Microsoft has developed a new AI model named Muse, which utilizes a training set from the little-known game Bleeding Edge. This model, detailed in a recent publication in the journal Nature, is designed to generate game visuals and controller inputs.
Originally released by Ninja Theory, Bleeding Edge was a 4 vs. 4 brawler that ceased updates less than a year after its launch. Microsoft incorporated a clause in the game’s End User License Agreement (EULA) allowing it to record online gameplay, which facilitated the data collection necessary for Muse’s training.
Applications of Muse
Muse aims to streamline the game development process. For instance, game designers can use it to prototype new ideas without the need for extensive programming and artistic resources. This capability could significantly reduce the time and costs associated with game iteration.
The model showcases a feature called persistency, which allows it to adapt to user modifications in real-time. As new gameplay elements are introduced, Muse can adjust accordingly, creating a realistic simulation of gameplay scenarios.
Current Limitations
Despite its innovative features, Muse has limitations. The generated visuals are currently at a resolution of 300 x 180 pixels and operate at about 10 frames per second. Microsoft is releasing Muse’s weights and sample data for researchers, allowing them to explore its capabilities.
For more information, visit the original article on Engadget.