![[PS5] Review of Choice of Life: Wild Islands](https://www.nogameover.net/wp-content/uploads/2025/07/PS5-Review-of-Choice-of-Life-Wild-Islands-360x180.jpg "[PS5] Review of Choice of Life: Wild Islands")
Yesterday, Microsoft Xbox introduced an intriguing development called Muse—a generative AI model intended to spark creativity in gameplay design. This launch was accompanied by an open-access article on Nature.com, a blog post, and a YouTube video that offered more insights. Now, if you’re scratching your head over the term “gameplay ideation,” you’re not alone. Microsoft describes it as the creation of “game visuals, controller actions, or both.” However, it doesn’t quite revolutionize the game development process as one might hope.
Despite its limitations, some aspects of the project spark curiosity. The scope of the training was massive, utilizing H100 GPUs. It took about a million updates just to stretch a single second of real gameplay into nine more seconds of simulated gameplay, closely mimicking the game engine. Existing multiplayer gameplay served as the primary training data source.
Instead of simply executing the game on a solitary PC, Microsoft employed a formidable cluster of 100 Nvidia H100 GPUs. This setup, although exponentially more costly and energy-intensive, managed to generate a rather modest output resolution of 300×180 pixels extending gameplay by nine seconds.
What caught attention was how Muse could replicate existing props and enemies within the game environment, mimicking their function. Considering the extensive hardware, energy consumption, and AI training involved, one might wonder why they didn’t just use traditional development tools to spawn these elements.
Though Muse did show impressive capabilities by accurately maintaining object permanence and replicating game behavior, its purpose seems inefficient compared to conventional game development methods. While future iterations of Muse might surprise us with more remarkable capabilities, for now, it joins a long list of attempts to simulate gameplay entirely through AI. Despite achieving some level of engine accuracy and object permanence, this method doesn’t seem practical for developing, testing, or playing video games. After hours delving into the details, it’s hard to grasp why anyone would prefer this approach.
For those keen on keeping up with the latest in tech, Tom’s Hardware offers an excellent newsletter delivering top stories and detailed reviews right to your inbox.
