On: GPT-3

[pharke]

I think the point Gwern makes is that if a government wanted to they could easily allocate enough resources to do this. More bluntly I think he's saying if the US decided tomorrow to begin a Manhattan Project for AGI there is a non-zero chance that they might succeed in 7 years.

[motoboi → pharke]

If they started a Manhattan Project for AGI, would we know by now? Or only after it explodes?

[pharke → motoboi]

Only after, secrecy and surprise is the whole point.