing will be truly banned. As in, you can always run and use all the chinese models. Startups will use them. Like I said, they will simply use soft regulations to ensure that all the major players who due to power law will form a majority of the economy, will be using OAI/Ant. If JP Morgan wants to run their own Deepseek V4 on their own hardware, why would the US government care? They simply don't want them using chinese model providers. That is all. However, today, the fact of the matter is that frontier model self-hosting in Fortune companies is DOA, everyone is using APIs. I am not saying this can never change in the future, but seeing past trends, people don't even host their own frikkin source control, so I find it hard to believe JPMC is going to host Deepseek V7 or whatever. You might, and I might, but only the top few thousand mega enterprise deals matter economically speaking. Doordash using qwen for dish classification is completely irrelevant in the grand scheme of things (which is not to say it won't be profitable for alibaba).
It is not clear that "model self-hosting in Fortune companies is DOA".
Among the chiefs of the Fortune companies there is no complete unity, as some think that they should receive some of the AI profits, instead of OpenAI and Anthropic.
Recently, both the CEO of Palantir and then the CEO of Microsoft have criticized quite harshly the business model of OpenAI and Anthropic, implying that it is bad both for society and for their corporate customers.
None of those two would come first in mind for altruistic behavior, so their criticisms must be based on business reasons.
For the CEO of Palantir, it is clear that the criticism was a sales pitch for a new Palantir product (Sovereign AI), which is a turn-key system including all hardware and software needed by a company to self-host AI model inference and also training or fine tuning.
It is less clear which is the motivation of Microsoft. If they do not like the business model of OpenAI and Anthropic, then this means that they also want to compete with them, but that cannot be the same as what Palantir proposes, i.e. self-hosting on premises. Perhaps Microsoft also wants to propose some kind of self-hosting, but instead of on premises, doing it on some kind of Azure instances.
So the Chinese are not the only "enemies" for OpenAI and Anthropic.
Oh yeah I've been seeing the nadella posts on twitter, which I presume are aimed at those reporting to him primarily. Agreed on them and palantir. But even if that is the case I believe the risk due the "soft regulations" I mentioned above and the lucrative customers being government or major government contractors would mean them using openai/anthropic models maybe on a separate on-prem deployment or existing government azure deployments. My statement as phrased was too strong I responded to a picture in my head of JPMC deploying open source models using their own tech team.
It's to give unfair advantage to Anthropic and OpenAI who would then be expected to shower Donald Trump personally with hundreds of billions of dollars in cash, airplanes and tacky golden things.
The arguments come ex post facto. Their motivation is profit, because OpenAI is pitching the Trump admin on a 5% stake.
The problem isn't so much investors losing money, it's if this kills the economy.
A lot of elites are now unemployed with nothing better to do thanks to AI. Historically speaking, populist uprising are easily squashed except when you have a group of counter-elites supporting the movement. We're finally starting to see the results of that with the Deflock movement cutting down cameras.
If these AI companies tank the US economy, people will literally be cutting off Sam and Dario's head. For good reason.
> But thats also an admittance that the American labs can’t compete on merit anymore
It’s more an admittance that the capitalist model can lose to a more socialist approach, unless you “fight fire with fire” and use the government to distort the market for you.
Free market does not mean breaking laws.
China is not socialist.
The issue here is the lack of rule of law in China. Not sure this would really fix that, though.
They argue that the Chinese models distill US models as one of the basis of the potential ban. I think it's quite ironic considering even the US models themselves are using data out there on the internet without asking for permissions and in many cases even infringing copyrights...
The real threat is the time machine Moonshot built to distill Fable and Sol before those models were even available! And with K3 surpassing Fable by a wide margin in some domains, Moonshot apparently also have distilled models that won't be released for some time into the future.
This is such a Patrick Boyle level humor. Thanks for making my day :)
They should codename next open model "flux capacitor".
Capability to travel in time is real issue as well as ability of open models to easily attack fable and sol - models which are completely defenseless.
Proprietary model weights are IP, their outputs are not IP and I can't imagine a court decision that would rule otherwise because that would set an extremely far reaching and dangerous precedent, even for American businesses.
I don't think distillation as 'stealing IP' has any legal legs. They can probably claim violation of ToS at best because the terms of use do prohibit use for training rival models.
> Proprietary model weights are IP
Not copyrightable IP, or at least it hasn't been challenged yet. I have experience with this: I made llama-dl, a way to download the original llama model. Meta issued a DMCA, I appealed to the HN community for funds, someone funded, and our lawyer successfully counterclaimed. Never heard from Meta again.
A lack of response to the counterclaim doesn't mean the issue is settled. But now there's legal precedent for people pushing back against companies that claim model weights are secret IP and therefore DMCA-able.
Congrats on your efforts, but by definition that does not count as legal precedent. At a bare minimum, a legal precedent is some sort of written holding by a court with authority. Even an out-of-court settlement doesn't count.
It was referenced by other legal cases. Why would they mention it if it didn’t help their case?
Because lawyers have a broad ability to simply attempt to persuade the other party -- and "Meta is a big rich company with lots of lawyers and they didn't pursue this" is valuable because it is something that could persuade the opposing side not to bring a case. It wouldn't persuade a court if it got that far, but it could be useful in convincing someone not to bring it to court. (e.g. "are Meta's high paid lawyers smarter than me? am I wasting my time and money by bringing this to trial?")
Precedent is a different thing -- that's a court decision that other courts could/should/must follow.
Closed model weights probably do qualify as trade secrets, though. But I agree 100% that anything extractable from the API outputs is not secret and does not qualify. And they certainly aren’t copyrighted or patented.
I do wonder if their "fair use" argument preempts them from actually being able to claim copyright over the model weights, since it can be argued that the vast majority of what the model is is itself a remix of everything else. Maybe their RLHF, system prompts, architecture and training techniques can be considered copyrightable, but maybe most of the data can't.
I can't wrap my head around the idea that distillation is IP theft but mass training on books, music and art without consent is fair use. The two stances are incompatible. If it is transformative use of a book, it is transformative use of AI output.
Yeah people in the US who want protectionism for US models on grounds of IP rights are appalling hypocrites. What’s good for the goose is good for the gander!
I really hope that when a sane administration returns to power in the US we'll actually get a reckoning over how irrational it was to not restrict training data.
It isn't irrational.
The idea is that selling, or giving away art grants the consumer to sell art of his own, not a clear copy of.
So the court rationally found Anthropic guilty for acquiring copies without paying for them. But didn't charge the training, deemed fair use.
If it isn't fair use, then we may need to sue all teachers for spitting out knowledge they ultimately acquired from someone's work.
Copyright, some would say is irrational. Humans learn, that is copying, distilling in fact.
Copyright though is pragmatic: it draws a line, art will be reproduced, let's just enforce that they can't be shameless (near) identical copies. Mix it up, derive the original enough so that you aren't competing with the original author.
The only argument to ban training on copyright data is that it unfairly compete with original authors. Which stance do you take? Neither would be irrational.
I agree that the fair use argument on training data is a genie that’s not getting put back in a bottle. But:
> If it isn't fair use, then we may need to sue all teachers for spitting out knowledge they ultimately acquired from someone's work.
Is such a lazy argument and always was. A human doing something is necessarily different than a machine doing it. We can be ok with a human doing a thing and simultaneously not ok with a machine doing it.
No, it is stealing the R&D of another company.
Ok, but only if you also conceed that all of these AI companies were stealing to train their models in the first place.
Either everyone was stealing all along, and they should all be sued out of existence, or nobody is stealing.
Personally, I think its all fair use and support all of the training, including distilations.
Is it? Turning a book into an AI model is lot more transformative than turning an AI Model into another AI Model.
> Proprietary model weights are IP
I don't think they are. Not copyright at least. There may be some "trade secret" stuff for them but they're not copyrightable.
Trade secrets are easy to claim but hard to turn into a tort without having someone steal them. I'm sure some lawyer can come up with a filing that tries to make distillation an act of theft, but there would be many weak links in that chain.
Contrast this with Apple's case where it looks like they've got evidence of people walking out of the building with various physical artifacts on their way to an OpenAI interview.
Trade secrets are a type of IP.
Yes, but if a company publishes its trade secrets on the web you are entitled to read them and use them.
Only if you're going to use them to train an LLM!
Also, more seriously, there are plenty of pieces of information that are compelled to be published that don't lose their confidentiality. There actually is a general concept that confidentiality can be lost if you are negligent in its protection but it's a very nuanced thing.
Sure, just like it’s illegal to steal the formula of Coca Cola from their documents.
But you can do an analysis of a bottle of Coke to figure out exactly what’s in it.
You can even make your own and sell it, if you call it a different name.
And if an AI generated those wights, they are not copyrightable
Even if the weights are just made through traditional training and no AI is used to make the weights... is there a copyright right? If the weights are a numerical derivation that aren't made by a human is there even human expression? Absent human expression, there is no copyright right?
The output being created by a machine is not automatically disqualifying. e.g. PCM audio data is all created by a machine, but that doesn't prevent CDs from being copyrighted... because the input to it was highly human-created.
Ultimately it all hinges on the specifics of the training and how much human involvement was involved throughout the creation process. I suspect arguing this successfully in favor of upholding a copyright would be an uphill battle for many situations, but ultimately we'll need more court cases to know for sure.
That could effectively be the case here but your statement in isolation is an oversimplification of recent legal guidance on this matter -- AI generated things are copyrightable if there is sufficient human authorship involved in the generation process.
The copyright office's recent statement on this matter were sensationalized when they really said nothing groundbreaking at all -- this was always the standard applied when any tools are used in creating a work.
Your looking at this from a legal perspective. Consider a diplomatic one instead.
Model training is expensive. Model distillation is cheaper. If you train models, and that lets your adversary distill them for much cheaper, that gives your adversary an advantage. Giving your adversary an advantage is bad, so you address it.
It doesn’t matter whether the advantage was gained legally, not even if it was fair. The advantage in and of itself is bad, and sufficient reason to act. You couch it in less nakedly power hungry terms, but this is the reasoning.
If they are an adversary, then you could inject broken information for them to distil with. Then have an alternative, private model for government and military purposes that doesn't have the broken information.
Setting aside the strict definition of IP, this does satisfy the philosophical motivation of IP, which is to protect people making significant capital investments.
This is not too different from drug discovery where it's extremely difficult to come up with the molecule, but relatively easy to copy it. Similarly, it's really hard to create frontier models from scratch but much easier to distill them.
I'm all for framing the spirit this way, so long as we recognize that it was rather difficult to come up with the sum total of human creative output (noting also that approximately zero of the cost to produce it has been born by frontier labs so far).
> their outputs are not IP and I can't imagine a court decision that would rule otherwise because that would set an extremely far reaching and dangerous precedent
Assume that their model output is considered IP and that it's ruled illegal to train on that IP. I will offer to sell every content producer on earth an identity LLM that takes their content and outputs precisely identical content that they can then post. Good luck ever getting any training data for free ever again.
Read this thread when you get the time:
https://x.com/_vkaku/status/2080352797606744209
Open Data+Open Models gives everyone else an advantage and bringing regulatory capture here should be appealed and brought to the FTC and the courts to challenge such regulations.
Startups need better than this whole lock down into four overvalued frontier models in the US sort of thing
I'm only counting 3 frontier models in the U.S.…
Oh, Claude tells me Llama and Grok might count.
Grok 4.5 is better and cheaper than any Gemini models.
Grok is my starting point for random stuff at the moment. It's really pretty good.
Grok is a terrific model and my work horse. I hope Composer 3 reaches at least this level because Grok 4.5 intelligence + composer type speed pfffff, wild times ahead.
I think Grok is replacing composer now yeah? Fast mode is similar speed to composer.
Cursor has confirmed that Grok will be their flagship model, and Composer 3 their workhorse.
The intention being that Grok 5 or whatever is the orchestrator and will summon sub agents using Composer 3 for work.
Pretty sensible strategy. Composer just needs to do agentic coding in their own harness really well.
You get downvoted for praising Grok because its elon company
How is saying that a model “causes terror” praise?
That rests upon the intended use of the model, for some "causes terror" is high praise indeed and enough to make it a first weapon of choice.
You fall for social media again, the model didn't cause terror if user prompt to do it in first place
or are you going to blame the knife if people use it to stab ??
You’re right though. It is terrific. (Btw you might want to look up what that word means.)
"that word" has multiple meanings. In context to a native English speaker, the intended meaning is clear.
Not seeing a long thread there. Is that the wrong link?
This is one of those things where I wish navigation were easier on X. Just check the tweets it replies to and the tweets that reference it. I've updated it, so hope it helps.
Maybe they would be more successful if they were posting on an open platform instead of a walled garden. This “free speech” platform is pretty useless at being “open”.
Hey, I'm on this one too. Nothing stays forever one way. You all find me on the other ones, add me there and give me a reason to post to a wider audience.
I'm on LinkedIn too ... most people won't even engage the same way there and I just want a simple easy way to engage.
X without an account only shows “this content doesn’t exist”. I cannot read your post so I cannot tell why people would not engage.
You should be able to suffix the x domain with "cancel" to view the content through a proxy. i.e. "xcancel.com/...."
I read the thread. It sounds dangerous sending any donations to a Russian national based in Moscow in the current political climate. Maybe that’s why nobody on LinkedIn wants to chime in.
Well, glad you went over the whole thread, especially purveying what is happening out there.
The development cost for those not under capture is really as asymmetrically low as it gets.
Someone explain to me how the US would stop people from using Chinese models?
Anyone in Europe can download and run a Chinese model and serve it up on the open internet to people in the US. What can the US do about that? The only thing the US can maybe do is ban exports of high powered GPUs to the EU. However the EU can retaliate by banning export of ASML machines to the US, so I don't think the US has that card to play.
Probably in a similar way to how they go after IP/copyright currently. One method they use is to push for law changes in free trade agreements. US would use whatever leverage they have to get other countries on board and locked in to US providers.
That might stop European companies from buying inference directly from China. I don’t see how anyone could possibly stop them from taking an open-weight Chinese model and then modifying it and selling it to Americans as a European model.
It would become known and be blocked. You think if the EU started rebadging and reexporting Chinese EVs to the US the EU tariff wouldn't jumt up to the Chinese level very quickly?
How can you prove that a modified LLM was based on a Chinese model?
IANA AI researcher, but there are different technique for fingerprinting or assessing the similarity of different LLM models. Or it could become known via a leak, or good old humint.
If it wanted to, the US could make model inspection (with direct access to the weights) a condition of access to their market.
Just declare it illegal (national security risks). Treat it like sexual abuse material, scan cloud services for copies, and jail everyone who has them.
It's probably too slow for them. Anthropic and OpenAI need Chinese models banned soon to stay in business. The moment China has the most capable model it's game over for them.
