Video games in the age of artificial intelligence: a revolution that raises questions

Ultra-reactive conversational robots, automatic generation of images, code or even scenarios… The video game sector is increasingly adopting these systems implementing artificial neural networks allowing for learning capacity.

“AI is really a turning point” for the video game industry, believes Julien Millet, artificial intelligence engineer, founder of the United Bits Game studio, present at the French Pavilion at Gamescom.

The less predictable game

In a sci-fi setting, a character in a black cap and drawn features, standing behind the counter of a ramen bar, responds immediately when a player asks him if he is well: “I’m not very well. , I’m worried “.

With this video of a few minutes, the American electronic chip giant Nvidia presented in May ACE, a program intended for developers, allowing them to “deploy intelligent characters” in their games.

No more interactions not exceeding a few lines of automatic dialogue: equipped with a microphone, the player can now chat with the character of the game, thanks to an AI operating on the same principle as Chat GPT.

Like Nvidia, many companies in the sector are advancing in this segment, making games more immersive.

“It brings unpredictability and therefore makes the game more real,” comments Sarah Brin, of Kythera AI, a company that offers an artificial intelligence service for character movements.

“Huge potential”

At Gamescom, a show that has brought together hundreds of thousands of people in Cologne since Thursday, professionals and amateurs, sometimes disguised as their favorite character, meet in huge halls, where each video game studio has its own stand. Visitors line up to test the new products in preview.

Unveiled in Cologne, the game Club Koala, from the Singaporean studio Kunlun Group, allows you to embody an avatar in a cartoon world. It promises the player to interact “with unique characters (…) animated by AI”.

“AI is now an integral part of everyday life. We see its huge potential to take the gaming industry to the next level,” said Fang Han, CEO of Play for Fun Studio, which owns Kunlun Group.

AI at the service of history and to create code

The use of these technologies goes beyond simple adaptation to the player. It gradually enters the process of creating the game.

“We use artificial intelligence to generate storytelling lines to enrich the story in the game, or even produce code,” explains Linus Gärtig, from the Berlin company Ivy Juice Game, who met at the show.

AI also allows producers “to make their vision better understood”, thanks to image generation models, which instantly produce an illustration from a text, according to Julien Millet.

What, however, threaten certain professions in the studios, such as the “concept artist”, whose role was precisely to translate the designers’ directives into images.

“I’m really worried about young people who pay thousands of euros for schools to become concept artists. What will their options be? asks Julien Millet.

The use of these technologies could also come up against the delicate problem of intellectual property.

The AI ​​actually train on pre-existing images or texts, sometimes protected. However, for the moment the law applying to the images thus generated is not clear.

“If you’re a big publisher and then you use generative AI, and it turns out that it violates some copyright, then you’re vulnerable,” says Sarah Brin of Kythera AI.

Unlike most competitors, the company therefore refuses to train its model on open databases.

In the United States, artists have collectively filed a complaint against Midjourney, Stable Diffusion and DreamUp, three AI models formed using billions of images collected from the internet.