Human Compatible

Last updated November 9, 2026

By Stuart J. Russell

Computer scientist Stuart Russell, an AI researcher at the University of California, Berkeley, argues that the standard way of building artificial intelligence, giving machines fixed objectives to pursue, will become dangerous as systems grow more capable. He proposes instead machines that are uncertain about what people want and learn human preferences by observing behavior. Russell explains the principles behind this approach and discusses misuse of AI, automation and the problem of keeping control.

Buy on Amazon

Human Compatible has been recommended by