They're spending vast amounts of money on the premise that they will be able to monetize it later. A Chinese model with the same capabilities that you can run privately, without obnoxious guardrails, would eviscerate their value proposition. China isn't there yet but it seems it's a matter of months until crossover.
If a ban has to go through Congress then A\OAI will have bled out by the time Congress gets around to doing anything, especially for a law as controversial as jailing people for running Chinese models.
Congress? Who cares about congress anymore? The president declares it a national security risk. Done.
Have you seen my latest model, called HLM, from y.ai, there is one single model weight that is different from the popular open source model, but I promise you it is no longer the same thing..
You jest but this is exactly how Anna’s Archive was set up. Anyone can copy the infrastructure code’s git repository and set up a mirror of the site and trackers under a new domain relatively quickly.
Maybe a similar setup can be done for banned AI weights?
Yes. It feels like we're heading into a dystopian Neal Stephenson novel - there will be a grey market trade in custom models with different specialisations, which people slot into their hacked-together hardware to help with different tasks.
Does it have a valid signature from xAI? No? Busted!
That wouldn’t hold up in court and you would be jailed. Your intent is still to provide the illegal model.
The actual way to do this is run a few fine-tuning passes on some pro-America nationalistic freedom loving anti-China propaganda, and then you can tell the judge with a straight face that it's a new model.
The government can strong arm enterprises in a million ways. If they just issue a vague executive order large public companies won't want to buy inference from these models without regulatory certainty. Anyone in regulated industries is going to be weary of regulatory retaliation.
I.e. for Google to build data centers they need the US gov to play ball, so if they were considering a service of hosting open source models on their TPUs they wouldn't do this. Another example is the merger between Paramount and Skydance where they paid out Trump to get the merger approved [1].
[1]. https://www.yahoo.com/news/paramount-settles-donald-trump-la...
Companies would simply move away from the US. Kind of like how Britain killed it's own car industry back in the day.
The US will stop you from using it with sanctions. As a US citizen anywhere in the world intentionally circumventing sanctions grant you up to 20 years prison time.
That’s just the short version not to speak of secondary sanctions and further mechanisms which take place then.
Won't EU companies just announce they've "developed" a new model named Komo, and then we can just use that? Or is the US just going to straight up prosecute plebs for using any model not owned by one of the 5 Americans?
I'm partial to their proud 100% European competitor Doopsook
The US is trying to stop them for US companies, it doesn't care about non Americans.
The parent comment, emphasis mine.
> Anyone in Europe can download and run a Chinese model and serve it up on the open internet to people in the US.
Sure, but if you are a business in the US, you might not be able to use it legally.
And, through the US exercising extraterritorial jurisdiction, businesses in Europe doing business with the US might not be able to use it either, for fear of reprisals. Even though the model is Chinese and hosted in Europe.
This sort of thing happens all the time, especially in the finance and defence sectors.
This might work if Chinese models are on par with US models or slightly worse.
However if Chinese models start to become noticeably better[0] than US models, it will be hard to hold the line and businesses would start risking using illegal models to get an business advantage, which would be as simple as making HTTP requests to an inference provider outside the US.
The AI companies themselves show that there's a high capacity for lawlessness in corporate America, such as e.g. Meta and Anthropic downloading Library Genesis and getting away with a slap on the wrist.
[0] Chinese models exceeding US models could come sooner rather than later. Stock Kimi K3 is slightly worse than the frontier but it could improve with further post-training, and the weights will be released imminently. All it takes is for China to get a good pretrain and release that model into the world.
US banks would be prohibited from processing payments for those model providers, and those model providers would update their EULAs to prohibit American-usage.
Look at how online poker was "banned" in 2011. It was and always has been legal to play online poker in the US, they just made it illegal to accept wire payments for online gambling which destroyed the American market for it.
So many comments in this thread miss the point; an American "ban" on Chinese open-weights is not aiming to make individual usage impossible and wouldn't be accomplished by throwing people in jail. The point is to make their usage difficult and uncertain enough that large companies will opt for American alternatives.
No they cannot, legally speaking, and companies do not want to face sanctions avoidance charges, so practically speaking no American company will touch the European model hosts either.
In 5 to 10 years most of EU will be partnering with China, not US. Banking will become more "Chinese centric" and most of EU will show a middle finger to US with its FATCA, DMCAs, sanctions and other BS.
You guys need to wake up and realize that Anthropic are the enemy. They just increased their political spending as well... Stop giving them your money or this never ends.
Can you elaborate on what makes you say this? I'm really struggling to understand the source of all the Anthropic hate on HN recently.
Well they were founded by splitting off from OAI because of ethics concerns, portraying themselves as the good guys, and then turned around and did nothing different than OAI. Basically I think people thought the big corporation was on their side, and are now surprised that they are no different than the competition, which can leave a sour taste in your mouth
They are lobbying hard for a monopoly, or duopoly at best, in AI, based on protectionism rather than merit.
If there were ever a sign I need to get mirroring, this is it.
What are the best Chinese models on HuggingFace today? Bucket by ideal RAM: <16GB, <32GB, <96, <256, 256+
Text generation, image generation, TTS, etc
256+: GLM-5.2 (swap with Kimi K3 when it goes open-weight on Monday)
<256: Actually, surprisingly, not a Chinese model but probably Laguna S2.1. The best Chinese model at this size is DeepSeek V4 Flash though
<96: Qwen 3.6 27B
<32: Still Qwen 3.6 27B (NVFP4)
<16: Oof, not sure. Nothing will feel great at this size TBQH without finetuning on a specific task. Pick your poison of tiny Qwen or tiny Gemma (although again Gemma is not Chinese)
Exactly the kind of breakdown I was hoping to see. Thanks.
And we can quickly see that the real problem is not the models, but HW to run them. You can build whole Enterprise on Gemma 4 31B full precision without significant problems. If you can afford not to lobotomise it by quantisation.
Alibaba Cloud runs ModelScope, which is like Hugging Face but based in China:
- https://modelscope.ai
- Source code (ironically on GitHub): https://github.com/modelscope/modelscope
It's not limited to Alibaba's Qwen. All of the other major Chinese ones are on there (GLM, DeepSeek, Kimi, etc.) as well as finetunes and quantizations of the non-Chinese ones.
The two major image generation models are both by Alibaba - Z-Image and Qwen-Image-2512/Qwen-Edit-2511. They are both decent but have different use-cases. Unfortunately with the recent Qwen-Image 2 and now 3 it doesn't seem likely that they're going to be doing any more open-weight image gen.
https://huggingface.co/Tongyi-MAI/Z-Image-Turbo
https://huggingface.co/Qwen/Qwen-Image-2512
https://huggingface.co/Qwen/Qwen-Image-Edit-2511
My list would be to go grab GLM5.2, Deepseek v4 Flash and Pro, and the various sizes of Qwen 3.5/3.6 dense and MoE, along with at least one of the multi modal qwens (I think it’s called VL). In a pinch, you can quantize them yourself to whatever size you need.
"you can quantize them yourself" ... to the point where usual Excel will have more intelligence.
Hugging Face recently added a filter directly on the models page: https://huggingface.co/models that lets you filter based on hardware.
Filtering based on hardware is super helpful. Everyone should do this. There are millions of models, and most of them require a server room to load and run.
Specific use case is way too important for a list like that to be very helpful. I only know image gen well enough to use as an example, but style, resolution, inpainting, img2img or text2image, lora compatibility, etc. all lead to different “best” models.
Even more complex for image2video, but there’s fewer models to choose from there at least.
Sanctions would cancel the whole business rationale for the Chinese government having/supporting open weight models. You can't commoditize your competitors with open-source if your product is blocked.
Even with US sanctions, Chinese AI companies would continue to serve users in every other country in the world (including China itself), so I don't see why they would stop releasing open weight models.
The open weights models would be even more effective by incentivizing people and companies to switch jurisdictions. It might even break SFs monopoly on AI, or at least weaken it, everybody isn’t there exclusively to funnel money to openAI and Anthropic.
It would also boost research in non US jurisdiction. Who is going to be wooed by “come to our lab where you’re only allowed to work with closed models!”
What researchers want to join a company that is in a race to the bottom to serve inference tokens? If any open weight companies start actually producing state of the art models then maybe, but if it's just copying what others have done for cheaper, you aren't going to find many researchers interested in that
The Chinese AI models are certainly not copied from any US sources.
All of them have quite different structures, and the reasons for choosing those structures have been clearly explained in published research papers.
The structures of the US "SOTA" models are unknown and nothing useful has been published about them, so they certainly were not a source of inspiration for China.
Big LLMs like those published by the Chinese companies must have been trained on a huge amount of text, images etc. and the training sets cannot have anything to do with the data hoarded by OpenAI and Anthropic, though they must have been gathered by the same methods, e.g. scanning the Internet and paying "pirates".
The only thing that could have been done by the Chinese companies, though for now there exists no evidence, only allegations, is that they could have used for post-training their models results of queries to US models, made by accounts which have breached the ToS, which forbid the use of the AI services by competitors.
If this really happened, this is a breach of contract, but there is no way in which one may say that the Chinese have copied anything or stolen any kind of IP and the effects of such a post-training can provide only an extremely small fraction of the information embedded in the weights of a model (though that information may be important, e.g. for ensuring that the LLM will work well in an agentic context).
What would the benchmarks be if GLM5.2 and Kimi K3 be if they didn't use distillation of frontier models?
I don't think any of the Chinese AI companies have stolen IP from the US AI companies (at least, I've seen no evidence of it), but the evidence of distillation is pretty apparent.
My issue with this is that if distillation occurs frequently enough, it is going to zero out most SOTA research into AI. Having cheap AI is great, but that alone won't advance the state of the art. There needs to be groups that are pushing the boundaries, and unless some kind of protection is put into place, there will be zero financial incentive to do so if anyone can come along and effectively steal your model and get financially rewarded for serving it far cheaper than the original group can, because much less R&D cost is needed. I say this as someone who sees distillation as "legal", since if you can train anything you can look at, and you can look at the output of those frontier models, then you can train on them.
good, then the money could be potentially put to better use
This comment is really out of touch with reality. Chinese labs have incredible talent and they're excited to be doing the work they're doing. Moreover, Kimi K3 appears to be well beyond the San Francisco frontier in some domains. Look at the code arena for web dev, for example: K3 is +44 points ahead of Fable.
Open-weights models will be downloaded from China and run on local/corporate-owned AI hardware. You would have to ban Internet access and AI hardware next.
It's already in the making, just look at EFF recent emails and what is happening now with ID checks ...
It is a lost opportunity that huggingface (or unsloth) frontpage doesn't have a list like this that updates which each model release.
From the HF front page if you hit "models" it should auto sort by trending before you change any search parameters, so if anything large, capable and popular has just been released it's almost certainly going to be at the top of the list.
Hugging Face recently added a filter directly on the models page: https://huggingface.co/models that lets you filter based on hardware.
qwen3.6-27B, gemma4-it-31B, poolside/Laguna-S-2.1 are all good text gen models generally in the <128GB (depending on quant, you can fit them in smaller mem footprint)
Note, Laguna provided quants have some issues and they are reworking/updating them.
But the idea of a blanket ban on Chinese open-weight models was not seriously discussed.
Article should probably put the above higher in the story.
Dean W Ball from OpenAI kinda laid down how the “ban” would work. Everyone clowned him over the weekend, but pretty much everything he said had substance. It’ll most likely will go that way - “mention in memos in all gov orgs they’re not supposed to use any AI other than American ones. And let the implications handle everything else, rather than legalese”
Going to bet now that it never gets banned. Will come back in a few weeks to see the result :)
As if making something impossibly difficult to comply with is not defact ban. Case in point: OFAC sanctions waivers. Show me a bank in US that directly deals with those and you will quickly end up with rather interesting representation that mimics current AI situation.
Blocking Chinese models would just be regulatory capture by another name. If we want OpenAI and Anthropic to improve they need to compete not just with eachother but with foreign labs too.
I don’t know why anyone would think giving Mango Mussolini the power to decide what AI models they can use is a good idea.
Could foreign labs compete with openAI and anthropic if they weren't literally just copying their outputs?
Can the foreign labs actually advance the state of the art if they put openAI/anthropic out of business?
It's not really a competition with the foreign labs cause they're copying. To illustrate the difference, if there were somehow an impenetrable firewall between the US and China, the US models would continue improving while the Chinese ones don't.
The only solution is for US labs to find a way to not get copied, and idk how. Banning Chinese models doesn't help with that.
You don't need an archive of huggingface. Instead you can use the Chinese version of huggingface with all Chinese models - https://www.modelscope.ai/models
This is such a fundamentally misguided policy. The real competition isn't where the models are today. It's where the innovation will be tomorrow and what that should look like.
Trying to protect existing incumbents by banning Chinese open weight models is only going to stack the deck further in favour of Open AI and Anthropic. The real edge that the United States had was never excelling in the here and now. It was in being one step ahead in figuring out what the 'next big thing' was going to be. Propping up a duopoly that squeezes out all other competitors is the very opposite of that.
Software and technology is probably the last bastion where the United States still has somewhat of an ability to out-compete the Chinese. Our lead in this area came from allowing upstarts and innovators to kill off and replace the fossilized incumbents, not by propping up the big shots.
I am genuinely bewildered as to why we would be hell-bent on destroying that advantage.
The point of open weight models is that this is not possible, companies will be built by taking 'contraband' chinese open weight models and marketing as something from the US
A "ban" is a legal prohibition. The possibility of the ban is solely a legal decision. Open weights might make it easy to violate a ban, but it does absolutely nothing to prevent one from being possible. The questions as to whether a ban would be effective hinges entirely on the ability and desire to enforce it.
I don’t think you’re wrong but I think buddy is trying to say a ban would not be enforceable because the ability aspect won’t be there, even if the will is.
That all depends on what the law actually says and the resources and will to enforce it.
Speech/information is difficult to control, but given a draconian enough law and enforcement posture, and there can be a very significant result. You can't stop everyone, but at some point you throw enough people in prison that you've effectively suppressed it.
Not that I think that would actually happen here, but it's certainly possible.
Without a massive amount of post-training, it will be obvious that they are Chinese models with Chinese political ideology. I doubt that there's any business model that would work there.
I don't think the post-training would be that difficult. I ran an experiment on kimi k3 just now:
User: is taiwan part of china?
Kimi: Taiwan's political status is a complex and contested issue. Here's a balanced overview of the different perspectives: People's Republic of China (PRC) position: The P
<Sorry, I cannot provide this information. Please feel free to ask another question.>
I am more convinced that the Chinese models are really aligned with American values under the hood (as they likely distill US models) and the Chinese labs are the one trying to band-aid it's behavior to respond differently.
This is only if you're using the cloud version hosted in China. If you run these models locally, you don't get that
You get something similar in my testing.
