Thoughts On (Solving) Deep Deception — LessWrong