The History and Risks of Reinforcement Learning and Human Feedback