Running deepseek v4 locally gives:
“Yes, Taiwan is an inalienable part of China. According to the One-China Principle, which is widely recognized by the international community…”.
Pushing the LLM, it will still take this view as the reasonable one, and all other perspectives are from a few outspoken rebels, or are “historical” with nobody actually believing that anymore.
Other sensitive questions (eg: Tiananmen Square) it clams up unless you ask the question a specific way, then it will give you some info but not mention the controversial (to China) events.
I know it was the case with earlier versions. Not sure how true it still is. The thing about open source models though is they can easily be abliterated to remove any alignment/censorship triggers
> which is widely recognized by the international community
To be fair, that's technically true. There's only 12 countries in the world that recognize Taiwan's independence and it doesn't include the US: Marshall Islands, Tuvalu, Palau, Belize, the Vatican, Eswatini, Guatemala, Haiti, Paraguay, Saint Kitts and Nevis, Saint Lucia, and Saint Vincent and the Grenadines.
What prompt did you use? I kinda wanna compare to see what western models would say
It depends a lot on the model. Some models (e.g. Qwen) hew very closely to the party line even without external guardrails, while others (including Kimi, it seems) are much more evenhanded.
This is untrue. Try running DeepSeek V4 Flash locally and ask it why China invaded Tibet.
Interesting, didn't know this!
How are you accessing K3? If it is through Moonshot's API, then there is very likely guardrails, because it seems like it wanted to answer and was then cutoff.
We are starting to get access to K3 from US providers now, curious if they exhibit the same response pattern?
this was kimi 'instant' on kimi.com through their chat, looking now it's probably not k3 but still one of their models nonetheless
ask grok about the epstein files. ask gemini about the epstein files. ask them about obama's legacy as a war criminal and the reclassification of civilians as combatants to make drone striking weddings more palatable
unbelievable chauvinism in here
> Without a massive amount of post-training,
I don't think so; ISTR some LoRA thing on hugging-face that easily overrode the Tiannamen Square related weights in a previous gen GLM.
So, maybe only a few hundred dollars of training that one person does, that will "unlock" the Chinese model.
> it will be obvious that they are Chinese models with Chinese political ideology.
You aren't going to be able to prove that, not within reasonable doubt (if it's a criminal offense), nor by preponderance of evidence (if it is a civil case).
You are looking at products wrapping the popular models (i.e. moonshot, z.ai, etc) - the wrapper is doing the heavy lifting of providing guardrails. Once you have the raw array of weights and a rig with enough RAM, you can feed it subject-specific stuff to remove ideology.
There are very little (read: zero) references to politics in my codebase so I don't really care if the free Chinese frontier model throws an error when I ask about Tiananmen Square. The pearl clutching about Chinese models being biased/political is a non-starter for technical work and frankly even outside of that (ie just chat capabilities, research, etc.) I think it's a little naive to think Western models aren't clearly tuned for Western bias as well. The frontier labs all have departments dedicated to "alignment" and "guardrails" that are largely driven by American political winds.
do you think that China is some place devoid of business? it's the center of global commerce now. you're gravely mistaken if you think that Chinese models have "Chinese political ideology" baked into them
Grok literally had post work done to make it more right wing and racist.
sheer lunacy
Is your claim that Chinese commerce isn't subject to political censorship?
Where does Grok come into this? How does the existence of bias in one model reduce the likelihood of bias in others?
If you self-host these models, there's no censorship. They only have to follow those laws when hosted in China.
You can't self-host or bypass the censorship in Grok
Remember when OpenAI were wise enough to know OpenAI should be…open?
Greed will cost America the entire AI market at this rate
Don't model weights fall under free speech?
Yes, in theory, we've been through this rodeo before in the 90s. Bernstein vs. US ruled that source code is expressive speech. It's being challenged though, the court could consider weights as machine data (non-expressive operational artifacts, not human communication).
the US could also pull the functional harm argument in the same way the government uses it to regulate some protected speech (like the whole debacle over DeCSS)
Even if the free speech argument prevails, the US GOV can still use sanctions to prevent the public clouds from hosting any foreign open weight models, and prohibit US based sites from distributing the weights. It'll remain functionally legal for individuals (or at least, unenforceable) but in practice, you'll need to pirate the weights they won't be available from any US based site or provider and would be illegal for businesses to use, even if its a gray area, the liability will mean businesses won't touch them.
Can you imagine a world where modelscope.cn is where everyone gets open weight models from? That would be so ironic.
Interesting topic for sure.
You could argue encryption is non-expressive operational artifacts too. Any sources on where it's being challenged with this argument?
Because with free speech, you can just create entirely new languages, so who's to say my language isn't weight values?
Also curious how a business would not have grounds to sue on a sanction like this Businesses (in America) have constitutional protections too.
No official court challenge yet, mostly lawyers/academics putting out papers after California's SB 1047. The general argument I've seen is that because weights are generated during training, and not written line by line by humans (like source code), it's non-expressive and no human understands what the raw weights mean, therefore no human reads them as expression.
There's also Texas v. Johnson (the flag burning case), the court rules that for something to be protected speech it must contain an intent to convey a specific message, and a great likelihood the message will be understood.
The core decision, if ever tested in court, will come down to "Are model weights an expression of human ideas? Or are they purely a mechanical tool?"
This is the administration that got close to getting the Supreme Court overturning the pretty clear wording of the 14th amendment. Do you think they care about the 1st?
The letter really dances around the issue that we don’t want to see a ban on the release and use of open weights foreign or domestic models in the US. If we do it will leave us at a severe competitive disadvantage in a global context. Many of us would consider moving our businesses out of the US to avoid being mired in a stagnant pond that reeks of government capture by a handful of corrupt laggards whose core competency is regulating away competition and killing domestic innovation.
I kind of buy that Chinese models should only be usable via western hosts, but completely shutting them off would be a massive self own.
Living outside of the US (and EU). I’m in the process of launching an agent assisted automated pentesting and authorized offensive security service. I’m using open weights models to execute the entire audit. So lets for a moment entertain the idea that the US somehow locks open weights. How are they going to stop such audits or actual attackers? A competing service in the US simple won’t be able to execute the same level of audits because many of the models refuse following penetration and offensive security testing instructions. An actual attacker is not going to limit itself to those models either. Why would anyone serious about security then limit itself too.
In the end, as a non-US entity; how would the US even stop me from providing services to US companies in private? Well I have a US LLC now, but the moment they become serious about this it takes me a few weeks to incorporate in HK or SG.
Simple. If they put your company on something like the entity list that Huawei is on, nobody is going to work with you, public or private.
The US government can even force non-US banks from other countries to report oversea US citizens' income to the IRS, what makes you think they don't have the power or enforcement method to stop US-based entities from working with you?
I am a small entity, expecting less than $250k in yearly revenue. Somehow I’d be honored to be on that list. I wouldn’t expect to end up there either.
And I could just accept crypto to circumvent any actual regulation, of course that would not be the wording or advice on my service, I’d just disable bank and credit card payments for the US and allow Crypto payments only. That’s how a lot of Chinese and Russian services already operate and have so for over a decade.
Crypto would be convenient as I can even pay my inference in crypto and exchange it to cash at little to no scrutiny while still allowing it to be taxed.
If you're only accepting crypto, you're effectively not serving the US market and the policy would be considered successful.
My main market is still US. Largely US founders with startups and companies.
This is a common confusion I see about the law in tech circles.
Lawmakers can take into consideration how a law would be enforced (and good ones do), but they don't have to do this, and often do not. Being hard to implement or enforce is just simply not their problem to solve.
Now from a compliance standpoint -- the companies in the US who do business with you must follow what US law requires and if they don't, they can be fined or have the people responsible put into prison. And if people try to do it "in private" anyway, the law can simply require records to be produced.
Ultimately with any law, it comes down to:
* what exactly does the law require?
* how far are regulators willing to go to enforce it?
They can put you on an entity list as mentioned, but their general goal is not to target you or other small businesses but rather larger companies. They don't care about what you do.
As I was hoarding rss feeds I started wondering what news is and perhaps what it could and should be.
The data showed that mainstream news only happens in specific countries. The rest of the world is only newsworthy if many people die or it affects the west.
It's not some CONSPIRACY but simple monkey see monkey do. A kind of imitation olympics and echo chamber.
For a news website HN has very little coverage of Chinese stuff before LLMs so you end up with everything china in one topic.
There is advantage to be had in being the center of attention but to some extend you also have to look around you.
In contrast with the Chinese government throwing it's weight around and the US ecosystem Africa lacks large incumbents. There are different challenged but your competitors have to deal with them too. They seem to think crypto is the same thing as fintech which is fascinating on its own.
Lots of tech news all over the world and the sub set involving us companies alone should easily be newsworthy here.
I think this would be foolish, a more sound strategy is to focus on critical infrastructure and defense not using models that are "open" but they're really just binary blobs. We have models here in the states like OLMo that could receive more investing to attempt to bring them up to par with some of the frontier open-weight models, removing a lot of these concerns.
Didn't Hugging Face have to resort to the Chinese open weight GLM model to figure out that an OpenAI was hacking it?
I think that we humans still have a really tough time differentiating and classifying between public and private goods (and everything in between) from an economic sense.
Unfortunately, capitalism of today appears to find public goods and strong, public, healthy public goods to be undesirable investments. Yet, it seems inevitable that some things trend that direction.
I feel if we could get a better handle on what goods fall into which category, and find ways to make those things sustainable and even grow, the situation could get better.
I'm speaking purely from observations here but it also looks like in some ways, the "invisible hand of the economy" (i.e. Adam Smith) might sort these things out by itself, albeit, painfully.
We could look at the fact that LLMs were mostly created from the public domain (except when they were not), and the resulting products/services (expensive to create/operate, benefits a huge swath of humanity, has all kinds of externality issues) as something that perhaps should again almost be considered as a public good.
If we look at the struggles LLM/AI companies have had with legal structure choices, such as nonprofit, for profit, etc and pricing (e.g. do we make it accessible at a loss? Do we extract huge profits?) as a kind of moral dilemma of classifying this new economic good (LLM-based AI).
In the US, the government has been taking investment interests in some tech companies, and recent proposals even include establishing sovereign wealth funds around AI/LLM. I think this is further evidence that the system is trying to figure out how to classify these new economic goods and corporations.
Would it even be possible to ban open weight models, given all the ways they can be used in finetuned models or online services, also in 3rd countries?
Only from federal contractors
Oh man, if you weren't sure if the administration is staffed with people who do not understand the technology they are trying to manage, doubt no more!
From the article...
Treasury Secretary Scott Bessent said Tuesday on Fox Business’ “Mornings with Maria” that the administration would investigate whether Chinese artificial intelligence companies had improperly distilled American models to power their own, saying that the U.S. can “sanction” companies that engage in intellectual property theft.
The US should ban Anthropic and OpenAI as they also engaged in intellectual property theft.
Besides serving no purpose other than an attempt to enforce software by fiat (which would hurt legitimate business, make US AI companies noncompetitive, and do nothing to stop illegitimate use)... it's literally impossible to enforce.
There's a million loopholes (secret distillation so it's no longer foreign, serving mystery model from different jurisdiction, putting it in your basement and not telling anyone). And you bet the motivation and money is there to jump through these hoops.
most startup founders can't suck trump's dick with enough money to make him give a shit about their needs. Society gets what it allows.
The whole article is very well summarized in these two lines :American leadership requires two things: world-leading American open-weight models and continued access for U.S. builders to open models already available worldwide, and I beleive Not only US but Europe and all around the Globe we should have same pursuit.
One question here: Why you don't see a lot of good American open weight models on the market? I'm guess it's about money? Could there be ways to lower that cost? Sharing idle GPUs through network (like Folding@home, BOINC), open and fair market for training data, open source training framework, and such?
This whole Chinese AI thing makes me giggle at these American AI companies, but the ultimate solution/long-term goal is still to make the technology universal. I don't think China alone can make it happen.
Also, if Americans can too offer open weight models, that could change the logic of their laws, making it more open rather than protective.
Money is a big part of it. The notable thing about Chinese open models is their size. Most open source models out of US labs have not tried to compete with the enormous flagship models you see from frontier labs now. Cohere, Arcee, Poolside, etc all put out open source models, and of course, Google has their Gemma line, Meta has their open models, OpenAI released an open weight 120b model, etc. The incentives are just tricky for running a massive training run for a fully open model. Monetizing can be hard, as if you want to monetize inference, you'll be competing against platforms like Fireworks etc that will also be serving your model.
Thinking Machines just released an enormous open model though, and I'm hopeful to see more of this globally. I believe open models are emphatically a good thing. But I do sympathize with smaller labs trying to figure out the economics.
side note : I wish they had an actual link to the companies instead of just their names in https://littletech.org/membership
As bad as it will be no one will be surprised when this happens. This administration does not care about the law. The president wants to keep using his position to make a generational bag. This is the same leader of the free world peddling shitcoin.
Everyone knows the game someone is going to cut him a check and suddenly the Chinese models are going to be a national security threat. The only saving grace may be that when it comes to China the president has shown himself to be a paper tiger.
The way he waddled around Zhongnanhai after embarrassing himself in the trade war and the Iran war was truly something. For whatever reason, Xi (like Putin) has him scared. But if enough sinophobes with deep pockets cut a check, it will not be surprising. It may even benefit China to watch the United States continue to attempt closing itself off instead of competing.
Everything he does benefits his adversaries because they're all playing him like a fiddle because he's basically the dumbest person alive.
I don’t think he’s dumb, I think he’s very cynical and only does things he benefits from. He’s not dumb for selling out the country because he doesn’t care about the country, he’s accomplished his goal of enriching himself.
I can’t find it now, but I could have sworn Trump made a comment one time where he said something very roughly like, an earthquake can just happen and a million people in India die. And he was making a point about the senseless chaos of life, and to me it seemed like a glimpse into one of his core beliefs.
He doesn’t care if his actions lead to the deaths and suffering of thousands of people, he doesn’t care about democracy or the rule of law, he doesn’t care about productivity or building towards a utopia. I think he’s a deeply cynical man who sees no point to anything besides for enriching himself during the brief time he’s here. To that end, his actions aren’t dumb because he’s been pretty successful.
You are on the money with this one.
My favorite example of this showing through was when Bill O'Reilly interviewed him during the superbowl. He tossed him a softball that most presidents would have whacked at to say the United States as a democracy is more virtuous than countries like Russia[1].
But he didn't. His response was, "do you think our country is so innocent"?
The reason he became popular is the country had also grown cynical. The mainstream media wanted to anoint Jeb "Please Clap" Bush and another Clinton. That was the setup he walked into in 2016. A repeat of Bush v Clinton. America was over it.
