London meetup: "The Friendly AI Problem" — LessWrong