Review for NeurIPS paper: Accelerating Reinforcement Learning through GPU Atari Emulation