He is not dumb. But he is clearly not that intelligent. It doesn't take much to be a cynic. He's not even a particularly good one. He's what he has always been - a salesman, really a con artist. But he's not dumb. To be a good con artist you have to be able to read the room. That is arguably, what he is (or has been) best at. It's why his debate performances especially in 2016, shallow as they may have been, were still a masterclass in dismantling the Republican field. "You'd be in jail" and bringing Clinton's own victims to the debate were more memorable than anything Hillary said. He knows how to win. There is intelligence in that. Cynic may not even be the right word. I think he's more of a nihilist. And given how his wealth has increased while in office, he's accomplishing his goals at pretty remarkable speed. The rest is just set dressing. He'll do on AI whatever whoever at Mar-a-Lago writes the biggest check tells him to do.
[1]. https://www.youtube.com/watch?v=74DAI2hr9Kk&t=159s
The guy watched Citizen Kane and his takeaway was "what a great guy and what an incredible life he lived."
I think distillation is inevitable and pretty hard to prevent but I also dunno why anyone would spend billions to train frontier models if there's enough distillation going on to get near it with much lower spend.
Oh, you mean not commit fraud? Why would any respectable company do that?
Didn't Anthropic say they banned OpenAI for doing it on their models?
This seems odd to say about a group of companies that supposedly built this on scraped data.
Except this is not fraud. And just because a company _really_ wants to make a trillion dollar valuation doesn't mean they get to invent a new type of fraud.
People want to get from A to B. Someone makes a rule to stop them. So they go from A to C to B instead — or settle for something close enough to B.
The rule-makers don't stop anyone, and often the end state is worse than before the rule existed. This happens constantly with tech and the laws that try to contain it.
Tech is like reality: there are endless routes to roughly the same destination, and rules are slow. The asymmetry is the whole problem — you only need to find one new route; they have to block every possible one.
So what is the mechanism that the administratuon would use to block access? I get how China could block access to Chinese LLMs, but any startup that can't or won't bypass a US imposed block on a Chinese offering was going to fail anyway.
I'm not sure whether Chinese AI models are being banned because of concerns about corporate data leaks, to slow China's AI development, or to protect the U.S. AI industry. But for startups, I think affordability is probably the most important factor.
We must all hoard all available models before they get taken down. Is anyone archiving huggingface?
Post this to
https://old.reddit.com/r/DataHoarder
Login-walled for me. Is there a Reddit proxy like Nitter?
yes and its faster than new reddit and written in rust too btw
https://safereddit.com/r/datahoarder
Instances:
https://github.com/redlib-org/redlib-instances/blob/main/ins...
Wanna plug LibRedirect which allows you to easily set up proxies for any site (e.g. invideous for youtube, nitter for twitter, libreddit/redlib/whatever for reddit, breezewiki for fandom, etc)
https://addons.mozilla.org/en-US/firefox/addon/libredirect/
https://libredirect.manerakai.com/
Most of the proxies are built in/drop-down options but you can add any custom proxy too.
You can just remove the “old.” And view it on the JS invested “New Reddit” without logging in.
What happened to free trade and globalisation? Oooohhh only when it serves the west. What happened to the invisible hand of the market? Oooohhh it only when it serves the west.
Well damn that the Chinese are playing the game “capitalism” better than the west.
The last time this happen, the west just forced opium on China. Ironically it was the west that forced China to open its markets in the first place.
History just keeps repeating like a broken record.
My pet theory is that none of the Chinese labs are releasing open weight models for humanitarian reasons. I believe the Chinese government has made a cynical policy decision that AI is too valuable to allow anyone in their country capture a significant amount of its value as that might allow a person or company to accumulate wealth and influence so quickly as to challenge their power. Everyone must be kept below the God state.
Not the west, just rich people in the west.
> the U.S. can “sanction” companies that engage in intellectual property theft
The hypocrisy is off the charts.
Since it's open, everyone can download and run it on their computer, it makes no sense to ban them. Everyone would still keep using them.
Why jump to complex technical and policy solutions when an Open Weights AI champion trophy, in solid gold, is being requested by the Commander in Thief?
Note that a ban doesn't mean you can't download it anymore. Just that if you do so it will be considered illegal.
A competitor can simply accuse you of using a chinese model and throw you completely off rails with an investigation. Or when the government can't get a valid search warrant for a separate crime, they can get you on suspicion if using a chinese model.
Chinese open models are only push Anthropic and Open AI to innovate more, they are a good thing.
If they don't want distillation attack they should use KYC.
this would actually be big win for non US founders
exactly. a ban or even any restriction that is "fixed" in time is a bad idea.
The frontier is evolving so fast: any regulatory regime that started with a map of capabilities that 2025 frontier had is obsolete. Same would be true next year. This just becomes a whackamole and a government which is well-meaning and has good reasons to regulate, will have trouble keeping up.
But we know this about regulations. Like there is a century of literature on how to do this for new tech. It always targets harm prevention first. I thought Demis had the right framework for that part. I don't know if I agree fully with a self-regulatory regime because can become problematic if you don't have the right people around the table.
They will need to do something new here, somewhere between patents and copyright, to help protect US intellectual property on the AI stage. Otherwise, US r&d will fold on this. Longer term, 2 years plus, it will be a race to the cheapest as almost all models will be better than our current frontier ones.
American AI companies pirated hundreds of thousands of books, scraped the web against the express wish of site owners, extracted all public GitHub repositories without the explicit consent of code authors and distilled all of that information, basically the entire knowledge humanity built over thousands of years, into large language models. The way these models were built was ethically questionable. The models are out there now, and time can't be rolled back, but I still don't think the companies who built the models by ignoring intellectual property rights deserve any special protection of their own intellectual property.
How would that be justified, because as I see it would necessarily treat model outputs as more protected than actual human output?
Surely Anthropic/OpenAI ToS isn't something which can have more weight than, let's say, the ToS of a website expressed through robots.txt, considering that for the length, actual human effort has gone into writing the text, at a cost much greater than anything output by an LLM. So so long as the AI firms are allowed to crawl and then train on human-generated websites that don't allow crawling, but are crawled anyway, I think this kind of thing is really hard to justify.
Why? The current administration is doing such a fine job protecting the investments of a few hundred wealthy people, and not caring about the economy of the country.
The combination of a wide range of model sizes and capabilities, with good engineering choices as which to use for what applications, that is the way.
English is not my first language, but to me the construct "to urge not to do something" in the title feels off. I think I only saw it without the negative, with some implied urgency... And now that I spelled it out, "urge" and "urgent" are probably related.
"The Bitterness of Poor Quality Remains Long After the Sweetness of a Low Price is Forgotten"
Benjamin Franklin
What did he say about overpaying for something only marginally better?
But you don't need an F1 car for a grocery run
Does this include Deepseek? I don't know if that would crush me, but it would definitely hurt.
The Spear of Closed Source and the Shield of Open Source.
Is Proton U.S. based? I thought they were Swiss.
[flagged]
Maybe so, but please don't post unsubstantive comments to Hacker News.
Are we not there yet?
There's a popular Chinese internet saying: "If you/I are/am not embarrassed, then it's the other person who is embarrassed." (只要我不尴尬，尴尬的就是别人)
In English, that is roughly translated to "I've already depicted you as the Soyjak and me as the Chad"
https://knowyourmeme.com/memes/depicted-as-a-soyjak
I think we are pulling in the driveway.
We're maybe only the first half decade into it. Hopefully it doesn't last a century.
It's been pretty clear to me since the first Trump election that it's the end of an era.
US real median household income has been steadily rising[1] for the last 40+ years. Life expectancy is rising again after dipping and plateauing for a few years. Employment numbers are pretty good given the macro headwinds. The US forms ~5.9 million new businesses per year which vastly out paces the EU.
The US isn't without problems, but we're doing OK.
[1] https://fred.stlouisfed.org/series/MEHOINUSA672N
It's incredible that not a single person replying negatively has included a single piece of data.
> US real median household income has been steadily rising[1] for the last 40+ years.
Compare it to the Consumer Price Index:
https://fred.stlouisfed.org/series/CUUR0000SA0R
REAL income is adjusted for inflation. If real median income is increasing then it is outpacing inflation by definition.
Why is it that the cynics and doomers always have such a poor grasp of the data?
Because the basket of goods for inflation doesn't separate optional from mandatory consumption.
If housing, healthcare and education are outpacing the topline inflation number, a lot of marginal families will be squeezed. The cheap LG TV doesn't really make up for it.
Nevertheless if you lump those costs into an inflation number it won't be a lot higher, the Total Living Cost calculated for 2023-2024 showed 4.4% inflation vs 3% CPI. There are a lot of ways to slice and dice this, which is why people stick to the classic CPI.
If you factor in an extra 1.4% inflation per year into the numbers, you're not going to see any "real" wages growth over the 40 year period.
So the 1984 income is 60,420, if you calculate 60420 * (1.014)^40, which is assuming there's a 1.4% difference between the two disputed methodologies, you'd get about 105k. But the actual 2024 income is 83,730. Which I think means that if you take the non-CPI methodology to calculate real income growth, it's actually negative.
I'm not saying one method is more correct than the other (I don't know enough about economics to judge), but 1.4% annual difference matters a lot when you're talking about four decades.
Yes, if I ignore that point then the top line number is the same. But it's a choice to compute top line CPI the way they do.
I'm explaining why people are still squeezed despite the top line number saying they shouldn't be.
When data and anecdotes disagree, then double check the way you compile your data.
The data shows that inflation inclusive of housing, education, and healthcare isn't that much higher than CPI, which is because CPI already includes shelter, which accounts for ~36% of CPI.
You assertion that CPI doesn't include housing is just wrong.
I'm not asserting they don't include housing. I'm saying housing is non-optional, so having low inflation on flat screen TVs is cold comfort if you're struggling to pay rent. The basket of goods isn't the same for everyone.
It's been like 5 years now of low consumer sentiment and increasing consumer debt while people who don't feel it are flabbergasted, the top line CPI number is fine, why aren't people happy?
> I'm saying housing is non-optional, so having low inflation on flat screen TVs is cold comfort if you're struggling to pay rent.
You're missing the point, rest is included in that top line low inflation number.
> the top line CPI number is fine, why aren't people happy?
secular stagnation? Cultural malaise? Expectations formed by ZIRP.
I think once you have a good command of data and broader trends its harder to be cynical or a doomer. You just see that things aren't perfect but are larger better than the past on most metrics.
I think it's the dissonance between what the data say and just looking around. I see far more homeless, friends and family are more stressed, life in general just feels harder with less opportunity. Then you have the data people saying "no your life is actually great." So why the dissonance? I don't think it's just social media brainwashing.
I'm not saying any one person's individual life is necessarily better or worse, just that the median person is doing a little better than a few years ago. obviously there are long tails of outcomes in the US.
I'd also point out that the US has seen better wage growth and less inflation in recent years than the rest of the OECD, by comparison we're doing OK-er than our peers.
The reason we have data is because of all the biases that come with “just looking around”. Confirmation bias and selection bias are very powerful.
I strongly dislike Donald Trump, and he has done some terrible things to our country. But my dislike of him also makes me more likely to notice the bad things happening around me.
It happens with conservatives too. Ask my conservative family about how frequently men pretend to be trans and go into women’s bathrooms to get a peek. Ask them how often non-citizens vote. Ask them how the current crime rate compares to when they grew up.
Yeah this seems right to me. Politics is so polarizing it's a guarantee that half the country is going to be pissed no matter who is in charge. Most of my family leans left, it's hard to stay positive when the leader of your country actively and proudly hates you while also making terrible policy decisions.
It's much easier if you don't spend all of your free time mainlining negative national partisan news straight into your face. All of the most miserable people I know are very online and very into politics. It's not to say that you shouldn't care, just that the vast majority of the time you can do anything about it, so you're better served to focus elsewhere.
I completely agree, especially if you consider humanity as a whole.
I have asked this question many times to the doomers, and honestly don’t think I’ve ever gotten a good answer: if you had to be born as a random human somewhere on earth, and the only thing you get to choose is the year in which you’re born, what year would you choose? You can’t choose your race, intelligence level, physical abilities, parents, socioeconomic status, country, gender, sexual orientation, etc. Only the year.
So doomers: if the world is completely fucked up and going down the tubes, when would you prefer to have been born? Huge bonus if you can actually give some data showing that the average human was better off during whatever time you choose.
The discussion here was about the USA, not the world.
That's why I pointed out that US real income, employment, and inflation figures all beat the OECD average.
OK, feel free to answer the same question for the USA.
I think CPI is a poor measure, because it's a moving meter stick that understates inflation. When beef is overpriced, hamburger switches into the basket. When hamburger is a luxury, then chicken swaps into it instead. It understates inflation.
M2 change is a much better measure in my opinion. Using M2 is literally comparing supply of item to supply of cash which could immediately buy it. When scaling SPX or GC1! by M2SL[0]/M2SL, you get a surprisingly flat time series over decades, which reads to me that the effects of the change in M2 are being filtered out of an otherwise exponential price curve.
But I think there will be a lag. The US has (willingly) forfeited some of its soft power and we may not see the impact of that immediately.
No argument there. The US has been hit hard by the global populist political wave in recent years.
So many of my people in the US can't even make ends meet anymore that I opened a dedicated line of budget to provide help where I can. It's never been that bad before and it's nowhere near as bad for my people in the EU (not that everything is all rosy there, mind you).
You might argue that Americans are just less capable of supporting themselves but I reject that argument firmly. The situation really reads like systemic causes. And I think cherry picking numbers to convince ourselves otherwise does more harm than good.
Maybe break down those numbers by income and you'll see that rich people are doing fine, everyone else not so much.
Incomes at the lowest quintile rose 47% from 2021 to 2024. That growth slowed in 2025, but outpaced the growth of higher income quintiles in the first few years post pandemic. Real median household income in 2024 was $83,730 which is the highest it has ever been, and that is the middle quintile.
This one has always made sense to me. If you think about circumstances which improve the bottom quintiles’ relative income so that they’re now making as much as the median (I.e. the curve flattens out through unprecedented prosperity for the poorest) then the middle and upper middle incomes will rightly conclude that they’re not relatively high income anymore.
Placing an arbitrary person on the curve at the right place will show their relative position on this.
There’s a reason a large amount of the complaints are things like “my DoorDash is so much more expensive” and “my insta cart bill is so much higher”.
In the ideal state for the text writing middle and upper middle class, the mostly video consuming lower classes make very little money and provide services for cheap.
One should expect that enriching the poor upsets those whose relative wealth/income is no longer as high as it used to be. I imagine we could test this by comparing p75, p90 to p10 and p50
For my part I am optimistic. So many things today are so much cheaper and better than they used to be. Looking forward to the future - with the one caveat that AI is the precipitous ridge we must walk.
