Google's new AI 'world model' has seemingly spooked videogame investors, but it's hard to know what it will actually lead to

Google launched a new AI product this week, and as Seeking Alpha points out, videogame-related stocks like Unity and Take-Two took a dip. If those share price fluctuations really were a reaction to Project Genie, which was first revealed last year and is said to generate interactive "worlds," it's awfully premature—this doesn't do anything to GTA 6's prospects, let's be real.

But The Verge gave it a go, and was able to produce what the site called "bad Nintendo knockoffs," which is certainly interesting.

Google calls its current Project Genie model, Genie 3, an "experimental research prototype" (a funny name for something it's selling access to for $250 per month as part of its "AI Ultra" subscription) and says it enables users to "create, explore and remix their own interactive worlds."

"Unlike explorable experiences in static 3D snapshots, Genie 3 generates the path ahead in real time as you move and interact with the world," Google said in a blog post. "It simulates physics and interactions for dynamic worlds, while its breakthrough consistency enables the simulation of any real-world scenario—from robotics and modelling animation and fiction, to exploring locations and historical settings."

As a recent GDC survey showed, creative workers in the games industry are becoming increasingly hostile to generative AI, and Genie 3 has predictably not landed for everyone as the great benefit to humanity that Google says it's aiming for. "For once, CC: [email protected]," joked developer Rami Ismail about The Verge's Genie-generated Mario 64 and The Legend of Zelda knock-offs. (The Verge said that before it published its article, Genie 3 stopped accepting prompts to recreate Mario 64.)

The model does appear to be more advanced than what we've seen before, holding onto continuity in the examples shown (though they are brief clips) and apparently simulating physics reliably, though The Verge quickly found the expected jank, saying that one demo failed to maintain continuity and that the overall result was "much worse than an actual handcrafted video game or interactive experience."

Things like that horrid Darren Aronofsky slopumentary make it hard to feel optimistic that all this is just about researchers advancing human interests, and not big tech companies pursuing total cultural domination. But for the sake of a thought experiment, if generative AI tools weren't at all controversial and models like Genie 3 worked perfectly, do you think you'd use them? (For entertainment, that is. I'm setting aside possible research uses.)

I'm finding it hard to answer that question myself. As a kid, I of course wished that Star Trek's holodecks were real. But now, facing the possibility that I will in the future be able to type any scenario into my PC and instantly play it out in a custom simulation (though perhaps without Star Trek's corporeal holograms), I struggle to imagine enjoying it.

Would I care about the scenario or story knowing that there's no person on the other end intending anything by it? After the novelty wore off, wouldn't I just get bored of asking it to make me new Sherlock Holmes mysteries to solve? I haven't replaced listening to music I like with asking an AI music generator to generate music I like and then listening to that. Why would I do that for games?

Maybe I'm being too reductive, and the reality is that—as so many generative AI proponents like to say—this kind of thing will be used as a tool by human creators to expand what's possible, rather than as the end product. But even then, prompts are tiny compared to what comes back from an AI data center. Will I really accept a creative process that relies so heavily on machine predictions? What do you think?