Microsoft has introduced Muse – the first-of-its-kind artificial intelligence model capable of generating gameplay

Microsoft has made significant progress in the field of generative artificial intelligence for video games by introducing its latest model, Muse, capable of generating gameplay. Muse is designed to generate ideas and support developers in expanding their projects. This model, known as the "World and Human Action Model" (WHAM), was developed in collaboration with the game studio Ninja Theory, renowned for its multiplayer battle arena game *Bleeding Edge*.

Microsoft's Vice President in charge of gaming AI has revealed details about the new generative AI model. She noted that Muse is capable of understanding game physics and how the game interacts with player actions through a controller in a 3D environment. Additionally, it can help adapt older games for modern devices, making them more accessible and playable.

Microsoft is confident that Muse can bring significant advantages. The company believes the AI model will contribute to:
- Recreating and enhancing classic games that are no longer supported on modern hardware;
- Developing new gameplay mechanics during the prototyping phase;
- Generating fresh game content;
- More efficient concept testing;
- Accelerating the game development process.
Muse was trained on an extensive dataset of gameplay from the Xbox game *Bleeding Edge*. The project was carried out by Microsoft's research team in collaboration with Ninja Theory. The model had access to an amount of gameplay data equivalent to seven years of real-world playtime, allowing it to be trained on a billion image/action pairs.
Microsoft emphasized that this model is not designed to create full games or replace developers. Instead, they hope the technology will be beneficial for both players and game developers.

During the press conference, Microsoft demonstrated how Muse functions in real time. The artificial intelligence created game graphics and even interacted with objects that developers added to the game. Currently, this is an early version of the model, working with low resolution and at 10 frames per second. Microsoft released several short GIF animations demonstrating Muse's capabilities. Currently, the AI is limited to generating at a resolution of 300x180 pixels, but future improvements in quality are planned.