I keep wondering why the US claims it needs to keep ahead of China, lest all be lost. This is not a zero sum game, and the difference between first and second place seems likely to be a few months at most. A few month lead doesn't seem to me like that much of an advantage, even if it is maintained over years (which is not a given, the lead could go back and forth).
And finally, this is not a race with a single goal (like landing the first man on the Moon). One side could be ahead with regard to, say, math proofs, while the other side could be ahead with chemistry applications or something.
If AGI is the goal and something like the Singularity is even a remote possibility, the difference between first and second place could be effectively infinite. Given China’s suppression of human rights, it’s not a chance I would want my government to take.
Authoritarianism requires a boogeyman to point and shriek at. It is required to have a big bad scary evil (and RED) enemy to justify the constant erosion of democracy and civil rights.
That's pretty much it. The reasons don't really go much deeper. There are deeper threads you can pull at and more well reasoned and thought out objections to China, but those are academic. The people in power don't care/can't comprehend anything more sophisticated than "communism bad".
Fear mongering is a great way to obtain and maintain political and economic power. It can work with internal or external opponents but if you want to avoid or minimize internal strife (which can lead to instability and a loss of power) then it is best to use an external opponent. This is not a new phenomenon, it's been part of the political playbook for centuries.
China is spending about 15% of what we’ve spent on datacenter buildout. If we’re in a race with them it would seem like they aren’t even trying to beat us.
This feels like an attempted primer and justification for major bailouts to the big AI companies. And this is all a problem that the companies have created for themselves by going so hard into AI investments based on wild speculation about the near future of AI. And the bailouts will probably happen because A: these companies have politicians in their pocket, and B: most of the government / politicians will frame this as a national security problem we can't afford to ignore. This all serves to entrench and centralize power and access to data for the major AI companies that managed to stay in the game up to this point.
How would a bailout help? At the end of the day these companies need products people want to pay for. We know from the dot com bubble that you can burn money for years but eventually investors want to see some returns.
Keeps them running until a breakpoint they hope is coming where all the AI investments and infrastructure they have places them at the top of the world. I (and it seems most tech companies) believe high quality data corpus and access to compute will become strong forms of hard power in the near future. The question is if they can survive long enough for that to come fruition, and if it's before all their investments deteriorate and deprecate.
>set to warn of the risks posed by the artificial intelligence market
Nuanced, but the title doesn't match the article. All markets are risky, but not all markets are dangerous. All things considered, a bubble popper resulting in slower growth (as stated in the article) is a risk that is not dangerous.
No, they're referencing Secretary of Education Linda McMahon saying A1 (repeatedly) when she meant AI. Re-read the original comment, that's a 1 (one) not an I (the capital form of i).
I would be more surprised if they DIDN'T have anyone assessing the risk of an AI bubble. The report deals with characterizing how deeply embedded in the broader economy AI companies are, and the risks of heavy infrastructure investment, concentrated market power, reliance on private financing and potential shocks like supply chain disruptions, geopolitical tensions and electricity shortages. What is the issue (if any)?
No doubt this will be used to justify a government-funded capital lifeline to OAI and Anthropic who are still bleeding cash (irrespective of whether inference is profitable) and may have issues raising more money in the private markets if they are signaling a delayed IPO.
The proposed “donation” of a 5% stake to a sovereign wealth fund creates a direct incentive for government cash infusion.
I really can’t begin to describe how angry this possibility makes me. And I don’t think I’m alone. Keep pushing the envelope Sam / Dario and see what it gets you. Doubling down on a losing bet just digs your hole deeper.
What happens when the government sinks half a trillion dollars into this and we still don’t see an ROI / true agent autonomy? Then what? Ask for another trillion dollars and hope you can stumble on a research breakthrough equally as revolutionary as the transformer?
What's jarring is nobody is asking the hard question - how much was expended to make the breakthrough of Transformers?
It certaintly was nothing close to half a trillion.
Clearly more money is not the path to the solution. Furthermore China is doing pretty well with a fraction of the spend. America may have money but money needs to go toward productive projects - this requires ideas and vision. Which cannot be bought actually.
Well absent a research breakthrough the only option is “scaling” which according to Sam Altman was logarithmic growth in model ability with number of weights.
Of course any stats undergrad could tell you this was a fairytale. Increasing number of weights only works until you exhaust the signal in the data. I’m not sure how he was allowed to get away with such a blatant lie but here we are.
This lie is effective because on a small time scale it’s impossible to tell the difference between logarithmic growth and logistic growth. If you maintain a fixed training data size, increasing the size of the model will get you logistic growth in model capability meaning that past a certain size you get effectively no gain in performance because you’ve already squeezed out 99% of the signal.
“The intelligence of an AI model roughly equals the log of the resources used to train and run it.”
He is playing loose with the language here because the only way this statement holds is when resources = breadth and depth of training data - not compute / model size.
Ai guys would get torn apart by any sane copyright legal system. They have to go all in on the political favoritism play. They stole the content from the commons and continue to resell laundered IP via tokens
I keep wondering why the US claims it needs to keep ahead of China, lest all be lost. This is not a zero sum game, and the difference between first and second place seems likely to be a few months at most. A few month lead doesn't seem to me like that much of an advantage, even if it is maintained over years (which is not a given, the lead could go back and forth). And finally, this is not a race with a single goal (like landing the first man on the Moon). One side could be ahead with regard to, say, math proofs, while the other side could be ahead with chemistry applications or something.
