Intrigued by the potential of generative artificial intelligence for video game design, studio Blizzard Entertainment trained an image generator on their hit titles. Feeding the machine assets such as World of Warcraft’s warlike orcs, Diablo’s spooky dungeons, and Overwatch’s hilarious heroes, Blizzard can easily create new idea concepts for his art.
Because generative artificial intelligence creates art faster than humans can, studios like Blizzard, a division of Activision Blizzard, are using the technology to make design and development less tedious and more fun to make video games. I hope it will.
Blizzard Chief Design Officer Allen Adham told employees about the initiative last month in an email obtained by The New York Times. Its internal tool is called Blizzard Diffusion. It’s a play on Stable Diffusion, his one of the popular image generators that allows anyone to turn text into art.
“Be prepared to be surprised,” Adham wrote, adding, “We are on the verge of a major evolution in how games are built and managed.”
Generative artificial intelligence, the technology behind tools like ChatGPT and Midjourney, uses significant computing power to identify patterns in text and images and generate new content from data.
Some researchers have warned of the technology, warning that it could help piracy, job displacement, and the spread of false information. But video game developers are already relying on artificial intelligence to help non-player characters make human-like decisions, and leveraging generative AI to speed up the creation process in a labor-intensive industry plagued by delays. I believe we can improve.
Microsoft, which makes the Xbox console, has a games AI division, and Ubisoft has developed a tool called Ghostwriter that can generate basic dialogue for games like Assassin’s Creed. Several startups say their technology will make it easier to design the appearance of non-player characters, known as NPCs, which give video game worlds a heavy feel.
Chris Lee, former studio head of Halo Infinite at 343 Industries, said generative AI could improve game development by reducing the human effort required to create giant open-world games.
“Game developers have historically been unable to meet the demands of their audiences,” said Lee, now the head of immersive technology at Amazon Web Services.
Halo Infinite was supposed to be the flagship launch game for Xbox Series X in 2020, but after an eight-minute preview was released, its graphics were derided by fans as flat and ugly. The studio ultimately delayed the game’s release by another year.
According to Lee, the game’s developers were miserable because they had to move pixels slowly every frame, even when working on placeholder encounters. “Loading this huge world is a pain, like specialized data entry,” he said.
Generative AI can also streamline quality assurance testing. At a recent conference for game developers, Kate Rayner, technical her director at Gears of War development studio Coalition, explained how AI can be used to detect bugs and glitches so that players can expect more on launch day. Talked about how to reduce the crashes you encounter.
Much of the promise of generative AI is speculative, and Blizzard has already abandoned its patented machine learning techniques to create environmental textures like stone and brick.
Andrew Guerrero, Blizzard’s vice president of global insights, said the tool takes up too much artist time to be effective. But he said another AI tool is helping to fit cosmetic headpieces onto World of Warcraft player models.
“Our goal is to eliminate repetitive, manual processes and allow artists to spend more time creatively,” Guerrero said in a statement. “Our goal with AI is and will continue to be to make creative work easier.”
An internal email about Blizzard Diffusion states that it is used not only for characters and their outfits, but also for generating concept art for game environments. It also touches on the potential of tools for “autonomous and intelligent in-game NPCs,” “procedural-assisted level design,” AI-assisted “voice cloning,” “game coding,” and “anti-toxicity.” increase.
Ubisoft’s AI dialogue tool “Ghostwriter” was requested by writers facing the daunting and sometimes tedious task of cramming more than 100,000 lines of dialogue into an open-world game, the company said. .
Many of these lines are character background chatter that helps simulate the living world. A single everyday interaction may require a dozen or more variations.
In Ghostwriter’s promotional video, the employee begins with the prompt “I used to be an adventurer like you” (a nod to Skyrim’s infamous line) and polishes up on some AI suggestions. A famous adventurer in this land” and “I remember when I was young and strong.”
These simple conversations are how people start careers in video game writing, and developers argued on social media that automating these tasks could threaten such jobs. Simon Johnson, an economist at MIT Sloan School of Management who has published a new book on the impact of automation, said it was a bad idea for tech companies to invent algorithms that mimic humans.
“We should focus on machines that help improve human capabilities, not replace humans,” he said.
Yves Jaquier, executive director of Ubisoft La Forge, the research and development team behind Ghostwriter, said that when motion capture was introduced decades ago, it would replace video game animators. He said he had similar, but unfounded, concerns.
“There may be more technology involved in the future, but that doesn’t mean humans won’t need to be involved,” Jaquier said in a statement. “Artists, writers and programmers have always been at the center of the development process, and AI can now better assist creators in their workflow, but it is the personal artistic vision and perspective that is integral to the creation of games. “
Another unavoidable concern with AI-generated content is copyright. In one high-profile lawsuit, Getty Images accused Stable Diffusion of scraping 12 million images of him from its photo database.
Activision Blizzard employees received an email earlier this month from the company’s chief technology officer, Michael Vance, warning them not to use the company’s intellectual property in external image generators. (Microsoft is looking to buy Activision for nearly $70 billion, but regulators want to block the deal.)
“These new tools come with new and unknown risks. We will proceed cautiously to avoid pitfalls,” Vance wrote in an email obtained by The Times.
Some Activision Blizzard employees noted that the company’s AI tools don’t always perform as promised, and have struggled to find bugs or work properly with the game environment.
“The leadership’s focus on AI doesn’t feel like it’s solving problems that individual contributors care about,” said a former World of Warcraft engineer last August. Valentine Powell, who retired from Blizzard, said. “I feel like we’re ignoring their problems and focusing on hype words that we think will sound impressive to our shareholders.”
Small video game studios without the resources to create generative AI tools are looking to startups for help.
Scenario, which raised $6 million in seed funding in January, will create an image database to turn text prompts into assets that can be incorporated into games, such as a lizard in a space suit.
That image generator is used in several smaller games. Times tests sometimes spat out animated characters with unrealistic hands, which was a common weakness of image generators.
Another startup, Didymo, said Soleil Game Studios, which makes fighting games based on works like Naruto and Samurai Jack, has created hundreds of non-player characters using AI-powered generators. said he did.
“We have automated the process so that you can eliminate the mundane work because it gets boring after a while,” said Sean Cooper, developer of Didimo.
Video games can take years to make, and there are very few public games using generative AI in the early stages of the hype cycle. But some companies like Niantic, the creators of Pokémon Go, are experimenting with tools for creating marketing materials. News release in the tone of Dr. Seuss.
Niantic also used ChatGPT during the development of Peridot, a new augmented reality game leveraging nostalgia for 90s digital pets.
Kelly Santiago, head of production at Niantic, said filling a room with creative talent is expensive, and the technology has condensed the time needed to generate ideas.
“For me, I’m glad you took the first six hours of brainstorming for yourself,” she said.
