Home Illustration AI is blowing up the world of illustration. Here’s how artists are rushing to catch up

AI is blowing up the world of illustration. Here’s how artists are rushing to catch up


This is a scenario that would have been unimaginable a few years ago.

Earlier this month, a popular Korean-language artist who goes by @ato1004fd on Twitch live-streamed an 11-hour sketching session, allowing his 22,000 followers to watch as they built an image of a popular video game character. Genshin Impact.

But by the time @ato1004fd had finished the digital painting, a dishonest viewer had already grabbed a photo of the work in progress from the feed, used AI to “complete” it, and posted their own version on social media, before back and accusing @ato1004fd of being the copier.

“Bro, when you ask your fans to cry about art theft, [be] reasonable,” the forger wrote. “So you took an AI image as a reference but at least admit it.”

A backlash ensued Twitter where users accused the impostor of gaslighting the original artist. Eventually, the forger deleted his account.

Recent developments in machine learning programs have transformed AI into an impressive artistic tool capable of overtaking – and undervaluing – human artists, setting off an earthquake in creative circles. Concerns are highest among graphic designers and commercial illustrators whose livelihood is tied to their ability to produce content to client specifications.

“It’s the Wild West right now,” said Liz DiFiore, illustrator and president of the Graphic Artists Guild. People are mostly worried about their jobs.

DiFiore cited information from pricing guidelines his organization released last year, which indicated that the median salary for an illustrator was $50,000, about $10,000 less than what the Bureau of Labor Statistics estimates as the median annual salary of artists.

“Wages have gone down over the years,” she explained. “And counterfeiting is probably the biggest reason why illustrators are seeing the value of their work drop.”

The anime art impostor used an image generator called NovelAI, a monthly subscription service that promises algorithm-assisted authorship and storytelling. The program includes an “image to image” service that allows users to upload images that can be edited through text prompts given to the machine. “The AI ​​basically learned to create images, like a person would,” the company explained in a recent blog post ending with the words in bold: “Please use our tool responsibly.”

AI image generators learn to create artwork based on large datasets with billions of images pulled from the internet. Most artists are unaware that their paintings and prints are housed in this vast online archive; companies believe they are not required to notify subjects of their extensive archives because the information is widely considered to be in the public domain.

Greg Rutkowski, dragon’s breath (2016). Courtesy of the artist.

In September, Polish artist Greg Rutkowski, a major name among fans of fantasy art and games, criticized this system after learning that it had become one of the most popular prompts for game creation algorithms. pictures. His name had been used over 93,000 times and people were beginning to confuse the computer output with his own work. A platform, Disco Diffusion, even suggested his name as a prompt.

“The AI ​​should exclude living artists from its database,” Rutkowski said at the time.

Popular programs like DALL-E, Midjourney and Stable Diffusion already have policies in place that prevent consumers from using their products in certain ways, such as appropriating images of celebrities and politicians.

Mat Dryhurst and Holly Herndon

Portrait of Mat Dryhurst and Holly Herndon. Image courtesy of the artists.

Because the powerful and wealthy have an escape from online image scrapers, many artists have started asking companies to remove their work from AI training models. Artists Mat Dryhurst and Holly Herndon recently launched Spawning, a tool intended to allow users to set permissions on how their style and likeness can be used by machines.

“After years of beating our drums about the importance of training data, there is now an appetite for discussion,” Dryhurst told Artnet News.

“I understand why artists involved in commercial tasks would be affected,” he added. “If the fear is that these machines will create satisfying art (pretty paintings, sculptures with modernist tones), then yes, that fear is valid. I believe this kind of art will be automated in the next two years.

The spawn includes a search function that crawls through a dataset called LAION, a 150 terabyte dataset that most AI image generators use for training. A quick search for living artists returned thousands of images in the archive of works by Peter Doig, Damien Hirst, Jenny Holzer, Alex Katz, Jeff Koons, Kerry James Marshall, Cindy Sherman and many more.

Screenshot of results for "Cindy Sherman" spawn

Screenshot of results from “Cindy Sherman” from Spawning, a website by Mat Dryhurst and Holly Herndon.

The rapid adoption of artificial intelligence seems to have taken most major galleries by surprise, which typically exercise strict control over anyone using their artists’ images. Representatives from Pace Gallery, Gagosian, and Sprüth Magers declined to comment on their policies regarding AI image generation.

Legal experts say artists, illustrators and galleries seeking to challenge AI image generators would have an uphill battle in court. According to Megan Noh, attorney and co-chair of Pryor Cashman’s Art Law Group, there aren’t many precedents for copyright infringement and artificial intelligence.

“The field is very dynamic and evolving,” she said, explaining that plaintiffs would likely have to prove substantial similarity between an original artwork and a generated replica that goes beyond a familiar style. “An artist doesn’t have copyright protection over style,” Noh explained.

Lacking legal remedies, commercial artists have turned to professional organizations for help in lobbying the tech companies behind the AI ​​image generators. DiFiore said his guild has seen an increase in membership over the past few weeks due to concerns. “I spoke to at least three new members who said they joined the guild because we were talking about AI”

Despite the concerns, DiFiore still hopes the technology can be used to improve the lives of the graphic designers she represents. “AI could eliminate some of the drudgery associated with our jobs, like resizing images, generating ad designs, and reviewing compositions from different angles,” she said. “These are things that often take time in the parts of the job that we enjoy. Any professional would be happy to see efficiency come into their work, while maintaining their salary. »

Follow Artnet News on Facebook:

Want to stay one step ahead of the art world? Subscribe to our newsletter to receive breaking news, revealing interviews and incisive reviews that move the conversation forward.