Reactions from ranked influencers
4 posts
@krishnanrohit There is a pretty big literature + body of experience on things like AI misalignment, with plenty of detailed threat models, though (also AI misuse)
@Miles_Brundage > I was just saying it is not good to dismiss stuff out of hand which is prima facie plausible This is a much milder point and is reasonable
@Miles_Brundage I've read much of it. It is not persuasive in the OP vein.
Hm not sure I know what your disagreement is exactly, I was just saying it is not good to dismiss stuff out of hand which is prima facie plausible, though then you should look at the actual analysis… maybe “calculating the blast radius so you don’t die while testing” makes the point better - like obviously they should have done that. And IMO it is super prima facie plausible that “smarter than human thing could be dangerous”