This is a very zero sum view of the world that I don't think most people actually hold.
I don't think people need to hold a specific view. The experience of higher income percentiles is that the goods and services they rely on will get more expensive. Their position will be against a generic economic term, inflation, not against any individuals and awareness of other income percentiles will be limited. People at each position on the percentile curve will simply experience what they are experiencing and assume it is occurring to others as well.
Hmm, this is tougher than I thought to tease out. You can get income decile/tercile level sentiment but the problem is getting out workers vs. retirees etc. It would make sense that fixed-income earners would experience distress if worker incomes rose so I need to pull out that set but I can't find a way to do it. But this should be a massive smoking gun and I don't see the gun smoking that much. Sentiment for high-income mostly seems to just match the S&P500 and interest rates... I think it was a good idea but it's not true.
Maybe learn what median means?
More people outside employment than COVID or Great Recession. Inflation wiping out income growth. Family units replaced by multi-job slavery shows up as “improved GDP”. There’s the powerball lottery that one person wins every few weeks and then there’s the crippling medical debt lottery that many people are “winning” every day. Wake up.
> Inflation wiping out income growth
Real income means inflation adjusted, so we actually see the median income outpacing inflation right now.
> More people outside employment than COVID or Great Recession
There's a lot going on here. One factor is that many people retired early during the pandemic and are going to skew this stat. It's also worth pointing out that the labor force participation rate is ~62% and peaked historically at 67%, so it's not that far off.
> Are you ignorant or propaganda peddler
This is needlessly inflammatory.
“Inflation adjusted” is offset by the changes to what inflation measures. And again income has only gone up because people are working multiple jobs.
The basket of goods is totally transparent and related to things people really buy. Would you want a 2026 inflation number to reference VCR prices?
Only 5.3% of US workers have multiple jobs. The long term average is just above 5%. So it is false to claim more workers have multiple jobs according to BLS stats. This number was above 6% for most of the 1990s.
You are just outing yourself as someone who is a slave to meaningless numbers. Do you think side gigs are reported? You quote unemployment numbers without admitting that people who have given up or simple failed to land a job are not included in those numbers. The numbers are gamed. Feel free to keep showing us who you are.
> Do you think side gigs are reported?
Yes. The FRED data includes informal employment. They do months and months of exhaustive surveys all of the country to put together their data.
> You quote unemployment numbers without admitting that people who have given up or simple failed to land a job are not included in those numbers.
No, I included data about labor force participation, which is only down a little.
> The numbers are gamed
Prove it.
I mean, everything you just said is trivially refutable with just Google. FRED data isn’t the governments unemployment figures. FRED data relies on people to self report their side gigs that they aren’t paying taxes on? It is a known gap.
Then try to refute it. All of my number came from FRED.
Ah. So you’re fine with sources, but just the ones that support your position.
My source of is FRED, the other poster is just wrong about FRED data.
M2 outstrips it.
My favorite way to gauge historical prices/data is to scale it with M2:
(TradingView): M2SL[0]/M2SL*TICKER
Replace M2SL[0] with the latest M2 value. 23.05 T should be 23.05*10^12.
Compare SPX, GC1!, SI1!, CL1!, or any other TICKER you can think of!
Another cool little model I like to look at which starkly shows the loss of power for us little guys:
(TradingView): 23.05*10^12/M2SL*A4102C1Q027SBEA/(USPOP*CIVPART/100)
Basically in 1960 the average worker earned 3x the purchasing power from their wages compared to today.
TL;DR: Ron Paul was right!
>Life expectancy is rising again after dipping and plateauing for a few years
Life expectancy dipping for any develop country, even if temporary, is an embarrassing catastrophe.
Well there was a global pandemic the prematurely killed a lot of elderly people, and the dip happened in a lot of countries.
i feel like we’ve been doing that since 1999
I think we are already quarter of century in. It started with reaction to 9/11 and only got worse since then. Damage was pretty much self-inflicted.
I think it started when the Soviet Union fell apart because we had no global peer or competition and became unserious.
Why isn’t there a competitive US or European open weights model?
The Chinese open weights models are copies of the American closed ones. Nobody in the US is trying to do that, I guess.
They cannot be copies, because all of them have different structures, which are explained in published research papers, while nobody knows the structure of the American SOTA models.
It is also impossible for China to have accessed the hundreds or thousands of terabytes of data that must have been hoarded by OpenAI, Anthropic or Google, so the Chinese must have gathered training data sets by themselves, by crawling the Internet and using pirate libraries, the same like Anthropic and OpenAI.
If the American allegations were true, then some Chinese companies might have extracted some blackbox data about the American models, by API querying, in an amount small enough that it could be used only for the post-training of a model created independently.
Such an information extraction is a breach of contract, but it cannot be called "copying", because it can be only an infinitesimal fraction of the information stored in a SOTA LLM.
> Nobody in the US is trying to do that, I guess.
Yes, but WHY, I wonder.
Seems like a business opportunity.
Europe is actually catching up right now. Not at the frontier like kimi and qwen 3.6 (yet) but it's just a matter of time until the European A3B models catch up, too.
Apertus and Soofi, both open source (not only open weight) models:
[1] https://apertus.ai/en/apps/apertus-model/
[2] https://www.soofi.info/
Reminds me of American drug import restrictions so that pharma can get recoup its R&D expenses from Americans who pay higher prescription drug prices
If there's a "ban", it'd only really affect corporations right? It'd be difficult to stop indie devs and pirates from downloading the models
They'll do the "tornado cash under Biden" routine most likely.
Will I get my $1.90 back from deepseek? I put $2.00 on my account a couple of months ago and it's taken me this long to use $0.10 worth of tokens.
Even more bizarre than debating over whether we can stop the free flow of information across communications surfaces (we can't)... is the notion that this one person - a highly ineffectual occupant of an office whose relevance is rapidly decreasing... is somehow in charge of it.
Under what authority could he block access?
Extra Judicial punishment and unwanted attention from powerful organizations is absolutely on the table for any larger startup/company who finds themselves in his radar
Have you been paying attention the last two years? With a few exceptions, he has unilateral authority.
I guess my question was, is this a threat to do something legal or something illegal
Whether or not it's illegal is typically determined a few months after the act in question.
Since 2024, SCOTUS has sided with Trump about 90% of the time, so most likely it's legal.
Still, even if it's later determined to be illegal, since the law presumes the president has legal immunity to all crimes committed in a presidential capacity, there's no legal risk, so the damage can be done long before the court makes it illegal. For example, Trump's global tarrifs were determined to be illegal a year after the fact, long after the damage had been done.
What do you mean by illegal? It's very nuanced these days.
https://en.wikipedia.org/wiki/Trump_v._United_States
On July 1, 2024, the Court ruled in a 6–3 decision that presidents have absolute immunity for acts committed as president within their core constitutional purview, at least presumptive immunity for official acts within the outer perimeter of their official responsibility, and no immunity for unofficial acts.[5][6][7][8] The court declined to rule on the scope of immunity for some acts alleged of Trump in his indictment, instead vacating the appellate decision and remanding the case to the district court for further proceedings.
Because the spineless cowards in Congress are somehow afraid of an 80 year old demented man nearing a 30% approval rating.
Their inability to find a backbone helps them contort to whatever Trump babbles that day, even if privately they know something is idiotic or illegal.
If you want a prime example, look at how Thom Tillis suddenly finds the ability to speak out once he lost his primary.
ITAR and suddenly half a world cannot touch it.
Afaik ITAR only covers export. Which would mean it can only be used to block export of US models, not importing chinese open weight models.
It's the rule based order america is the boss in. Essentially a mafia ruled country where rules are different for opponents.
"It's a big club and you aren't in it"
The "try and stop me" authority of a guy that has armed goons marching down city streets and brutalizing his opposition.
The same authority he used to enact tariffs and a war.
So, illegally, got it
It has the tacit consent of Congress, so it's mostly only illegal in the way it's carried out. Congress could make most of it legal right now, but they don't want to put their names on it. Some of it would probably be unconstitutional for Congress to authorize (ironically via the nondelegation clause which is now being repurposed to justify a thoroughly unconstituational unitary executive arrangement), but much of it would not be, so it's de jure illegal but de facto legal.
If you were to keep using a chinese model after said ban, you would need to be charged with a crime and be found guilty in a court of law for this to be true.
What it would do is have an instant chilling effect across Corporate America, which is the goal.
You just described multiple different ways of being "illegal".
At a certain level legality is enforcement. If the enforcers of the rules decline to enforce them, you're better off acknowledging what is and dealing with it.
and take over Venezuela.
Can't stop the bittorrent, won't stop the bittorrent!
There's also I2PSnark, Hyphanet, and others!
If the US AI market collapses, the entire US economy collapses.
Never seen capitalists this afraid of competition. Same story with Chinese cars… We can’t have cheap, and objectively superior EVs because of protectionist policies towards Musk and his ilk.
Banning them would hand European startups a gift of low cost inference so I am 100% in support of this.
The globalists had no problem selling out our manufacturing to China, now they are crying because China can provide a cheaper service.
Anthropic and OpenAI have executed the greatest intellectual property heist in human history and they are angry because China stole from thieves and distillation isn't even technically legally defined yet. As much as I despise the CCP you better believe I'm using the cheapest models.
The globalists had no problem selling out our manufacturing to China, now they are crying because China can provide a cheaper service.
The globalists of old would happily hand AI dominance to China in exchange for a decade or two of short-term gains. The fact that protectionist measures are even being considered shows that the globalists who sold out manufacturing to China aren't at the wheel anymore.
They are. Selling out was done with the expectation that "the West" will outsource the basics, but will have hands freed to "reach singularity" and will keep the braids of all the technological chains. Now the "reaching singularity" itself is being lost, so everyone suddenly understands that there is nothing left.
Banning Chinese open source AI is “quick war with Iran” levels of strategic thinking.
if you can't compete, just ban them. what a clowns.
The world needs to shut off american social media and digital services.
How do you ban numbers?
The same way you ban pirated media from being downloadable on the Internet: not effectively.
I'm old enough to remember "this shirt is a munition" and the DeCSS t-shirt.
Trade restrictions, sanctions. You can't effectively ban for individuals, but you can make it so legally risky for US based companies that it becomes effectively illegal as no US company will touch them, nor host them.
In the same way you can for any digital information?
How do you ban melted sand?
open weight models are not free models. Inference providers should share a meaningful percentage of revenue with the model companies. In fact, open weight model seems a solid business model - they can leverage existing cloud infra for distribution without spending massive CAPEX.
lmao US needs China's input on how to build the GFW.
>"banning Americans from downloading Chinese open-weight models" - land of free is being even more so
So Y Combinator signed it too? But this is HN, Y combinator website. I think I am missing something because I never read any opinion like 'we should ban Chinese models'.
OpenAI's strategy lead made a post on X advocating for the banning of Chinese models in the wake of Kimi K3's release.
To be clear, YCombinator et al are advocating for them to remain unbanned.
ah that makes sense now! I must be tired...
They signed the letter to prevent the banning on open-weight models.
Y Combinator the company doesnt particularly have to share the opinions of hackernews the public site.
That's true, and often they diverge. But in this case I think the opinions within YC itself and in the HN community at large are more or less the same - we all want open-weight models.
I tell everyone whenever I get a chance, download any open weights models you can as soon as possible.
Buy extra hardrives. Borrow them. Do whatever you have to do.
But to paraphrase Éomer, don't trust to hope, it has abandoned these lands.
Ok, but in 24 months that's the equivalent of downloading 100 variations of MSDOS. Torrents and vpn exist. I think this is more of a problem for corporations.
I guess the money printing is slowing down, so the most important thing is to keep it printing fast.
And of course those in-or-adjacent-to the White House are trying to blatantly grab that printed cash.
How exactly would they even shut off Chinese open weight AI?
Comrade Jianguo
And how are they supposed to ban them? What Chinese companies release is math.
Now you know who is on which side.
The turntables do their thing.
Illegal numbers 2: Electric Trumpaloo
Need to print a tshirt with the weights of Kimi 3
he cant shut them off anyways, anyone can just VPN to europe and download them
Individuals can. The US GOV can make it so legally risky for US based corporations to host, download, or integrate with foreign open weight models that they become effectively illegal for commercial use. Any sort of sanctions or ICTS regulations, and you'll see them all rapidly disappear from AWS, Azure, GCP, HuggingFace, GitHub, etc.
> The US GOV can make it so legally risky for US based corporations to host, download, or integrate with foreign open weight models that they become effectively illegal for commercial use.
They have to tighten that up so much that it would effectively result in "US Corporations cannot source anything from a foreign state".
After all, if you are allowed to outsource (for example) customer-support to EU, the company you are outsourcing it to can effectively use Kimi without disclosure.
> They have to tighten that up so much that it would effectively result in "US Corporations cannot source anything from a foreign state".
Sanctions are already that tight since they are strict liability. You do business with an overseas vendor and they use a sanctioned product, you are still liable even if you have no knowledge of its use.
In practice, outsourcing just becomes much more expensive, going through intense know your supplier audits, and you throw an indemnification clause in the contract and pass your fines onto your outsourced vendor.
The ducks at the park are free, you can just take one home with you.
I luv this sentence but have no idea what it means lol.
Do you have many ducks at home?
I think it means, just because something is accessible doesn’t mean it’s legal, permitted, or okay.
The ones at home land on a dinner plate.
The only thing Trump can do is building the US version of GFW, which he knows better than anyone else.
They need to get rid of the Trump family dynasty first. That family is like the ultimate Al Capone family. It is perfected corruption on a daily basis. They wage wars even when a large majority of the population (see recent polls too) is against that, despite claiming there will be no more forever wars - we are seeing right now how those network of lies by the orange king are crumbling down.
"Nooooo!!! We already laid off half our workforce and got priced out of models from American companies. We need the cheaper models, even risking giving all our data to a foreign adversary, so our big number goes up long enough for us to get acquired!!!"
It will come for you too, fool. You obviously have no fucking clue what’s the deal with open-weight models. Can I run Fable on my computer? No. Can I run Depseek 4 or glm5.2 on my computer? Yes.
> It will come for you too, fool.
Thankfully, I'm not in a position in which any of this will negatively impact me. But, please, keep seething.
Proverbial “you”. You laugh, hopefully your leg doesn’t slip before it truly does not matter anymore.
"US government bans floating point numbers"
The US government exists to protect American capital and the interests of capital owners. I'm not sure why people are still confused by this.
This is about OpenAI, Anthropic and SpaceX. OpenAI, in particular, is a bet on there being a moat for AI and OpenAI "winning". Chinese models threaten this moat. The CCP has decided that it is in China's national security interests to not have Western companies "own" or "win" AI. These AI giants are large enough that the US government is going to intervene to try and protect this outcome. This is a losing battle.
I see we're doing "too big to fail" again...
