AI still can’t create great video game worlds, and it may never be able to create them.

this is step backis a weekly newsletter that brings you the single most important story in the technology industry. For more news on the video game industry’s backlash against generative AI, follow Jay Peters. step back Delivered to subscribers’ inboxes at 8 a.m. ET. opt in step back here.

Long before the explosion of generative AI, video game developers were creating games that could generate their own worlds. Think of a title like minecraft Or even the original 1980 rogue That is the basis of the term “roguelike”. These games and many others create worlds on the fly using specific rules and parameters. Human developers work hard to ensure that the worlds you can create in games are engaging to explore and have plenty of things to do. At their best, these types of games can be replayed for years, as the environments and experiences feel new each time you play.

But just as other creative industries are pushing back against AI’s future downturn, video games will also see generative AI. However, we may never be able to match the best that humans can create today.

Generative AI in video games has become a lightning rod, with gamers angry about slop in games and half of developers believing generative AI is bad for the industry.

Either way, major video game companies are jumping into the murky waters of AI. PUBG Manufacturer Krafton is transforming into an “AI-first” games company, EA is partnering with Stability AI for “transformative” game creation tools, and Ubisoft is promising to “accelerate investment in generative AI for players” as part of a major reorganization. CEO of Nexon, which owns the company that created last year’s blockbuster arc raidersperhaps put in the creepiest way: “I think it’s important to assume that all game companies are using AI now.” (Some indie developers disagree.)

Big game companies often tout their efforts as a way to streamline and support increasingly costly game development. But the introduction of generative AI tools poses a potential threat to jobs in an industry already notorious for waves of layoffs.

Last month, Google launched Project Genie, an “early research prototype” that can generate sandbox worlds using text or image prompts that users can explore for 60 seconds. For now, the tool is only available in the U.S. to people who subscribe to Google’s $249.99/month AI Ultra plan.

Project Genie leverages Google’s Genie 3 AI world model, which the company touts as a “critical stepping stone on the path to AGI” that enables “AI agents capable of reasoning, problem-solving, and real-world actions,” and Google says the model’s potential uses “go far beyond gaming.” However, this attracted a lot of attention in the industry. This was the first real indication of how generative AI tools could be used for video game development, just as tools like DALL-E and OpenAI’s Sora showed what was possible with AI-generated images and videos.

In my testing, Project Genie was largely unable to generate an interesting experience, even from a remote location. The “world” does not allow the user to do much other than walk around using the arrow keys. After 60 seconds, you can’t do anything with what you’ve generated except download a record of what you did. This also means that you won’t be able to plug the generated content into traditional video game engines.

Indeed, thanks to Project Genie, we were able to produce a terrible unauthorized copy of Nintendo (which appears to be based on the online videos used to train Genie 3). This has raised a number of well-known concerns about copyright and AI tools. But they weren’t even in a world of the same quality as Nintendo’s world of hand-crafted games. The world felt silent, the physics were crude, and the environment felt primitive.

The day after Project Genie’s announcement, the stock prices of some of the biggest video game companies, including Take-Two, Roblox, and Unity, fell. As a result, some damage control became possible. Take-Two President Karl Slatoff, for example, strongly pushed back against Genie in an earnings call a few days later, insisting that Genie was still not a threat to traditional gaming. “Genie is not a game engine,” he said, noting that such technology “certainly does not replace the creative process,” noting that for him the tool is more like “procedurally generated interactive video at this point.” (The stock price rose again in the next few days.)

Google will almost certainly continue to improve Genie’s world model and tools to generate interactive experiences. It’s unclear whether they want to improve the experience as a game, or instead focus on finding ways to help Genie in its ambitious march toward AGI.

However, other AI company leaders are already pushing forward with interactive AI experiences. xAI’s Elon Musk recently claimed that “personalized,” “real-time,” and “high-quality” video games will be available “next year,” and in December he said building an “AI game studio” was xAI’s “key project.” (As with many of Musk’s claims, take his predictions and timelines with a grain of salt.) Meta’s Mark Zuckerberg, who is now promoting AI as the new social media after cutting jobs at Metaverse Group, envisions a future where people create games from prompts and share them with others in their feeds. Even gaming company Roblox is touting a way for creators to use AI world models and prompts to generate and modify in-game worlds in real time, something the company calls “real-time dreaming.”

But even in the most ambitious view that AI technology can feasibly generate worlds that are as responsive and interesting to explore as video games running locally on a console, PC, or smartphone, creating a video game requires much more than just creating the world. The best games have engaging gameplay, include interesting activities, and feature original art, sounds, writing, and characters. And it takes human developers, sometimes years, to make sure all the pieces work together properly.

AI technology is not yet ready to generate games. Anyone who thinks this is the case is fooling themselves. But AI-generated videos are still terrible, and they were still used to create tons of bad Super Bowl ads, so tech companies will probably continue to put a lot of effort into games made with generative AI. In an already volatile industry, even the idea that AI tools could rival anything made by humans could have major implications down the road.

However, the complexity of the game is different from AI video. AI video has improved significantly in a short period of time, but there are fewer variables to consider. AI game creation tools will almost certainly improve, but the results may never close the gap to what humans can create.