Deep learning - deeper flaws? — AI Alignment Forum