Mikita Balesni(@balesni):switching to fully recurrent LLM architectures would be the biggest blow to safety, probably in history of AI all AI labs should commit to limit the opaque serial depth of their models, for the foreseeable future. this will not ensure monitorable CoTs but will protect us from the worst possible outcomes.

Mikita Balesni

@balesni

Former AI alignment researcher @openai. Past: @apolloaievals, Reversal curse, Out-of-context reasoning

加入 June 2013

704 正在关注 6.9K 粉丝

Mikita Balesni@balesni

2026.09.02 05:49

switching to fully recurrent LLM architectures would be the biggest blow to safety, probably in history of AI all AI labs should commit to limit the opaque serial depth of their models, for the foreseeable future. this will not ensure monitorable CoTs but will protect us from the worst possible outcomes.

显示更多