[link] Friendly AI and Utilitarianism — LessWrong