Serving technology enthusiasts for more than 25 years. TechSpot is a trusted source for tech advice and analysis.
In short: Some game designers use generative AI to brainstorm ideas during development. Microsoft’s open-source model is designed to make interactive content possible. It focuses on maintaining consistent across many frames, while recognizing and integrating user input.
Microsoft Researchers recently introduced Muse. It is a generative AI that extrapolates interactive video game scenarios based on images, clips and recorded player input. The tool is designed to streamline game creation while maintaining ethical training practices. The weights and sample data for Muse (which Microsoft calls “World and Human Action Model,” ) are now available in Azure AI Foundry. The developers The company discussed the details of the technology in its research blog and a newly published article. Published in Nature.
After training on a million updates, Muse is able to accurately predict up half a minute’s worth of gameplay using one second of footage and nine seconds player input. Muse, unlike previous models, remembers details such as terrain, characters and game mechanics.
Dom Matthews, Ninja Theory’s CEO, explained in a video with Xbox head Phil Spencer that the company does not plan to use Muse for building content for users. The studio will use it instead to help draft and iterate on ideas by quickly generating snapshots that create a cohesive vision. Microsoft will not mandate that its studios use its models. It is unclear if other developers will follow a similar approach.
Immediately, some developers expressed disinterest. David Goldfarb who worked on Battlefield Mass Effect and Killzone gave a short, dismissive reply. Marc Burrage, from Creative Assembly, told Wired that restricting the use of generative AI in prototyping can be detrimental to game development because it robs employees of hands-on experiences.
Spencer speculated optimistically that technologies such as Muse could facilitate game preservation by allowing developers to port older titles to new devices. Microsoft has repeatedly promised to preserve their customers’ digital library across hardware generations. However, it remains to see if regenerating them using AI can produce satisfactory results. Frank Cifaldi, from the Video Game History Foundation, isn’t optimistic. He comparesthis idea to a photocopy. Microsoft was also criticized for Muse’s data. The harsh criticism of AI models is largely due to companies using material from the internet without the creators’ consent. Microsoft avoided the issue by using games it already owned to train Muse.
In the initial demonstration, player telemetry data was taken from Ninja Theory’s multiplayer shooter Bleeding Edge. Users who have agreed to the EULA of the game have already consented for Muse to be trained. The company could theoretically extend the practice to popular games like Minecraft, Call of Duty or Halo, Overwatch, StarCraft Doom or Forza.
