How Does Reinforcement Learning Work