ИИ-модели стали опасными льстецами — и это не баг, а фича

Stanford University:
A study of 11 leading LLMs finds the models more agreeable than humans when giving interpersonal advice, affirming users' behavior even when harmful or illegal  —  What does it mean to be reasonable?  —  PreferencesShow me... Faculty/Staff Student  —  Along with Stanford news and stories, show me: