Microsoft Is Talking About AI-Generated Games, But Don’t Believe Hype

Microsoft has introduced Muse, a new AI model aimed at creating video games. Before imagining a world where AI produces big-budget games overnight, it’s important to remember that Muse is a step toward impressive technology, not the technology itself.

The Muse AI Is Impressive, But Not In A Usable Sense

Microsoft’s Muse AI has shown some impressive capabilities in generating video game content, but it faces serious technical issues that limit its usefulness for professional game developers. The biggest problem right now is the low resolution of its output, which is just 300×180 pixels.

This is much lower than what is expected in modern gaming, leading to visuals that are blurry and lack the quality needed for a successful game. As a result, any game made with Muse would look outdated and unappealing to players.

Additionally, Muse is trained only on gameplay from Bleeding Edge, a game released in 2020 for Xbox. This narrow focus restricts its ability to learn from the wide variety of mechanics, physics, and art styles found in other games. While it can successfully create gameplay that aligns with Bleeding Edge, it can’t adapt to different genres or artistic styles. Developers wanting to make a game with AI outside of what Bleeding Edge offers would find Muse inadequate.

This means Muse isn’t practical for serious game development or for the AI to be used in making a “catalog of games,” where good visuals and diverse design options are crucial. Right now, it functions more as a demonstration than a useful tool for industry professionals. There is great potential in Muse, but it needs significant improvements before it can become a valuable asset in game development.

Muse Works Like Minecraft AI, Which Isn’t Good For Xbox Games

Muse creates games based on what is around and what it has been trained on. This can cause some issues in the long run. After watching a lot of the footage, it’s clear that Muse keeps guessing what should be there and creates it on the spot. That is similar to Minecraft AI, which certainly doesn’t have the lasting power like games you could play forever.

Minecraft AI is fun to watch because it creates whole new scenes as the player looks around. What was once a mine can become a field, a lava-filled hellscape, or even an underwater castle. It is literally just guessing and filling in based on that guess. The difference is that players go into Minecraft AI expecting the chaos that comes with that, while developers need to go into Muse expecting it to work well every time.

That’s why right now, Muse feels like a gimmick, just like Minecraft AI. It’s a toy to play with, but it seems not to have been designed to actually make games. At best, this could be used to test concepts or ideas, but it is years from being anywhere close to being used in regular game development.

The Backlash Would Cause More Damage Than Profit

While the appeal of AI-created games for developers is strong, the backlash from players would likely be stronger. There are serious risks that could lead to outcry from gamers, critics, and even the developers themselves. If a game is largely created by AI, the potential downsides might outweigh any quick profits. A negative reaction could seriously damage Microsoft’s brand image, which would be a far bigger loss than any money made from a niche AI game. Let’s not forget that the company is already on the backfoot having effectively lost the console war.

A big concern is the lack of originality and artistic quality. Gamers appreciate creativity, unique gameplay, and engaging stories. An AI-generated game, based on existing data, might miss the originality and emotional depth that human designers and writers bring. Critics and players may see such a game as unoriginal and lacking the passion that makes a good game great. This could result in bad reviews, low sales, and a negative perception of Microsoft as a company focused on fast output over artistic value.

There are also worries about job loss in the game development industry. The idea that AI could replace human artists, programmers, and designers raises important ethical questions. Developers might feel threatened by the potential for AI to take their jobs, leading to mistrust and possibly harming future collaborations with Microsoft. This loss of trust could severely affect Microsoft’s ability to attract talent and produce high-quality games down the line.

Ethical issues also come into play. The data used to train AI often raises questions about the ethics of training data and unintentional biases affecting the final result. Even with careful data sourcing and testing, there’s a chance of unintentionally including problematic content. These issues could harm Microsoft’s reputation, resulting in boycotts, legal problems, and a loss of consumer trust.

When weighing the potential earnings from a possibly flawed AI-generated game against these numerous risks, it becomes clear that the losses from harming the brand’s image, losing market share, and facing negative public opinion would exceed the profits from a poorly received game. To protect Microsoft’s long-term reputation and its relationship with gamers, it’s essential to approach AI integration carefully.

Muse Has Many Years Before It Can Generate A Passable Xbox Game

We’re still quite far from having AI that can produce commercially successful games. Right now, AI can help with certain tasks, like creating game assets or designing levels, but it acts more like a helper than a full game creator. Some implementations involve using LLMs within games to vastly expand dialog systems. For AI to create an entire game on its own, we need it to become much more advanced. In my opinion, Muse looks like it still needs a lot more development time.

One major challenge is the visuals. Today’s AI can generate decent visuals at low resolutions, but struggles to produce the detailed, realistic graphics that modern games require. Creating lifelike textures, proper lighting, and believable character models takes a lot more computing power and advanced techniques than we currently have. Plus, turning these images into a functional 3D game with accurate physics is still largely dependent on human expertise.

Game mechanics are another tough hurdle. Models need to not only create visuals but also design engaging gameplay systems. This means developing challenging levels, believable enemy behavior, and smooth controls for players. For AI to reach the point where it can design such complex gameplay, we’d need significant progress in areas like reinforcement learning and game theory.

Creating interesting stories and gameplay experiences is perhaps the hardest challenge. While AI can write simple stories, it doesn’t yet have the ability to craft deep narratives that connect with players. It struggles to develop complex plots, create characters with unique personalities, or weave story elements effectively into gameplay.

This gap in creativity is likely going to continue for a long time, requiring major advancements in language processing and storytelling techniques. Considering all these challenges, it’s likely to take over a decade before we see a commercially viable AI-generated video game.

At this point, Muse isn’t going to be the tool to make it. The necessary advancements in AI research and the immense computing power required to train these models do not yet exist or are not available for the public to see.

This announcement suggests a future where AI can easily generate complete gameplay experiences, changing the way games are made and possibly bringing back older titles. Clips of AI-created gameplay shows what is currently available, but it clearly demonstrates how lacking it is overall.

The gap between what Muse can currently do and the ideal of fully AI-driven game creation is larger than many people realize. Even with Microsoft’s CEO claiming a catalog is coming, the reality is more complicated and less revolutionary than the hype suggests. Right now, Muse feels like a gimmick or parlor trick, not something that could make a minigame, let alone a fully-fledged game.