How did the AI Safety talent pipeline come to work so well?View in threadAnton Zheltoukhov1y*033that's just false. alignment community even to this day is significantly less then 1000 people worldwide. i would not call that a success. i'm not blaming the people involved. they've done what they could. lets just not deceive ourselves hereReply
Alignment is not *that* hardView in threadAnton Zheltoukhov1y*021current safety techniques offer almost no value for the real alignment problems that we are starting to facehttps://www.lesswrong.com/posts/HmQGHGCnvmpCNDBjc/current-ais-provide-nearly-no-data-relevant-to-agi-alignment#ponxMRXsnDJNyY4YXReply
that's just false. alignment community even to this day is significantly less then 1000 people worldwide. i would not call that a success. i'm not blaming the people involved. they've done what they could. lets just not deceive ourselves here
current safety techniques offer almost no value for the real alignment problems that we are starting to face
https://www.lesswrong.com/posts/HmQGHGCnvmpCNDBjc/current-ais-provide-nearly-no-data-relevant-to-agi-alignment#ponxMRXsnDJNyY4YX