Reward hacking and Goodhart’s law by evolutionary algorithms — AI Alignment Forum