If AGI is the goal and something like the Singularity is even a remote possibility, the difference between first and second place could be effectively infinite. Given China’s suppression of human rights, it’s not a chance I would want my government to take.
Ah, Pascal’s wager used to justify trillions in spending.
Not very creative.
Authoritarianism requires a boogeyman to point and shriek at. It is required to have a big bad scary evil (and RED) enemy to justify the constant erosion of democracy and civil rights.
That's pretty much it. The reasons don't really go much deeper. There are deeper threads you can pull at and more well reasoned and thought out objections to China, but those are academic. The people in power don't care/can't comprehend anything more sophisticated than "communism bad".
Fear mongering is a great way to obtain and maintain political and economic power. It can work with internal or external opponents but if you want to avoid or minimize internal strife (which can lead to instability and a loss of power) then it is best to use an external opponent. This is not a new phenomenon, it's been part of the political playbook for centuries.
China is spending about 15% of what we’ve spent on datacenter buildout. If we’re in a race with them it would seem like they aren’t even trying to beat us.
This feels like an attempted primer and justification for major bailouts to the big AI companies. And this is all a problem that the companies have created for themselves by going so hard into AI investments based on wild speculation about the near future of AI. And the bailouts will probably happen because A: these companies have politicians in their pocket, and B: most of the government / politicians will frame this as a national security problem we can't afford to ignore. This all serves to entrench and centralize power and access to data for the major AI companies that managed to stay in the game up to this point.
How would a bailout help? At the end of the day these companies need products people want to pay for. We know from the dot com bubble that you can burn money for years but eventually investors want to see some returns.
Keeps them running until a breakpoint they hope is coming where all the AI investments and infrastructure they have places them at the top of the world. I (and it seems most tech companies) believe high quality data corpus and access to compute will become strong forms of hard power in the near future. The question is if they can survive long enough for that to come fruition, and if it's before all their investments deteriorate and deprecate.
>set to warn of the risks posed by the artificial intelligence market
Nuanced, but the title doesn't match the article. All markets are risky, but not all markets are dangerous. All things considered, a bubble popper resulting in slower growth (as stated in the article) is a risk that is not dangerous.
The administration is very knowledgeable about all things A1.
Not true at all, I hate misinformation like this. They like their steak well done with ketchup on it. No A1 sauce in sight.
Your parent comment may be an example of Poe’s Law.
No, they're referencing Secretary of Education Linda McMahon saying A1 (repeatedly) when she meant AI. Re-read the original comment, that's a 1 (one) not an I (the capital form of i).
We are both making separate but equality high-quality stable genius level A1 (A One) jokes.
They want it in every classroom, every day. If they fund steaks for school lunches some good may come of it.
I would be more surprised if they DIDN'T have anyone assessing the risk of an AI bubble. The report deals with characterizing how deeply embedded in the broader economy AI companies are, and the risks of heavy infrastructure investment, concentrated market power, reliance on private financing and potential shocks like supply chain disruptions, geopolitical tensions and electricity shortages. What is the issue (if any)?
No doubt this will be used to justify a government-funded capital lifeline to OAI and Anthropic who are still bleeding cash (irrespective of whether inference is profitable) and may have issues raising more money in the private markets if they are signaling a delayed IPO.
The proposed “donation” of a 5% stake to a sovereign wealth fund creates a direct incentive for government cash infusion.
I really can’t begin to describe how angry this possibility makes me. And I don’t think I’m alone. Keep pushing the envelope Sam / Dario and see what it gets you. Doubling down on a losing bet just digs your hole deeper.
What happens when the government sinks half a trillion dollars into this and we still don’t see an ROI / true agent autonomy? Then what? Ask for another trillion dollars and hope you can stumble on a research breakthrough equally as revolutionary as the transformer?
What's jarring is nobody is asking the hard question - how much was expended to make the breakthrough of Transformers?
It certaintly was nothing close to half a trillion.
Clearly more money is not the path to the solution. Furthermore China is doing pretty well with a fraction of the spend. America may have money but money needs to go toward productive projects - this requires ideas and vision. Which cannot be bought actually.
Well absent a research breakthrough the only option is “scaling” which according to Sam Altman was logarithmic growth in model ability with number of weights.
Of course any stats undergrad could tell you this was a fairytale. Increasing number of weights only works until you exhaust the signal in the data. I’m not sure how he was allowed to get away with such a blatant lie but here we are.
This lie is effective because on a small time scale it’s impossible to tell the difference between logarithmic growth and logistic growth. If you maintain a fixed training data size, increasing the size of the model will get you logistic growth in model capability meaning that past a certain size you get effectively no gain in performance because you’ve already squeezed out 99% of the signal.
This disproves point number 1 in Sam’s thesis: https://blog.samaltman.com/three-observations
“The intelligence of an AI model roughly equals the log of the resources used to train and run it.”
He is playing loose with the language here because the only way this statement holds is when resources = breadth and depth of training data - not compute / model size.
(this is only slightly sarcastic)
Ai guys would get torn apart by any sane copyright legal system. They have to go all in on the political favoritism play. They stole the content from the commons and continue to resell laundered IP via tokens