AGI and Friendly AI in the dominant AI textbook — LessWrong