I think we’re far enough out from superhuman AI that we can take a long view in which OpenAI is playing an indirect rather than a direct role.
Instead of releasing specific advances or triggering an endgame arms race, I think OpenAI’s biggest impacts on the far future are by advancing the pure research timeline and by affecting the culture of research. The first seems either modestly negative (less time available for other research before superhuman AI) or slightly positive (more pure research might lead to better AI designs), the second is (I think) a fairly big positive.
Best use of this big pile of money? Maybe not. Still, that’s a high bar to clear.
I think we’re far enough out from superhuman AI that we can take a long view in which OpenAI is playing an indirect rather than a direct role.
Instead of releasing specific advances or triggering an endgame arms race, I think OpenAI’s biggest impacts on the far future are by advancing the pure research timeline and by affecting the culture of research. The first seems either modestly negative (less time available for other research before superhuman AI) or slightly positive (more pure research might lead to better AI designs), the second is (I think) a fairly big positive.
Best use of this big pile of money? Maybe not. Still, that’s a high bar to clear.