How likely is it that AGI will first be developed by a large established organization, rather than a small startup, an academic group or a government?
-
Even if in principle AGI doesn’t need so much compute, vast amounts can compensate for lack of insight.
- And we likely will
-
First one is likely to be the worst one.
-
Funding is key.
-
It used to be much less clear, but scaling laws strongly imply that if the first AGI requires enormous amounts of compute, smaller actors would not have the funds
-
Manhattan-Project style thing could be possible for a government, but they’d need to get lots of top scientists, which is hard as those people are likely happy where they are (Rob speculating)
-
Academics don’t have the compute/funding