I bet it just doesn’t actually work, like most of their stuff.

I looked at some examples of what it generates, and it has some of the same problems of AI video: disfigured human crowds and it “forgets” the environment you are in, which sometimes results in the room around you changing from when you looked last time (in addition to these “games” being basically walking simulators without any interactivity or animations). This for now makes very boring “games” where nothing really matters because there’s no object persistency, and I don’t see how they are going to solve this issue because it looks like an inherent flaw of current AI technology (managing context windows is challanging also for LLMs).

IMHO this can be used at best a rendering technology, or to make a photo become explorable in 3d (with some made-up parts), but not games.

In case you want to also have a look at more than the cherry-picked examples posted by google, i found this video that has no commentary (not made by me - I hope it’s ok by the rules to post youtube links?) https://www.youtube.com/watch?v=jCTWCx8UPlI

google has figured out how to trigger a steam sale on gaming stock