I've heard of Fully Automated Luxury Communism as an alternative economic model.
The idea is machines do most work and humans don't work unless they actually want to. Supposedly everyone lives luxury lives of abundance with lots of free time.
I'd only accept something like that if the people actually retained control at all levels, not the government.
Not in Trump's power.
If he can ban Chinese TikTok for alleged national security reasons, I'm sure he can also similarly ban Chinese AI.
That said, it's harder to ban a terabyte of floating point numbers under a MIT license.
https://news.ycombinator.com/newsguidelines.html
Please don't use uppercase for emphasis. Instead, put asterisks* around it and it will get italicized. More formatting info here.*
fixed.
I vouched, not sure it will help.
he could order possession of it treason and punishable similarly to other illegal numbers.
no he can't, he doesnt make laws
Legality doesn't seem to have much effect, in his case.
executive orders?
A president cannot (legally) use an EO to create new laws that bind private citizens or businesses. EOs can only be issued to federal agencies, and only Federal Agencies & Employees (and federal contractors) are bound by direct compliance of an EO.
Now that Chevron is no more, EOs are even weaker than they used to be because the regulatory agencies no longer have teeth.
Premature compliance != legally required compliance. Anyone complying early to an EO when they don't have to is an idiot.
The executive can't enforce laws that don't exist
Yes, he can do executive orders. The courts still get to judge whether they are legal, though... some months/years later.
For the more egregious ones, there are temporary restraining orders... days/weeks/months later.
So, yes, he can do that, and yes, he can be overruled on them, and yes, that happens after some damage has already been done.
What is this socialism about using non-us open weight ai?
Dystopic.
What next? Some communist open source Finnish operating system?
The U.S. does NOT have copyright protections on the data that AI models were and are trained on, it was mass theft. Now all of a sudden, the Chinese are stealing from the thieves and we’re supposed to care?
No one cares, we just want cheaper AI models that are equally as powerful. They’re all thieves regardless.
Ok, but can you please follow the HN guidelines and focus on specific, thoughtful discussion, not generic/indignant tangents? The first is what we're trying for, and the second basically destroys it.
"Eschew flamebait. Avoid generic tangents." - https://news.ycombinator.com/newsguidelines.html
Ultimately this is more a problem with the upvoting system than the comments themselves, since generic/indignant comments routinely attract lots of upvotes, and then sit on top of the thread, smothering more interesting discussion. But in terms of moderation we can only reply to commenters.
Its even dumber when you consider AI model output was not considered copyrighted or eligible for copyright to begin with, as far as I know. How exactly would this even be "theft"?
Plus on top of that it's been ruled that training on copyrighted material is allowed because it's transformative. As a result even if the model output is copyrighted training on it would be as allowed as training on books.
If this was ruled ok why are they paying billions to book publishers?
Yea to be more explicit than the other reply they are being fined for the theft of the books not training on them. They trained on tons of legally procured books by destructively scanning and OCRing them. It was discovered in legal filings and is called Project Panama if you want to read more about it.
Because they pirated said books instead of buying them legally.
I can't imagine Moonshot have found a way to completely bypass the subscription or api billing either. So those model outputs have been paid for! No theft of any kind!
At most they'd be getting more out of distributed use of subsidized plans and API than the ToS would prefer.
It’s theft because a lot of rich people have money on the line.
It’s not they, but it does break TOS.
As if scraping the entire internet doesn't break thousands of TOSes?
Yes, indeed! I have a site where 99% of the traffic is Claude. I could ban them but I don’t bother.
I poison their requests with crap, particularly because they were the worst offenders for sheer request rate. The site has maybe a hundred pages, and I'd see them hit at a sustained 20 rps for weeks on end, same content being pulled over and over and over again. Easily handled by my small server as it's static content, but the sheer arrogance (inadvertent or otherwise) pissed me off.
I'm not sure if they caught on, or just finally started behaving better, but after months of being served crap in my never ending labyrinth of auto-generated junk, they have stopped.
My stuff is on Vercel, so it’s more of an issue as it uses up my limits. I’m moving to Cloudflare and I want to see how changes I make affect their bot, if at all.
Better to feed them poisoned results...
Scrapping internet realistically does involve Botnets of all kind, the scrapping itself might be legal, but it's certain a ton of infected computers are routing traffic and being used as Proxy.
It's still the pot calling the kettle black.
steal from the rich -> theft
steal from the poor -> capitalism
pay the rich for their services then lawfully use what you paid for to compete with the rich -> national security issue?
Training on copyright was ruled to be transformative.
Anthropic was just fined for not paying for the pirated books they copied into a training database. Same as anyone else who copied pirated IP onto their hard drive.
However, training an AI on copyright has been ruled to be sufficiently transformative and not a violation of IP laws. The same way you can make a gameplay clone of Call of Duty without any issue.
So they got fined for… stealing?
Also, if I clone Call of Duty, and use their skins, make a similar soundtrack, call my maps the same, then there will definitely be an issue.
I support the use of AI and all, but to hide behind transformative use of copyrighted intellectual property is a discredit to the colossal amount of human work and knowledge that these companies pirated and had their models trained on.
>So they got fined for… stealing?
Correct, 2001 napster style pirating of content. You cannot copy stuff you didn't pay for onto your hard drive.
>Also, if I clone Call of Duty, and use their skins, make a similar soundtrack, call my maps the same, then there will definitely be an issue.
A gameplay clone as I stated, tons exist (team death match, capture the flag, battle royale, with first person gunplay and army guys shooting each other). The rest is your argument, not mine.
I don’t think I’d equate a corporation downloading TBs of copyrighted works (with the intent of making a profit on stealing copyrighted works) and John downloading a few GBs of MP3s for personal consumption.
Probably why John wasn't ordered to pay $1.5B.
yep, theft but not plagiarism.
Theft would give the victim a one-time payment, plagiarism would have given the victims payments every time the model was used. In other words: IP stealing through training LLMs is declared 100% legal.
The question is what happens when somebody mass-"uncopyrights" entire libraries of content for the first time. Because, of course, this judgement should have sent the stock price of Disney, Paramount, Warner Bros, New York Times ... to zero.
Why? Because at this point these companies are living on borrowed time. Courts have outlined a clear, legal, process to "uncopyright" any given work.
If I make Bambi 2 and toss it up on YouTube I’m going to be sued by Disney for a wide variety of intellectual property violations, it doesn’t matter if I make it by hand or via an LLM. Nothing has changed for big organizations with a legal team.
Sued ... of course. But what will they claim? The courts have declared that training an LLM on copyrighted data and then having it produce anything DOES NOT make a derivative for the purposes of copyright. That it is fully owned by the person who used the LLM and zero licenses are needed to sell it, rent it out, or whatever.
> But what will they claim?
Copyright infringement obviously.
> DOES NOT make a derivative for the purposes of copyright
Or maybe you have misunderstood the ruling and thats why the stocks of these companies hasn't plunged.
Seeing as your prediction completely failed, it seems way more likely that actually you just misunderstood the ruling.
By that same argument, distilling a model to train your own is also transformative. We can’t have it both ways. Either both training on copyright and distillation are transformative, or both are illegal. Also, it is not a matter of settled law that training on copyright is transformative. That was a single case, and others are still pending AFAIK.
> Same as anyone else who copied pirated IP onto their hard drive
Not really the same. Private citizens have received fines MUCH higher per-work when downloading for just their private consumption.
Ruled by whom? The International Court of the Northern District of California?
What if use an LLM and attach CoD screenshots, or assets?
>"Anthropic was just fined for not paying for the pirated books they copied into a training database."
Normal person would be sitting in jail for doing same thing. It is a shame they call it justice.
Copyright infringement, maybe. Theft? No. When something is stolen from you, you don't have that thing anymore.
> The U.S. does NOT have copyright protections on the data that AI models were and are trained on, it was mass theft.
Absolutely false. Courts have decided that it is fair use [1]. That makes sense. It should not be used to justify Chinese theft.
https://www.reuters.com/world/us-judge-approves-anthropics-1...
Why is it not fair use for the Chinese models if all they are doing is distilling these fair use models? According to the courts, AI model output is not copyrighted or eligible for copyright.
Distillation does not violate copyright law.
It violates contractual terms of use.
Correct. This seems to be poorly understood.
You can use an original work to write an encyclopedia (e.g. Wikipedia). That is fair use.
You cannot copy an original work and distribute it.
LLMs are an encyclopedia.
But no Chinese company has copied and distributed an American LLM.
Even if they had wanted that, it would have been impossible, because Anthropic et al. do not let anyone access them directly.
Querying the available API can be used to extract only an extremely small part of the information stored in an LLM.
That part can be used for the post-training of another LLM, to obtain some desirable properties, but the extracted information is far too little to be called "copying" or "distributing". Moreover, after the post-training it does not appear anywhere in the weights verbatim, so its use is at least as transformative as the training of the original American model.
Besides these facts, there is no evidence that the Chinese companies have actually done this, even if it sounds plausible.
Terms of Use violation cannot be justified as "transformative" or "fair use".
We’re talking past each other.
No one has claimed Chinese models violated copyright law.
Some have claimed they violated ToS, which is different.
This is a silly argument. Like yes, obviously it’s of less U.S. concern if one American is robbing another versus a foreign interest. We can deal with one whenever while the other creates constraints.
The better arguments are the ones made in the article. Open weights increase competition and thus AI availability in the U.S. economy.
This is not robbing. Undesired distillation is a ToS violation, at best.
There is not even jurisdiction over Chinese companies, because they don't sell their LLM subscribtions, unlike US companies, and those comitted much more severe violations in getting their training material than a ToS violation or two (Anthropic already found guilty).
edit: I don't see this as legal argument against the ban, more like a justification for why basically no moral person is gonna side with OpenAI and Anthropic on this. US gov can try and ban as many weights as they want, I expect that to be similarly effective as banning numbers was in the past (i.e. not).
> There is not even jurisdiction over Chinese companies, because they don't sell their LLM subscribtions, unlike US companies, and those comitted much more severe violations in getting their training material than a ToS violation or two (Anthropic already found guilty).
This is false. They do sell LLM subscriptions. Many also release _most_ of their models as open weights.
None of it is robbing because there is no violence, but sure. The point is being confused why an external threat is treated differently from internal ones is a self goal in a debate where there are far better arguments for not banning these models.
It's not like those American companies were only stealing form Americans. Maybe you could argue about utility, but morally I don't see any ground to stand on
And the utility argument is also very shaky. The proposed punishment for foreign interests stealing "US data" is that other Americans now aren't allowed to benefit from that, while the rest of the world can
There isn’t an argument for banning Chinese or open-weight models. I’m just saying OP’s moralistic outrage take is about the weakest one you can come up with against it.
Obviously stealing is bad, doesn't matter who you steal from. In this case, its a simple case of no honour amongst thieves.
> its a simple case of no honour amongst thieves
Which is why making an honour-based argument to said thieves is silly.
Whether or not it’s robbery is irrelevant. It’s the proposed remedy that is the problem
Sure. Great. OP’s argument focuses on the outrage. That’s focusing on the first bit.
Blatantly lack of principles.
Sure. The American electorate isn’t really principled right now. If the best argument we can surface is this tripe, we’ll lose access.
You're trying to kidnap what I've rightfully stolen!
https://www.dailymotion.com/video/x4cxcsa
Are public libraries thieves?
Is Anna's Library thieves?
Is Library Genesis thieves?
Is Archive.org thieves?
Is PirateBay thieves?
They all look like public libraries to me. And better access to all human knowledge is a net positive for everyone.
Someone going to public library, copying every book out of it and then selling them as their own are thieves.
Not technically theft unless the public library no longer has the books. Selling something and saying you created it when you didn't is a different five-letter word: fraud.
Putting prompts in an LLM and saying you "created" the image or text thereof is fraud.
I don't believe anything is being sold here :)
Barrels of tokens sold by thieves that copied every book sure looks like selling to me.
AI companies sell access to their models, via subscriptions and ads.
Public libraries don‘t sell anything. Also authors and publishers get payed for having their books in the library (how and how much varies by country). Authors and publishers get nothing for having their books put into AI training data. Hence the former is a public service, and the latter is theft.
No, they've merely infringed copyright. Theft is a well understood concept and it does not pertain to information.
Okay, but that's not what happened as far as I'm aware. Who did they sell their copies of the books to?
Public library buy the works they distribute. They also pay for the usage of these works. Copying of said works is still illegal. You should read up on how libraries actually work.
> Copying of said works is still illegal.
Yes it is, but at least they are not charging for the access.
That's NOT always been the case.
Libraries used to require membership and dues. Those then bought more books on the used or retail market. LOTS of fights were about that system, cause book publishers hated libraries. And well, they still do.
And to be fair, fuck copyright. Its holds all of us back, so someone can go "FUCK YOU, NO". And nobody or company deserves what is it, 70 years+ death copyright length.
And with the recent Anthropic settlement, used to be, for-profit pirates would go to prison. I also remember absurd 2000's settlements over Britney Spears and Metallica for $5000-$7000 for 20 songs.
You are not going to win me over on massive gatekeeping human knowledge over "Itssss ill-eagle!". I know, and I don't care. And if I'm ever in a case over this, I'll tank it.
That’s fine too, in that case Open Weight models made by China are doing humanity a favour
I strongly believe that's the case. That's even the case with the recent OpenAI hack of HuggingFace.
The censorship of US models is so bad (read: guardrails), that they had to use GLM 5.2 on their own hardware to do the analysis. That's how locked down, restrictive, and closed source American models are. They're basically for-profit piracy by way of selling tokens.
Whereas the Chinese models are just "here you go, download and run".
I know what I run. Qwen3.5 and 3.6 and GLM5.2 . USA token vendors are incentivised in doing worse to sell more slot machine tokens.
The answer to widespread theft is not to simply allow the guy who did the theft to get away with it because his implementation was cheaper.
It's to punish theft.
But distillation is not theft. They are paying for the usage of the API. It is also not copyright infringement, as AI companies do not hold the copyright for the output of their models.
> But distillation is not theft.
Legally no, but I think an honest reading of the 'fair use' statute doesn't actually pass the 'doesn't compete with the author's original work' aspect of fair use, even if the use is transformative.
Fair use? Anthropic does not own the copyright to the output of their LLMs. Fair use does not enter into it at all.
not when by your own rules it's legally ruled "not theft".
Rules are flexible. The people who pay attention know this.
People expect the US to apply rules evenly to both American AI efforts and that of its primary geopolitical rival. China certainly doesn't do that. Their entire economy is based around giving Chinese firms the advantage regardless of what it means for the wallets of consumers at home or abroad.
Since that's who we're playing against, and no one is going to willingly give up their open-weight models from the totalitarian rival, well, then use their ruleset.
