Last night, Microsoft announced Muse, a new AI tool aimed at enhancing gameplay ideas.
They also shared some low-quality gifs showing AI-generated gameplay from Ninja Theory’s multiplayer game, Bleeding Edge. The tiny images likely aimed to downplay any awkwardness that often comes with AI-generated content.
Microsoft claimed that Muse could “radically change how we experience and preserve classic games” and suggest that this AI could make older games work on “any device.”
The reaction on social media was quick. Some pointed out that Microsoft was riding the AI trend while others feared that Xbox might use Muse to churn out low-quality content.
Fortunately, Dr. Michael Cook, an expert in AI for games, explained what Muse actually does. Cook has a decade of experience in AI and game design and has published extensively on the subject.
In a detailed blog post, Cook clarified that Muse doesn’t create original game ideas. Instead, it learned from seven years of video footage of Bleeding Edge to generate new gameplay scenes.
This process is similar to what Google did with the classic game Doom last year.
The main goal? To help developers visualize how gameplay might change with new features. Cook notes that Muse can show what a game might look like with adjustments, like adding a jump pad in a new spot.
Essentially, Muse serves as a tool to help developers see potential gameplay changes. Humans remain the key decision-makers.
According to Microsoft’s research, Muse must grasp concepts like consistency and diversity to function effectively. If a developer modifies a game, Muse should ensure that these changes persist and behave uniformly, no matter how players interact with them.
For instance, if a jump pad is added to a level, Muse has to predict how players would react to it and make sure it always works the same way.
Cook points out that the research is less about generating gameplay or ideas and more about how developers will interact with these new tools.
Interestingly, Microsoft’s blog post mentioned that they’re leaving the decision to use generative AI up to creative teams, without forcing it on them—at least for now.
Even though Muse’s current application is limited, Cook questions whether it’s a viable method for game development. He argues that while it’s impressive, the complexity of generating realistic footage remains a challenge.
At its core, using Muse is expensive and requires a lot of existing gameplay footage. Cook also wonders how developers can gather enough footage early in the game development process for Muse to learn effectively.
On the topic of game preservation, Xbox chief Phil Spencer expressed excitement about the potential for AI to make older games accessible on new platforms.
Cook disagrees strongly, calling Spencer’s thoughts “idiotic.” He believes that while any form of media creates some level of preservation, Muse does not provide a complete solution for keeping games alive.
Despite advancements in AI, Cook points out that there is still no reliable way to know what an AI actually captures. Muse’s ability to output grainy gifs of a single game isn’t enough to represent the full spectrum of gameplay experiences.
He emphasizes that a model like this is not the ultimate answer for game preservation, suggesting that while it might serve as a side note, true preservation requires more effective methods.
Recently, Take-Two Interactive’s Strauss Zelnick weighed in on AI, stating that “artificial intelligence is an oxymoron,” emphasizing caution in the technology’s use.
