註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Jonathan Whitaker
@johnowhitaker
Data scientist, AI researcher, aspiring gentleman scientist
加入 October 2015
1.2K 正在關注    10.3K 粉絲
TIL, Anthropic's production alignment procedure uses ~30M training examples :) I'm recording a video (maybe live) with @alexisgallagher today on the Open Character Training paper, if you're interested in this kind of thing :)
顯示更多
New Fellows Research: Can Claude autonomously align other AIs? We gave Claude 48 hours and 1 GPU to improve the alignment of small models. It researched and proposed methods, then trained and tested the models on its own. It worked surprisingly well.
顯示更多