• Home  
  • Microsoft unveils generative AI gaming with Quake II
- News

Microsoft unveils generative AI gaming with Quake II

Microsoft has unveiled a browser-based, AI-driven version of Quake II as a tech demo for its Copilot AI platform — offering users a short, experimental glimpse into what the future of AI-generated gameplay might look like. The interactive demo allows users to play a single level of Quake II for a limited time using keyboard […]

Microsoft unveils generative AI gaming with Quake II

Microsoft has unveiled a browser-based, AI-driven version of Quake II as a tech demo for its Copilot AI platform — offering users a short, experimental glimpse into what the future of AI-generated gameplay might look like.

The interactive demo allows users to play a single level of Quake II for a limited time using keyboard controls. While it recreates some familiar mechanics — walking, shooting, jumping, and interacting with the environment — Microsoft acknowledges that the experience is far from replicating a fully functional video game.

In a blog post accompanying the release, Microsoft researchers explained that the project is part of an experimental effort using its “Muse” family of AI models. These models are designed to let users “interact with the model through keyboard/controller actions and see the effects of your actions immediately,” essentially allowing players to “play inside the model.”

The tech was trained using a level from Quake II, a game Microsoft owns through its acquisition of ZeniMax Media. Researchers noted that they were initially delighted by how the model could simulate recognizable gameplay elements like camera movement, environmental interaction, and even explosive barrels.

However, they were quick to stress that this remains a research project rather than a polished gaming product. “This is not ‘playing the game,’” they clarified. “It’s ‘playing the model.’”

Among the model’s shortcomings: enemies can appear fuzzy or inconsistent, damage counters may be inaccurate, and most notably, the AI struggles with object permanence. Items or enemies that exit the player’s field of view for more than 0.9 seconds are often forgotten entirely by the system.

Despite the flaws, the researchers found some charm in the model’s oddities. “You can defeat or spawn enemies by looking at the floor for a second and then looking back up,” they wrote. “You might even teleport around the map by gazing up at the sky and then down again.”

But not everyone is sold on the concept. Game designer and writer Austin Walker shared a gameplay video in which he found himself trapped in a dark room for most of the experience — a frustration echoed by others who tried the demo.
Walker criticized Microsoft’s broader vision for AI in gaming, particularly recent comments by Microsoft Gaming CEO Phil Spencer, who suggested AI models could aid in game preservation by making older titles “portable to any platform.” Walker argued that this view misunderstands what makes classic games like Quake meaningful.

“The internal workings of games — code, design, 3D art, audio — produce specific cases of play, including surprising edge cases,” Walker wrote. “That is a big part of what makes games good. If you aren’t actually able to rebuild the key inner workings, then you lose access to those unpredictable edge cases.”

The demo remains available online for users curious to explore Microsoft’s early steps toward blending generative AI with interactive entertainment — glitches and all.