My colleague Jim has been playing Atari games with reinforcement learning from Clojure and wrote this blog post: https://whydoidreamofkites.blogspot.com/2018/04/using-arcade-learning-environment-from.html