This project explores deep reinforcement learning, hybrid actor-critic approach with A3C/PPO combined with curiosity for playing Super Mario Bros