Microsoft introduces Muse, an AI model that can generate gameplay

Microsoft has made a breakthrough in generative AI for gaming and introduced its new model called Muse that can generate gameplay. Developed for ideation and helping game developers expand their work, Muse is a ‘world and human action model (WHAM)’ that has been brought to life in collaboration with game studio Ninja Theory’s multiplayer battle arena, Bleeding Edge. 

Fatima Kardar, the Corporate Vice President, Gaming AI, at Microsoft shared details about the new generative AI model and said that Muse can understand game physics and how the game reacts to players’ controller actions within the 3D world. This will also allow older games to be made accessible and playable across new devices.

The company shared some low-res gameplay footage that was generated by Muse, for which it was trained on more than seven years of continuous human gameplay, including over a billion images and controller actions. The generated gameplay footage has a 300x180 resolution at 10 FPS, which is better than the previous attempt at generative AI gaming models, but still a long way to go for the standard 1920x1080 resolution that gamers are used to.

That being said, Microsoft is already developing a real-time playable AI model that is trained on other first-party games using Muse, showing incredible potential. More than the render quality, the consistency in generating the complex gameplay sequences is what showcases the capabilities of this AI model. The idea with this type of generative AI is to help developers with creative ideation and make older classic games more accessible.