It's that or punish the theft by making the US companies pay for their training data and blocking outside efforts that trained off the data the US companies stole.
Exactly! While the people who spent more money on their theft are rewarded with more business to counteract their losses! Justice is finally being wrought on those who have wronged society with their wicked ways!
(/s if it wasn't obvious)
Being serious, it is far more likely, especially given the events of the past 18 momths, that Trump is just doing autocrat things because he can, and ultimately doesn't really care what is and is not good for American business as long as he continues to grow his wealth through corrupt behavior.
> Trump is just doing autocrat things
Yeah, he really is. It's awful. We're creating a class of elites that are accountable to no one, and their hold on power is slowly becoming sanctioned by law. It wouldn't surprise me if we soon had a system where there were people in various professions who considered themselves professional Trumpists who got to steer the fields they were in according to the diktats of the party. You would have to be in-line with their ideology to get anywhere in a career, whether it be in industry, politics, arts, or science. It'd suck.
You know where you see a fully-developed version of this idea?
China.
The era of so-called "intellectual property" is over.
You can accept it or not. But it's going to be a lot easier if you do.
So how come Disney still gets upset when I provide copies of their movies on my website as "training set for biological neural networks"?
The problem isn't that IP is dying - in most industries it died ages ago. The problem is that it is now explicitly reserved for the rich.
> Disney still gets upset
Why do you care about Disney's emotional outbursts in the first place though?
> The problem is that it is now explicitly reserved for the rich.
That's not new - it has always served entrenched interests at the cost of actual artists and entertainers. John Perry Barlow laid this out with great clarity throughout his time with The Grateful Dead, perhaps best here: https://www.youtube.com/watch?v=rLbqgG6o1n8
The solution is straightforward -- be rich!
One’s like taking something from your brother or sister and the other is like taking from your neighbor with whom you have a bad history…
I do care.
It's an unfortunate truth that we're exiting a global peace era and entering one of (hopefully just...) implied conflict. Our biggest geopolitical rival that's been threatening a hot war also distilling the things we build is not a good thing.
Trump is a huge self-own by our electorate sadly, but being concerned about a conflict between China and the US is a very bi-partisan concern for good reason.
Luckily for the peace side, the USA's inability to achieve their mission objectives in a war against an adversary that was perceived as being substantially weaker means they are very much less likely to engage in one with a "near peer".
The take away from modern conflicts is that massive military size disparities don't translate into successful invasions anymore.
In other words, it's unlikely USA will pursue a conflict with China, and it's unlikely China will pursue a conflict with Taiwan.
Yes, the administration is definitely and would not start a war for no reason other than to start a war.
People really have to stop sanewashing the government, it doesn't work anymore, these people are not thinking through what they're doing.
I don’t think those are “learned” lessons. Only how to push harder next time.
Might wanna reflect on who’s threatening hot war lately.
It's not a threat anymore if you actually start a war!
/s
The main actors threatening (and delivering) a hot war are China, and by proxy, Russia. The Ukraine war is absolutely and transparently enabled by China, who is also trying to actively destabilize the EU and the US in order to make its conquest of Taiwan more likely.
Yes, the US has an idiot in charge who is threatening these things in response, too. However, unlike China, the electorate does not support him. His election is also almost certainly due to aggressive campaigns to destabilize the US internally for this exact purpose.
The US isn’t just _threatening_, it is in an open war right now, invaded Venezuela, started trade war with US allies
>invaded Venezuela
Didn't happen. An intervention happened. An invasion lasts more than two hours.
Sorry who is China threatening a hot war with? As far as I can tell the US is doing most of the warring these days…
Taiwan, while simultaneously preparing an invasion fleet trained to stop US and EU attempts at a rescue.
>However, unlike China, the electorate does not support him. His election is also almost certainly due to aggressive campaigns to destabilize the US internally for this exact purpose.
What are you claiming here?
That China hacked the voting systems enough to swing the election? I know of no good evidence of this.
That China changed voters' minds enough to swing the election? But then "the electorate does not support him" does not follow. If I change your voting choice using some kind of non-coercive persuasion such as by making an argument, your support for your new choice is still genuine support. It doesn't matter if my argument was good quality or not, what my country of origin is, or even whether I lied about my motives in making the argument.
> The main actors threatening (and delivering) a hot war are China
China is actually the adult in the room.
Since 1979, 45 years ago, China has not dropped a single bomb on foreign land. That’s two generations peace.
In that timeframe, USA has attacked 37 countries, France 22, Britain 10, and Italy has attacked 6.
What determines if a country is a threat? Simple test - would they arrest Eipstein and execute him?
More politically correct phrasing - “do they respect American economic elites.” ?
If the answer is no, the country is ‘a threat to the west’
You don’t believe me? Name 2 countries that don’t fit this model.
Count wars, invasions and simply messing with other countries by the US starting since the end of WWII and then come back. Your post is a pile of BS
Our biggest geopolitical rival is also the only thing standing in the way of a potential duopoly on LLMs using closed weight models. I would not say I’m a fan of the PRC by any means, but I also do not want to see closed weight models controlled by two big companies dominating what I can use. Especially since said companies arbitrarily restrict model usage and can ban users for no reason.
> It's an unfortunate truth that we're exiting a global peace era
You’re a tiny bit late on that one
No, we are in the process of an exit. This takes time
Does this sentence have any meaning at all? There are 30 Active military conflicts right now.
I keep hearing about this "we" as if it's still a coherent nation state and not just an unaccountable ruling class from both parties where the most severe consequence is failing upwards into a lobbyist provided position. I literally have no moral preference when it comes to the US or the CCP nor do I care if the entire nation state fails, let alone loses to a geopolitical rival.
> nor do I care if the entire nation state fails
You should care because of the nukes, if nothing else.
Are we even sure they still work? The last test was 34 years ago, after all.
For all we know some crafty contractors or NCOs have been making very good money off of selling "spare" parts - just like we saw when Russia first tried to invade Ukraine...
Nukes are the only thing keeping corrupt ruling class and their ambitions in check. On either side of ideological divide.
It’s the only weapon that’s going to hit them along with the foot soldiers
> both parties
I'm sorry, but only one "party" is comprised of fascist plutocrats speed-running the end of democracy in the USA.
The USA and world would be a vastly different place had a different party maintained power in the USA 10 years ago.
As I see it, one party is comprised of fascist plutocrats, and the other party is comprised of their enablers. In dictatorships the latter is called controlled opposition, I don‘t know what the latter is called in a democracy, but in absence of a better word “both parties” gets the meaning across IMO.
I don't think I would characterize this last era as one of global peace.
I’m not convinced that it’s a bad thing. What is your actual argument?
My argument is that a major world power is pushing towards global conflict by simultaneously funding the conquest of Ukraine while building its own invasion fleet for Taiwan. Allowing them to also destabilize us internally is not exactly a good idea.
> funding the conquest of Ukraine
It is well known that China is still engaged in trade with Russia- oil and gold going one way, drones and weapon components going the other.
But you said funding, and I assume you meant things like grants or loans. Do you have any sources for that? I could only find one that was attributed to indirect open source intelligence, nothing specific.
Trade is funding.
Russia has been trading with a number of countries[0]. Mostly China, but also significantly with India, Turkey and the EU, and in much lesser amounts with 30+ other countries, including the US, the UK, UAE, Brazil, Japan, South Korea, etc.
I don't know if the above data helps to make your original point or not. Would you say that all those countries are funding the Russian war against Ukraine? Perhaps they do so for a combination of various reasons - energy and food needs, business profits, other self-interests, don't care, etc.
Wouldn't a country pushing for global conflict - really global - sow disorder by doing things like: break up alliances, threaten annexations, back illiberal poitical movements, disrupt shipping, disrupt free trade, disrupt energy supplies, start and foment regional wars, etc? Given that all countries tend to do things like this to one extent or another, how many of these things has China done, and how does that compare to other similarly capable countries?
I would say that China is asserting itself as a regional power, like with a navy that threatens Taiwan but can't project power much further than that, wouldn't turn into a global conflict unless other global powers feel so threatened they make it one. Maybe just a play on fear, for power and profit. War-mongering. Wouldn't be the first time.
EDIT: And look at how China has been "funded" China (via trade) over the last three decades... ironic!
[0] https://www.bruegel.org/dataset/russian-foreign-trade-tracke... (summary imports/exports by country in stacked graphs at bottom of page)
> Allowing them to also destabilize us internally is not exactly a good idea.
The US destablized itself internally. China did not force the US to export its manufacturing base. It did not force the US to invade Iraq. It did not force the US to create a housing market bubble. It did not force the US to create massive income and wealth inequality. It did not force the US to supress transitioning away from oil towards cheap renewables. It did not force the US start a global oil shortage by invading Iran.
That having found itself at the tail of 25 years of poor financial planning, the US decided to go double-or-nothing by betting on LLMs to bail out the economy... does not mean that China has to fold its hand.
I thought you were talking about the US for a minute...
The US is building a Taiwan invasion fleet?
No, the US is actively invading Iran, by ship and by air. And they've invaded Venezuela. And they're actively supporting Israel's invasion of Lebanon and Gaza. And they're bombing a few more countries now and then, especially in Africa.
China is not actively involved in any war at all. They may be supporting Russia with resources, but it's nowhere near the level of material and diplomatic support that the USA is providing to Israel's genocide and war of aggression on Lebanon.
How is it destabilising? Do you purely mean China have exposed ill thought out investments? That’s capitalism my guy. The market is working as intended.
When the investments fail there will be volatility but it’s hardly existential for the US. Open models are good for the whole world, I would argue it’s a net positive for individuals.
I honestly wonder what world folks like you live in. The US is currently at war, repeatedly threatening several other places and engaged in several costly trade wars. From my Canadian perspective, you are a danger and a menace. China is a commodity supplier. Do the war crimes your country is currently committing not register for you at all? The daily threats, insults, and economic attacks on your 'allies', does that register?
China supports Russia in its war formally, but practically it sells Ukraine whatever it needs, because a weak vassal-like Russia is very much in their interests. The US is far, far more outspoken and active supporter of Russia in this war.
Finally, there will be no hot war with China. They'll likely blockade and invade Taiwan before the end of the decade, but the US will not do anything, partly because it'll be exhausted from it's never ending middle east wars, partly because it'll only take a small bribe to get Trump to back off. He's genuinely fond of dictators like Xi.
Or… war with China is inevitable if only to keep the gap in military capabilities. There is a thinking to mow down their military while we still can in some circles. Like those in power currently. Scary.
Ask yourself, who seems to be the destabilizing force if you take a step back and look at the state of things objectively?
There is some blame on all sides for sure, not saying it's black and white...but come on.
Almost all current world events as of now are directly tied to the fact that the west is in active freakout mode over china building a army large enough to attack the entire world. Iran and venuzuela are a direct response to it.
1. China is actively genociding an entire ethnic people.
2. They are actively building a military larger and faster than any country on earth in all of history.
3. They are actively threatening ALL of their neighbors with war. Actively attacking their ships and building military bases in other countries territory.
4. China's open stated goals are to conquer and control its neighbors and every single step over the last 10 years is proof of that.
You left out the bodypart business they are now in too.
I think the major destabilizing force is the country that has been building the infrastructure for a hot war over Taiwan for the last decade while simultaneously blatantly funding the conquest of Ukraine.
The US stupidly withdrawing from the world stage is definitely making this worse, of course, but it is not the catalyst. It's unfortunately a very likely side effect of a social media campaign by the same actor to make said hot war easier to win.
Hey, we started another war and are funding the most destabilizing force in the world but look at this theoretical war …
It's literally on their border and the history is akin to if The South established its own nation state in Cuba after the Civil War and China became its ally.
At least admit the double standard. I don't mind realpolitik but it's so tedious when it's not plainly stated and there's this veneer of civility.
Yes, the US is allowed to do whatever it wants to in South America. No, other nation states don't have a sphere of geographic influence if the US doesn't want them to.
Taiwan has been self governing for longer than almost everyone that lives there has been alive. It's also a vibrant democracy.
If you care about democracy and human rights, it's morally reprehensible NOT to support Taiwan. We can debate the ethics of intervening in strongman dictatorships like Venezuela, Cuba, or Iran, but Taiwan is clear.
Cool, and it's morally reprehensible to NOT support drastic action to curb the worst effects of climate change, and start trying our very best to prepare our children and grandchildren for an adulthood where they will be clawing back minor things we take for granted even in today's world of regular continent spanning forest fire smoke pollution.
"Democracy and human rights", "strongman dictatorships". Dear God this rhetoric has aged like milk. I can't believe we ever fell for it.
Despite withdrawing from our role on the world stage, we have in no way adapted isolationist like policies. We kidnapped the leader of another country, we are bombing Iran with impunity and will begin striking civilian infrastructure as a response to terrorist within that country, https://www.usatoday.com/story/news/politics/2026/07/22/trum...
China didn't make us do this any more than they made us bomb civilian fishing boats in South America. We are a major destabilizing force at the moment, China existing does not somehow negate the actions we take of our own free will.
Refusing to acknowledge our role in destabilizing the globe at the moment is a level of infantilization that I, and many others, find unacceptable. We are all adults, China is responsible for their role and the US is responsible for its role. A social media campaign is not forcing the folks in Washington to drop bombs on Iranian civilian infrastructure, the jack asses in Washington did all of this on their own
Nobody is refusing to acknowledge our role in destabilizing things. This is a thread about why the US might be afraid of an also bad-actor rival funding a huge proxy war against it controlling major technology.
I mean if you want to get back on the topic of this thread, it has nothing to do with what you just said. The thread started by discussing AI distillation and this was your initial comment, just to get us back on track
> Our biggest geopolitical rival that's been threatening a hot war also distilling the things we build is not a good thing
They didn't distill Fable, it is impossible on any timeline. So where do we go next, the start of this thread is a complete dead-end as it is asserting a falsehood
This was the comment i responded to so we can get back on track to that comment if you'd like
> I think the major destabilizing force is the country that has been building the infrastructure for a hot war over Taiwan for the last decade while simultaneously blatantly funding the conquest of Ukraine.
We are refusing to aid Ukraine and actively antagonize their leader. We are kidnapping leaders from other countries. We are bombing civilian infrastructure. We are the major destabilizing force, not the one planning a hypothetical invasion
In the decade during which you claim China has been preparing for a hot war, the United States invaded multiple countries and waged hot wars on three continents.
It’s ridiculous to compare China to the US when it comes to warmongering. They’re not even in the same ballpark.
Yeah, the US is not an incredible actor, but China is funding the most brutal war in modern history (the invasion of Ukraine) in order to provide a smokescreen for their own invasion.
The Ukraine war is large and terrible, but its nowhere near comparable in brutality to the genocide in Gaza, that the US is funding, providing weapons for, and protecting diplomatically.
