Friendly AI at the NYC Future Salon — LessWrong