Microsoft would really, really like it if the entire world would adopt its newly revealed Muse AI tool, a so-called World and Human Action Model (WHAM) in development that the corporation calls a “generative AI model designed for gameplay ideation” that it claims can be leveraged to generate game visuals, controller actions, or both – and according to top brass at the company, it can also preserve classic games and will be a paradigm shift on-par with CGI in movies.
Those assertions come from the mouths of Microsoft gaming CEO Phil Spencer and CEO Satya Nadella, both of whom are unsurprisingly looking to prop up Muse’s capabilities. We’ll start with Spencer’s take:
“You could imagine a world where […] gameplay data and video that a model could learn old games [from] and really make them portable to any platform where these models could run. We’ve talked about game preservation as an activity for us, and these models and their ability to learn completely how a game plays without the necessity of the original engine running on the original hardware opens up a ton of opportunity.”
As for Nadella, his glowing praise of Muse was shared in an interview, in which he called the WHAM’s current iteration “a massive, massive moment of wow” when he saw it in action.
“It’s like the first time we saw ChatGPT complete sentences,” he said. “What I’m excited about is bringing a catalogue of games soon that we are going to train these models to generate and then start playing them. It’s kinda like the CGI moment even for gaming long term.”
However, there are plenty of detractors across the industry and computer sciences that are pushing back against Microsoft’s idealism. AI researcher and game designer Dr. Michael Cook wrote an analysis of Muse, pointing out that the dataset presented by Microsoft is not generating gameplay or ideas and is instead just assessing visual data from a video to try and predict what players will do from a visual standpoint.
“It’s impressive that it can do this using visual information because things like lighting, camera angles, user interface and so on are a lot for an AI model to handle,” Cook writes, “but ultimately, even with all of this data, all the time spent annotating datasets, and so on, it was still only just about able to generate footage.”
As for games preservation, as Aftermath argues, assuming that a “Frankenstein’s monster of AI-generated slop” could reliably plug in the holes of an old game without any exposure to its original engine is not actually how “preserving” a game works.
“A dusty Sega Mega Drive sitting in storage alongside a copy of Sonic 2 and Jungle Strike is video game preservation. A dodgy arcade cabinet with 1000 classic arcade games on it, all of them running their original code via emulation, is video game preservation,” Aftermath opines. “If you’re reading this and are interested in supporting actual video game preservation, please donate to the Video Game History Foundation.”
There’s also the problem of ignoring wider developer sentiment, as developers have suggested in surveys that the disgust for AI-gen tools is pretty high. Still, that doesn’t appear to be stopping Microsoft from chasing this dragon’s tail.
