Machine recommendation system ethics, models of human reward, reinforcement learning framing; followup on GPT-3.