Phil Spencer Posits Generative AI Tech Preserving Old Games, Called ‘Idiotic’ by Expert

This week, Microsoft introduced its newest project in the field of generative AI, which has been met with varying responses, to put it mildly.

The technology known as Muse is an innovative AI model, being the first of its kind, capable of recreating a game using video recordings and player control commands as its basis.

Microsoft’s AI researchers are collaborating extensively with Ninja Theory in developing Muse, a project where the AI has been educated using seven years of gameplay footage and player data taken from Bleeding Edge, an online multiplayer action game for Xbox One.

Based on all this data, Muse is able to reproduce the game environment and player actions.

Although the technology is relatively new, Microsoft appears hopeful that, over time, Muse may prove useful for facilitating the process of video game development.

Dr. Michael Cook, both an AI researcher and game designer, clearly explains his work in a straightforward manner in his blog post. Instead of creating unique games, as some might assume, he clarifies that Muse is primarily designed to assist developers in predicting player reactions when modifications are made to the game environment.

In simpler terms, Cook states that they created a tool enabling game developers to modify a game stage by incorporating established gaming elements such as introducing a jump pad where there previously wasn’t one. They then passed this updated level to their model and requested it to depict a video of a player playing from the new position, based on its understanding.

In the video mentioned earlier, Phil Spencer hints at a possible application of Muse technology in the future. This application might involve feeding vintage game footage into Muse, allowing it to recreate these games on contemporary hardware, bypassing the need for the original software or game engine.

Cook calls Spencer’s comments “idiotic”.

He points out that it can be challenging to grasp what AI models are capable of versus their limitations, as demonstrated by the case of Bleeding Edge. Even after being trained on years of footage, Muse won’t capture everything seen in the game or all potential player actions. In simpler terms, an AI attempting to recreate an entire game using video and gameplay data would never be able to achieve absolute accuracy.

While this model may accurately mimic the original game software, it doesn’t represent the ultimate solution for game preservation, Cook asserts. A generative model that shows what game footage might have looked like could serve as an interesting side note, but it will always fall short compared to other strategies we employ in our efforts to preserve games effectively.

It seems clear that we’re likely not yet at the point where solid judgments can be made based on what Microsoft has shown us thus far regarding Muse. However, given time, this technology might prove useful in the realm of game development, though its practical applications could still be a ways off.

It appears to be quite an exaggeration, given the opinions expressed by Cook and others in the industry regarding its application for preservation.

What do you make of Microsoft’s Muse AI tech? Discuss in the comments section below.

Keep in mind that certain links on this page are affiliate links. If you click on them and subsequently make a purchase, we might earn a minimal commission from the transaction. For further details, please refer to our FTC Disclosure.

Read More

2025-02-20 23:06