Giving AIs safe motivations — AI Alignment Forum