Joar Skalse comments on Goodhart’s Law in Reinforcement Learning