Wei Dai comments on Towards more cooperative AI safety strategies