Reading List

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 (Stanford University) from Techmeme RSS feed.