The Future Of AI In Gaming: What's Here And What's Next

Forbes

2 days ago

Decades before AI became a big deal, the 1983 movie WarGames portrayed a mainframe computer supposedly using artificial intelligence to play a game of thermonuclear war. It accessed huge databases, ran automated simulations without human oversight, and (finally) learned that playing with nuclear weapons isn't such a great idea.
Setting aside the two now-tired tropes—that computers can turn on us and that only teenagers know how to use them—it was a fascinating early glimpse into how video games, or at least computers that play games, might use artificial intelligence.
Now AI-based advancements in games are accelerating. Game developers are using AI to make games more realistic, more complex, and more challenging. However, it's not all fun and games: AI also brings a few downsides, including impacts on the game-development industry, privacy issues, and—potentially—great increases in energy consumption.
Shall we play an AI game?
AI, using a very general definition, has been in games for decades, and some of the common AI-based elements include the following:
Leveling up with AI
While games have long been vehicles for pushing the boundaries of computing and creativity, not all game types take advantage of AI. Julian Togelius, associate professor in the department of computer science and engineering at New York University (NYU) and director of NYU's Game Innovation Lab, says popular games such as Doom, Quake, and Wolfenstein from the 1990s have certain designs and features that are expected of such first-person shooters, and that such games don't really benefit from using AI. 'If you took Doom and put some super-fancy AI into these games' monsters, what difference would it make?' Togelius says.
But AI can make other games more interesting. Ubisoft, a large game developer and publisher, incorporated generative AI features into a prototype role-playing game, in which a large language model (LLM) improvises dialogue for NPCs on the fly. And, in September 2025, a startup called Ovomind plans to release a wristband that will measure players' physiological responses. It will identify emotions such as excitement and stress, and use AI to create-personalized experiences.
Togelius suggests that augmented reality gaming, wherein digital content is combined with real-world content, is a good prospect for more innovative, immersive AI-based gaming. One early example is Pokemon Go, in which players use mobile phones to locate and capture virtual Pokemon characters in real-world settings. The game uses only location information, Togelius points out. He notes, 'There are definitely huge potential uses for AI in converting the world you see in front of you into game content.' (Most of the glasses still make you look like a total geek, though.)
AI in gaming challenges
Even though AI can make games more challenging and immersive, there are also downsides. Take personalization, for example. 'Games that can adapt themselves to you and produce a new, interesting world that it knows is interesting to you … that's going to come,' says Togelius. But it can also go too far. 'Some people would think it's extremely creepy.' Games have also been called out over concerns about data privacy and manipulation to encourage spending within games.
Furthermore, when incorporating AI and its underlying LLMs into games, developers have to think about consoles' limited memory. Most can't run an LLM locally, says Togelius, nor is it cost-effective to engage the model from a remote server. The $50–$75 people pay for a game doesn't cover the incremental cost to power the data center that would host it. 'But I do think LLM-based non-player characters will come,' Togelius says.
AI is forcing changes to how games are developed, too, not all of which are welcome. While more than half of respondents to a survey of game-industry workers said they use AI tools, others are concerned that—as in other industries—AI could replace workers. That fear has a basis: Microsoft, one of the largest game developers, recently announced huge layoffs in its gaming division.
Generative AI tools can create stories, graphics, and other content by sampling the work of writers and artists (which some contend violates copyright protections). AI tools can also generate game code and perform the quality assurance (QA) testing. But Togelius, who also co-founded modl.ai—a company that makes AI game-testing bots—maintains that QA, at least, is ideal for AI to handle.
'Game testing, well, it sounds like fun; you get to play games all day. Until you realize that you're playing the same scenario hundreds of times, trying to see if every potential subgoal can be released under various conditions, and every time they do an update. It's a really grueling thing, and not fun at all.'
Ideas you won't find anywhere else:
SAP Insights tells you what's new and what's next for your business. Original research, unique ideas, expert takes – sent directly to your inbox twice a month. Sign up now.