Rohin Shah comments on Reinforcement Learning in the Iterated Amplification Framework