註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Owain Evans
@OwainEvans_UK
Director of Truthful AI (non-profit AI safety research group) + Affiliate at UC Berkeley. Work: Emergent misalignment, subliminal learning. Prefer email to DM.
加入 April 2020
496 正在關注    22.6K 粉絲
New paper: Adding clean data to your dataset is assumed to weaken poison data. We show the opposite can occur. Clean data can *strengthen* backdoor misalignment, if the poison has an identifying trigger. We call this homeopathic learning.
顯示更多
0
14
515
59
轉發到社區