Google’s Project Genie Shows “Meaningful Progress in Content Generation,” But it “Can’t Replace Creative Vision,” Says Analyst

Last week, Google debuted its latest generative AI experiment, Project Genie. It's a tool capable of generating interactive three-dimensional environments made entirely through generative AI that you can explore for one solid minute.

All you need to do (after paying the $250 monthly fee for Google's AI Ultra subscription) is write a prompt to start generating your own game-like worlds, but besides the paywall barrier, there are more than a few limitations to this technology in terms of it becoming the new way to make games.

The fact that it runs at a slide-show speed of 24 frames per second, the fact that you can barely interact with anything in the environment, that the worlds it generates are devoid of any kind of level design, or anything to really do in them, all come to mind before getting to the fact that these worlds only live for one minute at a time.

Still, that didn't stop stock market investors from panicking, causing the shares of major video game industry companies like Take-Two, Roblox, and Unity to take a nosedive, as if investors suddenly believed the age of game development as we've known it was over, and Project Genie was here to usher in a new age.

Obviously, that won't happen, and even if it does, as SuperJoost Playlist author and co-founder of SuperData Research, Joost van Dreunen puts it, "World models like Genie represent meaningful progress in content generation. But they can't replace the creative vision, narrative depth, and intangible elements that make games memorable."

Dreunen, co-founder of one of the foremost video game industry analytics firms, before selling it to Nielsen, offered his take on Project Genie after it was revealed and so significantly impacted the stock market for a day.

The top-level takeaway is the aforementioned obvious, that GenAI models like Genie cannot do the job of top-tier game developers. "Creating compelling game worlds is harder than it seems," Dreunen writes, "much like how AI can't simply prompt its way to the next great American novel."

Even as the technology improves over time, as these GenAI models certainly will, so long as major players like Google and OpenAI continue to invest in them, Dreunen argues that people won't want to play the games they generate, because people want to experience what expert artists and creatives can do - not what a piece of software that's only iterating on the actual creativity of humans can do.

There's also the fact that, currently, players and game developers have made it clear that they have no serious love for GenAI tech and games made with GenAI. Dreunen cites the downfall of InZoi, but you need only look at the backlash Larian Studios faced after it even flirted with the idea of using GenAI for concept art to see how players feel about the technology.

The only hope that GenAI world models like Project Genie and future iterations that will come after it have, if they have any hope at all, is potentially one day making something that people actually want to experience. Even then, it won't come without actual human input, with creatives using technology like Genie as a tool, if these technologies ever do find a place in game development.

"The horseless carriage phase will pass. When it does, the winners won't be those with the fanciest AI tools. They'll be the ones who used those tools to make something genuinely worth playing."

Follow Wccftech on Google to get more of our news coverage in your feeds.