Review for NeurIPS paper: Provably Efficient Exploration for Reinforcement Learning Using Unsupervised Learning