Microsoft’s Muse: The AI Breakthrough That’s Not as Revolutionary as You Think
Last week, Microsoft made a big splash in the tech world with the announcement of Muse, a new “generative AI breakthrough” designed to aid “gameplay ideation”. The company showcased some grainy-looking gifs of AI-generated gameplay footage, based on Xbox studio Ninja Theory’s multiplayer game Bleeding Edge. But is Muse really the revolutionary technology Microsoft claims it is?
To get to the bottom of things, I turned to Dr. Michael Cook, an AI researcher and game designer who has spent years studying the intersection of AI and games. Cook has written extensively on the subject and has even built his own AI to see if it could win a game jam a decade ago. His insights are invaluable in understanding what Muse is really capable of.
According to Cook, Muse is not generating gameplay or creating its own original ideas. Instead, it’s been fed seven years of video footage of people playing Bleeding Edge to see if it can then generate further gameplay footage of it. This process is similar to what Google did last year to generate footage of classic first-person shooter Doom.
So, what’s the point of all this? Cook explains that the AI model is designed to predict what might come next if changes to a game were made. For example, if a human adds a jump pad to a game level, Muse can predict what the gameplay footage would look like. This is not about generating new gameplay, but rather about providing a shortcut tool for predicting and visualizing how gameplay might adapt to a particular input by a developer.
But here’s the thing: Muse is not a practical process. It’s impressive that it can do what it does using visual information, but it’s still only just about able to generate footage predicting player behavior. This system lacks any practical use, is enormously expensive, and requires any developer already have a vast vault of gameplay footage lying around for Muse to ingest.
Cook raises some important questions about the viability of Muse. For example, how do you get video footage of people playing your game in the first place? If you’ve been in development for a couple of months, you won’t have enough footage, and even if you make the systems able to run on less input data, there must be a minimum level required to understand the full game logic.
Lastly, on game preservation. Cook notes that you could imagine a world where, from gameplay data and video, a model could learn old games and really make them compatible with “any device”. But this is still a long way off, and Muse is not the solution to this problem.
Actionable Insights
- Muse is not a revolutionary technology that can generate new gameplay or create original ideas.
- It’s a tool designed to predict what might come next if changes to a game were made, providing a shortcut for developers.
- Muse is not a practical process and lacks any practical use, is enormously expensive, and requires a vast vault of gameplay footage.
- The viability of Muse is uncertain, and it raises important questions about how to get video footage of people playing games and how to make it compatible with “any device”.
Summary
Microsoft’s Muse is a new AI breakthrough designed to aid “gameplay ideation”, but it’s not as revolutionary as you think. It’s a tool that can predict what might come next if changes to a game were made, but it’s not practical, is enormously expensive, and requires a vast vault of gameplay footage. While it may have some potential uses, it’s still a long way off from being a game-changer.