The Trouble With "Good" — AI Alignment Forum