Every day I grow more and more glad that I turned down a Meta offer. It was probably a hire-to-fire offer anyway, not based on any engineering prowess on my part. Still, I couldn't be more relieved I dodged that bullet.
They have nothing else to do. Someone needs to be able to justify their position by creating stupid changes like this to create a line item on their LinkedIn.
Meanwhile, nobody seems focused on capturing CEO’s data for AI training.
It is going to be funny in a few decades when zuck transfers his shares and voting rights and estate to the ai bot, and makes himself functionally immortal. Or at least a sort of commissioned renaissance painting version of himself, probably.
Imagine in 300 years we are still ruled by zuck, ellison, bezos, musk, thiel, et al, just in ai model form empowered by estates worth more than entire nations and legal protections designed to outlast heat death of the universe. Assuming there is still a "we" living on earth. Charitable assumption I guess.
Wasn't it a few months ago that some engineer leaked that XAI was building 'Human Emulators'. This is either Meta's attempt at the same or just a blatant lie to make sure their engineers aren't slacking off. I've heard the workload has more than doubled for those who weren't laid off which is the only reason I think it might not be a employee monitoring system as I don't think anyone there can afford to not work hard.
> to improve the company's models in areas where they still struggle, like choosing from dropdown menus and using keyboard shortcuts
Seems like a strange approach in general. I'd have assumed you'd just have it use accessibility features to get at things, if there is no other interface.
Knowing how to make an accessible website is so rare that companies pay me money to do it for them. I wish it was good enough for people, much less companies, to rely on.
Even if no attempt at proper accessibility was made, it's still generally far easier to attempt to find an HTML (or other form of UI) element, than to attempt to scroll to the right spot and use visual inspection to find things.
‘Meta spokesperson Andy Stone said the data collected would not be used for performance assessments or any other purpose’
Horseshit.
1. Employees are being asked to train AI to replace them.
2. Performance assessments will 100% be impacted. No question.
Thinking back on the OTT interview experience that Facebook helped pioneer, imagine making it through that, getting paid a massive sum of money BUT barely getting by on it because of the location, then they drop this crap on you?
The framing here matters. Recording keystrokes and mouse movement to train autonomous agents means they're essentially trying to replicate the work, not just assist with it.
There's a pretty direct implication: if you work at Meta and participate in this, you're generating training data for a system that's intended to eventually do your job. Consenting employees are literally labeling their own replacement.
The thing that will be interesting to watch is whether this stays US-only for legal reasons or if that's just the pilot. GDPR makes this kind of collection significantly more complicated in Europe.
US mouse movements are obviously very vigilant, some people say they're the strongest mouse movements ever seen.
Since this is a serious website: I'd be genuinely curious how mouse velocity and trajectories differ between cultural and environmental settings
(apart from hardware, that's boring and should be normalized).
There was a time when studies made headlines that were exactly about the relationship between mouse movement, typing etc, and psychiatric disorders as well as physical health.
Obviously, both are related.
If you ask me, Ad tech would probably be able to tell your denominated faith using this data, when there's enough of it...
Genuine question: would right to left language based interfaces have different type of movements and thus training data than left to right language ones?
When you will think about it, what actually useful data are you getting from this exercise? It is like strapping camera on a manual laborer so you can see what he sees, but you don't get data about the touch and grip and you won't get data about why he is doing specific moves.
Every day I grow more and more glad that I turned down a Meta offer. It was probably a hire-to-fire offer anyway, not based on any engineering prowess on my part. Still, I couldn't be more relieved I dodged that bullet.
They have nothing else to do. Someone needs to be able to justify their position by creating stupid changes like this to create a line item on their LinkedIn.
Meanwhile, nobody seems focused on capturing CEO’s data for AI training.
The same company is trying to build an AI Zuckerberg...
It is going to be funny in a few decades when zuck transfers his shares and voting rights and estate to the ai bot, and makes himself functionally immortal. Or at least a sort of commissioned renaissance painting version of himself, probably.
Imagine in 300 years we are still ruled by zuck, ellison, bezos, musk, thiel, et al, just in ai model form empowered by estates worth more than entire nations and legal protections designed to outlast heat death of the universe. Assuming there is still a "we" living on earth. Charitable assumption I guess.
Not funny ha-ha though.
Wasn't it a few months ago that some engineer leaked that XAI was building 'Human Emulators'. This is either Meta's attempt at the same or just a blatant lie to make sure their engineers aren't slacking off. I've heard the workload has more than doubled for those who weren't laid off which is the only reason I think it might not be a employee monitoring system as I don't think anyone there can afford to not work hard.
[dupe] https://news.ycombinator.com/item?id=47851242
[dupe] https://news.ycombinator.com/item?id=47851948
Look at the HN ID number to see which is first.
This is the source/first
> to improve the company's models in areas where they still struggle, like choosing from dropdown menus and using keyboard shortcuts
Seems like a strange approach in general. I'd have assumed you'd just have it use accessibility features to get at things, if there is no other interface.
Knowing how to make an accessible website is so rare that companies pay me money to do it for them. I wish it was good enough for people, much less companies, to rely on.
Even if no attempt at proper accessibility was made, it's still generally far easier to attempt to find an HTML (or other form of UI) element, than to attempt to scroll to the right spot and use visual inspection to find things.
Training on future vi macros. Just
Taking screenshots too.‘Meta spokesperson Andy Stone said the data collected would not be used for performance assessments or any other purpose’
Horseshit.
1. Employees are being asked to train AI to replace them.
2. Performance assessments will 100% be impacted. No question.
Thinking back on the OTT interview experience that Facebook helped pioneer, imagine making it through that, getting paid a massive sum of money BUT barely getting by on it because of the location, then they drop this crap on you?
Big Brother is always watching.
The framing here matters. Recording keystrokes and mouse movement to train autonomous agents means they're essentially trying to replicate the work, not just assist with it.
There's a pretty direct implication: if you work at Meta and participate in this, you're generating training data for a system that's intended to eventually do your job. Consenting employees are literally labeling their own replacement.
The thing that will be interesting to watch is whether this stays US-only for legal reasons or if that's just the pilot. GDPR makes this kind of collection significantly more complicated in Europe.
Not sure how GDPR could help. They can make generating data for their model your actual job as per contract I guess.
Is US mouse movement different from European? Is it called sparkling mouse movement if it comes from California?
US mouse movements are obviously very vigilant, some people say they're the strongest mouse movements ever seen.
Since this is a serious website: I'd be genuinely curious how mouse velocity and trajectories differ between cultural and environmental settings (apart from hardware, that's boring and should be normalized).
There was a time when studies made headlines that were exactly about the relationship between mouse movement, typing etc, and psychiatric disorders as well as physical health.
Obviously, both are related.
If you ask me, Ad tech would probably be able to tell your denominated faith using this data, when there's enough of it...
Genuine question: would right to left language based interfaces have different type of movements and thus training data than left to right language ones?
As everybody knows, key strokes and mouse movements are the things that solve problems, definitely the data worth capturing for AI training.
Maybe they're building a simulation of the rich lives and behaviors of white collar office people in the early 21th century, with breathtaking detail?
I couldn't imagine life without my unique keystrokes and mouse movements.
Like a museum exhibit?
But you can put on 3D goggles, maybe there's even TTS narration.
Some call it museumverse.
When you will think about it, what actually useful data are you getting from this exercise? It is like strapping camera on a manual laborer so you can see what he sees, but you don't get data about the touch and grip and you won't get data about why he is doing specific moves.
More accurately learn which employees are inactive while WFH
I dont actually think its for training AI models. AI is the scapegoat - just like the layoffs