13h agoOpenAI's roon and commentator Siméon debate AI alignment, disputing whether models need strict corrigibility or more relaxed behaviorsOverly literal chain-of-thought behaviors in current models triggered the discussionSentimentSentimentPos0%Neg100%Users reacted negatively to the case against corrigibility by dismissing the argument as overcooked or expressing outright hostility toward the concept.3 comments with sentiment. View comments.