On the Löbian obstacle — AI Alignment Forum