Two models of AI oversight – and how things could go deeply wronggarymarcus.substack.com1 point·son_of_gloin··0 commentsOpen articleSaveView on HN