San Francisco: It sounds like a thought experiment reminiscent of Rene Descartes over the 21st century.
Simulated city citizens in video games matrix The franchise was awakened to a harsh reality. Everything was fake, the players told them through the microphone, and they were simply lines of code intended to adorn the virtual world. Empowering generative artificial intelligence like ChatGpt, the characters responded with panicked and mistrust.
“What does that mean?” said one woman in a gray sweater. “Are I the real thing?”
Released two years ago by an Australian high-tech company named Replica Studios, the volatile demo showed both the potential and consequences of enhancing gameplay with artificial intelligence. The risks go far beyond the unsettling scenes of a virtual world. What will the industry be like when video game studios become more comfortable outsourced voice actors, writers and other jobs to artificial intelligence?
With The Pace improving, large tech companies like Google, Microsoft and Amazon are relying on AI programs that will revolutionize the way games are created in the next few years.
“Everyone is about to race towards AGI,” says Tech founder Kylan Gibbs, using the artificial, common intelligence acronym. “There's a belief that once you do it, you basically monopolize all other industries.”
In the earliest time after the 2022 ChatGPT unfolded, conversations about the role of artificial intelligence in games were about how studios can help quickly generate concept art and write basic dialogues.
The application is accelerating quickly. This spring, the Game Developers Conference in San Francisco was greeted by thousands of enthusiastic experts looking for employment opportunities with an eerie glimpse into the future of video games.
Engineers at Google Deepmind, an artificial intelligence laboratory, lectured on a new program that could eventually replace human playtesters with “autonomous agents” that could run early builds of the game and discover glitches.
Microsoft developers hosted an adaptive gameplay demo, citing examples of how artificial intelligence can study short videos and generate design and animation levels that took hundreds of hours straight away.
The executive behind the online gaming platform has introduced Cube 3D. This is a generic AI model that allows you to generate functional objects and environments from textual descriptions in seconds.
These were not the solutions developers wanted to see after years of extensive layoffs. Another round of cuts in Microsoft's gaming division this month signaled some analysts that the company was shifting resources to artificial intelligence.
The studio is suffering as expectations for hyper-realistic graphics have turned into economic losses even for bestselling games. Also, some observers are worried that investing in AI programs in the hopes of reducing overhead costs could actually be an expensive distraction from industry efficiency issues.
Most experts have confirmed that artificial intelligence acquisitions will be coming to the video game industry within the next five years, and executives have already begun preparing to rebuild the company in anticipation. After all, it was one of the first sectors to deploy AI programming in the 1980s, with four ghosts chasing Pac-Man each responding differently to the player's real-time movements.
Sony did not respond to questions about the AI technology it uses to develop games.
“Game creators can always be central to the overall efforts of AI, allowing them to decide on using Generic AI, which best supports their own goals and vision,” said Yafine Lee, a spokesman for Microsoft.
A Nintendo spokesperson said the company has not commented further beyond what one of Miyamoto, one of the New York Times leaders, told the New York Times last year. “For example, there's a lot of talk about AI.
Over the past year, generative AI has shifted to a common tool within the industry, according to a survey released by organizers of the Game Developers Conference. The majority of respondents said their companies were using artificial intelligence, but the growing number of developers expressed concern that it contributed to job instability and layoffs.
All responses were not negative. Some developers praised the ability to use AI programs to complete repetitive tasks such as placing barrels throughout the virtual village.
Despite the impressive technical demonstrations at the meeting in late March, many developers have admitted that their program is still a few years away from widespread use.
“There's a huge gap between prototypes and production,” said Gibbs, who runs Inworld AI, a high-tech company that builds artificial intelligence programs for consumer applications in sectors such as gaming, health and learning. He appeared on Microsoft's conference panel, where the company showed off its adaptive gameplay model.
Gibbs said large studios can upgrade their technology in the face of millions of dollars in costs. Google, Microsoft and Amazon each hope to become the new backbone of the gaming sector by providing studios with AI tools that require studios to join servers under expensive contracts.
Artificial intelligence technology is developing so quickly that it goes beyond replica studios, the team behind the tech demos based on the “Matrix” franchise. Replica went out of business this year due to a pace of competition with large companies like Openai.
Eoin McCarthy, Chief Technology Officer of Replica, said that the demo's popularity means users are generating more than 100,000 lines of dialogue from non-player characters or NPCs.
As AI programs improve, costs have been declining in recent years, but most developers said they are not used to these infinite costs. There was also fear about how expensive it would be once NPCs started talking to each other.
When Replica announced it was closing the demo, McCarthy said some players grew up worrying about the fate of the NPCs. “Was they trying to live on or will they die?” McCarthy recalled the player asking. He said, “It's a technology demonstration. These people are not real.”
Large companies often throw away these moral questions in their presentations to studio executives.
Nvidia collaborated with a startup named Convai to infuse NPCs at a Cyberpunk ramen shop in real-time conversations. Verge showed a video showing Sony using Openai's voice recognition system and other technologies to create a version of Aloy, the hero of Horizon Forbidden West.
Some engineers are going even further, experimenting with AI programs that put faithful simulations of real people into the game. In late 2023, researchers from Google and Stanford University partnered with the creation of a generator agent.
“The generators wake up, make breakfast and head to work. The artists draw. The authors write. They form opinions, notice each other, and start conversations. In a virtual world inspired by the Sims, these agents were in relationships with each other and planning Valentine's Day celebrations at the cafe.
Some ethics experts praise the development of technologies that could put a certain amount of strain on acquiring human subjects. But others question the points of technology that can only replicate people's choices.
“Humans should be at the heart of our work,” said Celia Hordent, a user experience and cognitive science expert who has developed codes of ethics in the gaming industry. “It might be a better starting point to have a better process rather than thinking of AI as the whole solution.”
Many of the current programs that can automate game development are still quite expensive and full of glitches. Entrepreneurs preach patience, saying that the available models will likely take another five years to improve quality and reduce costs.
Gibbs said the adaptive gameplay model that appears during Microsoft's meeting sessions probably costs hundreds or even thousands of dollars per hour to run commercially. A similar program called Oasis has its own problems, he said. It generates content on a frame-by-frame basis, so players forget that visual information doesn't appear on the screen anytime soon, leaving players in a constantly changing environment.
The technology shows promise, but he said it is still the answer in search of problems.
“How do you push the research community in a more useful direction?” he asked. “It's a cheap way to make a game, but it costs 5,000 times more to run the game, so is it actually cheaper?”
Beyond the dollar signs, ethics experts focus on the question of how prepared the industry is for the perceptual characters and levels that design themselves.
Cansu Canca, director of responsible AI practices at Northeastern University in Boston, said normalizing technology presents risks to individual agents and privacy.
“My biggest concern is not that AI will gain awareness,” she said. – ©2025 The New York Times Company
This article originally appeared in the New York Times.
