The META bonuses went out and triggered multiple Karpathy-style exoduses. When knowledgeable insiders leave companies on the brink of success, it tells us that the people who have seen the next iteration believe that ideas will no longer be tied to infrastructure.
The META bonuses went out and triggered multiple Karpathy-style exoduses. When knowledgeable insiders leave companies on the brink of success, it tells us that the people who have seen the next iteration believe that ideas will no longer be tied to infrastructure. https://t.co/F1awIGkogM
@AndrewCurran_ they could probably afford a couple H100s for themselves with those bonuses though
@AndrewCurran_ Or they think they can participate in a vc cash grab
@AndrewCurran_ On recent Lex pod, Sama predicted intelligence becomes commoditized/ubiquitous, like electricity. This line of thinking seems to support that
@AndrewCurran_ that’s wild, who else could be thinking that 😅
@AndrewCurran_ I don’t think that’s necessarily true. They probably just wanna work on things they want to, and you can’t really do that if you are in a high pressure industry leading company.
@AndrewCurran_ Or just join another company.
@AndrewCurran_ This actually makes a lot of sense when you think of previous technological shifts including PCs, Internet, Mobile, and Cloud, and where opportunities to build has accrued long term
@AndrewCurran_ Or they can make more money elsewhere, or they are just bored with the job.
@AndrewCurran_ Extrapolating *a lot* here but I wonder if this implies future LLMs will be able to reduce the draw on compute and energy usage. It'd imply Altman might be wrong to think that achieving AGI can be best done by scaling up compute more and more.
@AndrewCurran_ I don’t know this person, but it sounds like the “build on top of LLMs” is the reason. Meta is building open source LLMs + internally using it as features in their existing products. Not really doing anything exciting in the application layer itself