Huh? I just showed you reinforcement learning from pixels on Atari with CPU...