[mannanj]

Isn't Facebook training their AIs on their finest engineer's computer use so the AIs can become better computer users?

In this case, the more insidious yet subtle risk and attack vector for humans using these Facebook computers, is that Facebook begins to use this data to discriminate (legally) on performance metrics. They can then use these to automatically disseminate performance improvement plans, lead to higher productivity (perceived, as whats measured no longer ends up being a useful metric) and control and urge people to do more of what they desire.

And my curiosity is: does what Facebook desire align with what the humans who work for Facebook desire? I think with AI, that's a no. The company desires as low a labor/workforce/compensation cost as possible, while the humans desire as much compensation as possible.

[pizzly]

What are the models actually learning from Meta's engineers computer usage data?

Are they learning to procrastinate by browsing hacker-news before starting to actually work so that in the future some models will replicate this behavior.

[mannanj]

Lol I hope so, that would be pretty funny.

I don't remember exactly the detail, or how much was disclosed. But I saw a leaked audio of the Zuckerberg meeting where he was sharing his insight into why they want to collect this data and I remember it being pretty vague.

If you want to use the computer the way the smart engineers use it, you probably look at everything.