Researchers at Stanford University and Google have unveiled an RPG-style experiment in which 25 ChatGPT generative agents were released into a virtual small town where they interact among themselves.