Charlie Steiner comments on What are good models of collusion in AI?