I’m actually quite surprised. Didn’t think I’d be saying that but looks like good job math wise (edit: obvious caveats apply: not my area of expertise, don’t know if novel, don’t know if there’s any related proofs). (As far as x-risks go though, something not well designed for self improvement is a lot less scary).
It’s not obvious to me that this work is relevant for AI safety research, but it seems mathematically interesting on it’s own (though it’s not my area of experise either, so I can’t evaluate it).
I’m positively impressed. Honestly, I’d never thought that SI/MIRI would deliver anything scientifically noteworhy. I hope this is the beginning of a positive trend.
I’m actually quite surprised. Didn’t think I’d be saying that but looks like good job math wise (edit: obvious caveats apply: not my area of expertise, don’t know if novel, don’t know if there’s any related proofs). (As far as x-risks go though, something not well designed for self improvement is a lot less scary).
It’s not obvious to me that this work is relevant for AI safety research, but it seems mathematically interesting on it’s own (though it’s not my area of experise either, so I can’t evaluate it).
I’m positively impressed. Honestly, I’d never thought that SI/MIRI would deliver anything scientifically noteworhy. I hope this is the beginning of a positive trend.