Game AI has developed significantly since its inception, but it can often still feel ‘robotic’ to players. Machine learning offers a potential solution to this by offering naturalistic behavior based in brain-like neural networks. By using a centralized model with asynchronous machine learning, processing power is freed on the client computer where the game can be run. This approach is used in the project for an original game with a symmetrical multiplayer design. Overall, the AI opponents did not perform as would be expected from a human player, but movement had some properties of natural behavior expected in human players.