Microsoft’s Muse generative AI adapts game code for any platform

Microsoft has unveiled a new generative AI model called Muse, which the company says has a deep understanding of the 3D world and game physics, and can react to player actions.

Not everyone was enthusiastic about the announcement. AI researcher and game designer Dr. Michael Cook noted that Muse doesn’t create new games, but only predicts how gameplay will develop based on video recordings of gameplay sessions. According to him, the AI ​​helps developers test changes faster, such as adding new elements or levels, but is not a generator of original ideas.

Muse’s process is based on training on large amounts of video footage. Microsoft used the multiplayer game Bleeding Edge for this. The AI ​​analyzes gameplay and predicts possible changes when adding new mechanics. However, the developer remains the main link, controlling the process and making adjustments in real time.

Cook emphasizes that Muse technology is not revolutionary and does not solve key development problems. Its effective operation requires significant resources, as well as careful instruction of game content. In addition, such projects are expensive and require large amounts of data.

Microsoft’s statement that Muse can help in preserving games has become particularly controversial. The head of Xbox Phil Spencer expressed confidence that AI will be able to port old games to modern platforms, but Cook criticized this statement. In his opinion, Muse does not reproduce the real gaming experience, but only generates visual sequences that do not always accurately convey the original logic and mechanics of the game.

Thus, despite Microsoft’s ambitious statements, Muse remains more of an auxiliary tool for testing and optimizing games than a universal solution for their preservation and adaptation. The problem of archiving and transferring classic games requires more complex methods than simply generating video sequences based on old content.