Microsoft Launches Browser-Based Quake II Demo, Powered by Copilot AI’s Muse Model

Microsoft has released a browser-based, playable level for the classic computer game Quake II. This serves as a technical demonstration of Microsoft’s Copilot AI platform’s gaming capabilities; however, as the company acknowledges, the experience is not as immersive as playing a well-crafted game.

The tech demo is part of Microsoft’s Copilot for Gaming initiative and features an AI-generated Quake II clone playable in a browser. The Quake II level is fairly rudimentary, with blurry enemies and interactions, and Microsoft has restricted the duration of time users can spend playing this tech demo.

In a blog post outlining their work, Microsoft researchers explained that their Muse family of AI models for video games allows users to “interact with the model through keyboard or controller actions and observe the effects of their actions instantly, essentially enabling them to play within the model.”

Muse AI Model

While Microsoft initially demonstrated its Muse AI model at 10 frames per second (fps) and a resolution of 300 x 180, the latest demo runs at a playable frame rate and a slightly higher resolution of 640 x 360. Nevertheless, it remains a highly limited experience that merely hints at future possibilities.

Microsoft continues to promote Muse as an AI model that can aid game developers in prototyping their games. When Muse was unveiled in February, Microsoft indicated it was exploring how this AI approach could enhance classic games like Quake II and adapt them for modern technology.

“Much to our initial delight we were able to play inside the world that the model was simulating,” they wrote. “We could wander around, move the camera, jump, crouch, shoot, and even blow-up barrels similar to the original game.” At the same time, the researchers emphasized that this is meant to be “a research exploration” and should be thought of as “playing the model as opposed to playing the game.”

The researchers acknowledged “limitations and shortcomings,” such as the fact that enemies appear fuzzy, the damage and health counters can be inaccurate, and, most notably, the model struggles with object permanence, often forgetting about objects that are out of view for 0.9 seconds or longer.