Thoughts On (Solving) Deep Deception — AI Alignment Forum