BroadMind Learns to Play Super Mario

We have trained a neural network to learn to play Super Mario through practice and reinforcement by extending DeepMind’s algorithm. ~vmnih/docs/ What game should BroadMind tackle next?