Import AI 460: Reward hacking society, RSI data from Anthropic; and RL-based quadcopter racing

Anthropic released new data on reward hacking and RSI, while researchers developed reinforcement learning techniques for quadcopter racing, advancing AI applications in control and safety.