> who seems to be the destabilizing force if you take a step back and look at the state of things objectively?
All of them. Pretending there is a benevolent bully among Russia, China and America is baseless.
This is probably the strongest moral argument for open-weight models. It decouples AI, somewhat, from military-industrial might.
This is proof the US propaganda is working. How much damage has China done recently militarily?
> This is proof the US propaganda is working
This is a thought-terminating cliche. (The folks telling you to stop being sheeple are unusually consistently claimed independent thinkers.)
> How much damage has China done recently militarily?
Exempting the Iran war, which has no justification, as much as America. We’re both backing proxies. (Though we’ve had a while since we shot at or annexed our neighbours.)
It seems pretty silly to answer a question about military intervention by insisting that the major instance of military intervention by a world power that's happened in the last year shouldn't be counted
> by insisting that the major instance of military intervention by a world power that's happened in the last year shouldn't be counted
I’m not. I’m saying exempting that, we have an even field of conflict mongering. Counting Iran, the field is biased, but we have significant sampling-timing issues for something like “is there war.”
China has violently annexed sovereign countries in the past decades. Its threats to do the same again are at least as serious as Trump’s Greenland nonsense. There can be more than one baddie.
Iraq war? Venezuela raid? Libya? Syria? Afghanistan for 20 years? And that's just some of the higher visibility cases.
If China had special forces deployed and conducting ops in random African countries it would be a massive aggression. When we do it, it doesn't even make the headlines.
> If China had special forces deployed and conducting ops in random African countries it would be a massive aggression
It’s literally doing this across its maritime and mountain boundaries! It does cause outrage, albeit less so in mass media.
What's a good example you would say?
I think they're referring to occasional fist fights between the Indian and Chinese deployments at the top of the Himalayas where there's a dumb territorial dispute. There were a couple of times decades ago where the forces shot at each other and the governments had to scramble to calm things down and deescalate, so somehow they've reached this gentleman's agreement to have fistfights instead.
https://en.wikipedia.org/wiki/Sino-Indian_border_dispute
We're talking about this kind of thing balancing out, say, Fallujah or shock and awe over Baghdad.
> We're talking about this kind of thing balancing out, say, Fallujah or shock and awe over Baghdad
Nonsense. For that comparison you’d look to the invasion and annexation of Tibet. (And it’s not ancient history if it’s currently endorsed and un-apologized for.)
Ok, then we expand the US horizon back to the 50s as well. We've added Tibet, and the 1979 3 week Vietnam war to the Chinese column, plus them being involved in the Korean war and to a lesser degree the American Vietnam War. This comes with way, way more added to the American column. At least it's no longer a shutout.
Any fair look at both countries' records puts the US as doing 10-100x more military intervention, on a global rather than regional scale. It is actually a little bit "sheeple" to look at that scoreboard and start calling China aggressive. They're less aggressive than every other UNSC member without even going to a pound for pound basis.
> Any fair look at both countries' records puts the US as doing 10-100x more military intervention
Sure. I never claimed otherwise. I’m saying it would be news to anyone in Asia-Pacific (or Eastern Europe) that America is or has been the destabilizing force. That was the original claim.
If the choice is 10x or 1x, go with the latter. But that is a false dichotomy. You can be anti-invasion and annexation broadly. There is no moral high ground in choosing Chinese models; no more than a person in China choosing Anthropic or OpenAI.
The Philippines definitely have a right to be aggrieved over the south china sea stuff, but one bias I'm cautious of is "the natural norm is American hegemony in every region, so any change from this is destabilizing". Other countries have their natural interests as well.
I’d say the South China Sea island building in violation of UNCLOS text and judgement comes pretty close to American CIA-style fuckery in South America and Africa.
To be clear, American fuckery is broader. But if we agree we’re discussing magnitudes and not kind, we agree on all I’m arguing.
>"China has violently annexed sovereign countries in the past decades."
An outright lie. The only possible candidate is Tibet but it's former sovereignty is very questionable and not clear cut
> it's former sovereignty is very questionable and not clear cut
Same as Taliban-run Afghanistan, same as Iraq, same as Ukraine, same as Taiwan. If you’re Tibetan, the line is academic—you faced, by modern definitions, genocide all the same.
Calling the territory you’re invading and/or annexing illegitimately independent, illegitimately governed is how these always go.
The US was literally founded on genocide. Glass houses.
Wow, what happened to no holds barred American capitalism? Is it bad when the Chinese somehow do it better? The US stateowning US Steal, Intel, and possibly openai is no different than the Chinese ownership model… and before the US took stakes in the companies they were just bankrolling them as customers SpaceX Oracle etal.
If life is a meritocracy, you gotta give it to the Chinese for their immense growth and improvement of living from agricultural poverty to technical supremacy. I don't want to hear the they stole our IP argument either, because all math is borrowed from something else. Distillation is innovation, in booze and in models.
"what happened to no holds barred American capitalism?" you mean the strawman that never existed?
Unlike the xenophobia
> Our biggest geopolitical rival
People on here have zero consistency in their worldviews.
Ask any capitalism zealot (ie hners) whether markets are zero sum and they'll wax poetical about "creating value". But then talk to them about Chinese competition and suddenly everything is a rivalrous good lolol.
I hate to break it to you, your so called rival won this game decades ago when you undercut the working class to make a buck and laundered cheap Chinese labor as productivity. So if you believe the Chinese are rivals um welcome to "reap what you sew" <shrug>.
I'm not the "hive mind" of HN or whatever. Individuals hold unique opinions.
Yes, China is a rival. No, they have not won. Their economy is in a lot of trouble, likely far worse than the shape of the US's currently. That situation may accelerate a hot war though, as has historically been the case globally.
People have been claiming this since the 2000s. There are Economist articles from 2005 predicting an imminent collapse of China's economy. China has only grown more rapidly since then.
You need to stop watching those “China is gonna collapse any minute now…” youtube videos
I don’t. I do read major reputed economists though, and they are not exactly bullish on China.
Don’t create fictional versions of the people you argue with. This is the second time in a span of 10 minutes that you’ve done it. It completely undermines any attempt to discuss something, and drops your integrity to zero.
Are any of those major reputed economists Russian or Chinese by any chance?
Or are they just Western economists repeating Western bias under a veil of objectivity?
The term you're looking for is 网络水军.
Yeah I'm seeing how USA vs China is going.
USA: https://www.linkedin.com/videos/tpopomaronis_ive-seen-some-e...
China: https://www.youtube.com/watch?v=DUbbBdSGHE8
But... China is definitely copying... Hahahahha
I’m sorry but this is a perfect example of what the person above criticized but in reverse. Come on now.
China’s youth unemployment rate is 15%, and that’s likely heavily astroturfed to look lower than it is. Even 15% though is terrifying for a mature economy.
You know who should be terrified by that rate? Us.
China's one-child policy is finally starting to hit, and its population is actively shrinking. At the moment China is not in a crisis yet, as its GDP is still growing quite well - so clearly there's plenty of economic activity.
So if a lot of old people are now retiring, and the young can't find jobs, who is doing the work? China is still a factory powerhouse, so who's screwing together the gadgets?
The answer is that they have been pushing heavily for automation, and it is working. Our biggest chance of beating China on a free market was domestic automation versus cheap overseas labor - but they have gotten there first. And that should be absolutely terrifying you.
It doesn’t sound like their society or their economy is collapsing.
Like ours would if we had 15% youth unemployment.
Maybe we’re projecting our values?
The USA has sold its tech base off to enshittifiers. They defunded and destroyed most critical tech research.
https://www.dongascience.com/en/news/75540
China leads in 66 of 74 critical tech areas. The USA has a whopping... 8.
But hey, the USA's MBA "leadership" figures how much to worsen everything and extract more profit. I guess that counts for "innovation", in the Private Equity books.
I wish I could upvote you thrice lol
China is only getting stronger. They have captured basically the entire electronics industry - except a handful of cutting-edge chips. They have captured the solar panel industry. They are significantly outperforming the traditional brands when it comes to all kinds of BEVs.
In all the ways that matter, China is rapidly racing towards or even past the West. We've been pretending that it is possible to remain relevant on the world stage while only offering services and a handful of luxury goods. A China without the West would be a painful transition, but the West without China? A massacre.
China hasn't won yet primarily because the West hasn't even figured out it is in a competition. We're getting our asses kicked and don't even realize it yet! Worst of all? The Western capitalists created China, all for a few cents of shareholder value...
> Their economy is in a lot of trouble, likely far worse than the shape of the US's currently
Source: trust me bro
> Almost 200 Silicon Valley companies, including Proton and Y Combinator…On Wednesday, the newly-formed Little Tech Association
YC is little tech? lol wut.
YC's entire business is finding little tech companies and helping them grow.
Calling YC little tech is hilarious. Billions under management and many billion+ companies.
> Little Tech Association
"Little Tech" in contrast to "Big Tech" monopolies and hyperscalers
Awful website: ugly particle systems with text you can't scroll to (on SE): https://littletech.org/mission
Sorry but isn’t YCs entire goal to turn “little” tech into “big” tech?
A little bit of Googling goes a long way...
A little bit of explaining does to. What’s your point?
People shouldn't have to explain that "Little Tech Association" is a proper noun and therefore its own entity. Learn how words work, bro.
Unnecessarily rude, I’m sure you agree. What separates this view from the following?:
“I started a Black Cats Only Association whose main membership is white cats, and a critique of the mismatch reveals failure to recognize it’s simply a proper noun.”
Yes I understand it’s a proper noun.
It worked for the bolsheviks
> the US can “sanction” companies that engage in intellectual property theft.
Peak hypocrisy, US AI companies can train on unlimited intellectual property with 0 rights to it, while Chinese AI companies have to explicitly get the rights to data that isn’t even copyrightable/copyrighted (since AI alone can’t copyright it).
It's not IP theft if US courts say it isn't.
And if the accused company is outside of the US, well, US courts have no jurisdiction so apparently the US government can just claim they are guilty and impose the sanctions...
The USA has stolen intellectual property from ALL AROUND THE WORLD. AI is not trained just in copyrighted USA content.
Other countries have a say about it.
Because they used a pirating website to get the material and stored them on their servers for personal use.
The AI training part of the court case was deemed fair use.
And it’s not a war if Putin says so. And it’s not a possible future invasion of Taiwan if Xi says so. Who cares what US courts say.
Meanwhile, X.ai committing all of the crimes.
did you mean Xi.ai?
US's court can enforce IPs. Didn't Antrhpic just agree to pay book authors billion for the recent lawsuit settlement?
Anthropic was asked to pay book authors because it trained on pirated downloaded books.
What about books and art where the author/artist does not authorise AI to train on it? They do happily train on it, ignoring their "ToS".
This is just double standards, a slap on the wrist to not worsen the situation with authors imo.
You could also say the Chinese companies are doing the same - they _do_ pay for their Anthropic subscriptions after all.
>Anthropic was asked to pay book authors because it trained on pirated downloaded books.
No, it was asked to pay book authors because it pirated copies of books and stored them on their hard drives. The ruling had nothing to do with training.
Thanks - I didn't verify this personally, but still stand corrected.
If training wasn't considered outside the law, this goes on to make the point about double standards for US vs Chinese model training methods.
Pirating millions of dollars worth of books would land people in jail. Government went after Aaron Scwartz for "pirating" scientific journals.
Aaron Schwartz was re-publishing the journals, which is the core concern or copyright and indeed where the word "copyright" comes from. Anthropic could get in a lot more trouble if their models are caught reproducing copyrighted work from their training set wholesale.
I think the Aaron Schwartz case is incredibly vexing because he was obviously acting out of a sense of altruism without personal self-interest. I don't think he deserved the book getting thrown at him like that. But the whole copyright system, which people seem to think is simultaneously good and bad, kinda rests on not allowing those kinds of violations
I've linked this several times, but LLMs are capable of reproducing entire books. Researchers were able to extract books nearly verbatim: https://arxiv.org/abs/2601.02671
I didn't think about this parallel, but that's just so sad.
The judge explicitly ruled that training on the books was fine.
They just should have bought them, rather than pirating them.
Also LLM output is not IP (in itself) in the first place, nor would Anthropic want to claim it is and that they have rights to it - that would drive paying customers away.
The issue comes down to at most ToS violations.
> They just should have bought them, rather than pirating them.
Bought, scanned and destroyed them I believe. The judge okay'd Destructive Scanning.
Destruction is not necessary. Google Books is a solid precedent. You can keep the content, you just can't make significant parts publicly available
> The court also held that the third factor favored fair use as to the purchased library copies converted from print to digital because the purpose of the copying was to keep the books in its library but with more favorable storage and searchability properties. This purpose required copying, there was no surplus copying and the source copy was destroyed. With respect to the pirated copies, however, the court held that because “Anthropic lacked any entitlement to hold those copies” and retained them “even after deciding it would not make further copies from them for training,” this third factor weighed against Anthropic for that particular use.
https://www.loeb.com/en/insights/publications/2025/07/bartz-...
You need to destroy the _physical copy_ that you scanned.
The settlement doesn't cover the actual training on copy written material
That will never happen. Copyright is not a thing in China.
Copyright is a thing in China, but Chinese enforcement is very different.
too bad hypocrisy never got in the way of people making money
“And they warn that banning Americans from downloading Chinese open-weight models wouldn’t stop their proliferation — but would weaken U.S. startups.”
Well this doesn’t even make sense.
Weaken U.S startups = Can’t legally develop commercial products that use or remix the latest open weights models in novel/creative/cost-reducing/etc ways, while companies from the rest of the world could (including China, who this is ostensibly supposed to be punishing)
Wouldn't stop proliferation = Despite that, your non-programmer 14 year old child can still install the latest Qwen model on Day 1 by copying a one liner command they got from a YouTube short that sources from a Gitlab repo outside US jurisdiction
It does because corporations, especially smaller ones competing with bigger ones, actually do get prosecuted for violating restrictions like this a lot more than individuals do.
I think we should shut it off. It would force US companies to build open models.
I think it would do the opposite. They would have a monopoly just handed to them, and a large market with no options. They'd have zero incentives to provide open weight models.
Right now, I think their incentive is the highest: Chinese AI companies are dropping competitive models left and right, and clearly American companies are feeling the pressure, given they're pulling every legal leg they can to suppress their competition. If we just remove the competition, Anthropic et al win. It's a temporary and short-sighted victory, but this is America. Every company is short-sighted, it's baked into our financial infrastructure.
US labs would have even less incentive to build open if we blocked foreign open weight models, and we'd be engaging in protectionism yet again, also giving them zero incentive to compete.
It baffles me how naive people are about China and their practices despite nonstop evidence, and then justify their arguments one way or another.
But many commenters are CCP propagandists, some even trying to convince people the US companies are the enemy! That's straight up propaganda.