Last fall, the popular fantasy artist Greg Rutkowski — whose name had been included in more text-t0-image AI prompts than the likes of DaVinci and Picasso — teamed up with other artists to ask Stability AI, the company behind the popular open source image-generator Stable Diffusion, to remove his and other artists’ work from their models.
“I probably won’t be able to find my work out there because [the internet] will be flooded with AI art,” Rutkowski told the MIT Technology Review last September. “That’s concerning.”
By November of last year, Stability AI had responded to Rutkowski’s concerns by, amid a slew of other product changes, “nerfing” users’ ability to generate images in the style of specific artists. Of course, that’s a few steps below removing his work from training materials entirely, but it’s something.
Unfortunately for Rutkowski, however, it appears that he’s run into a new problem: open source AI.
As Decrypt reports, a version of a smaller model called a LoRA — basically a resource for Stable Diffusion that gives it specific areas of training — has now been made available to specifically churn out AI-generated work inspired by Rutkowski’s work.
“Since this artist name was one of the most used in 1.5 prompts and now it’s mostly gone, I thought it was a good idea to train a LoRA,” the maker of the Rutkowski-“inspired” AI wrote in the model’s description, as seen on the AI image generator-sharing CivitAI. “This isn’t perfect but it’s close enough.”
“Enjoy!” they added.
Considering that Rutkowski has been quite vocal about his AI objections, this seems wildly uncool. If there’s any bright spot for the artist, it’s that the developer behind the bot says he’ll take the model down if Rutkoski requests it.
“If he contacts me asking for removal, I’ll remove this,” the developer, who goes by just “Lykon” on CivitAI, told Decrypt. “At the moment I believe that having an accurate immortal depiction of his style is in everyone’s best interest.”
Still, even if Lykon removes their model from the platform, there’s really no way for the artist to protect against the creation of similar open source bots in the future. Open source models are incredibly hard to police, and at the end of the day, the cat is out of the bag. Without draconian regulation or legal precedent to force their hand, any requests by artists are going to be almost impossible to enforce.
It’s also a reminder that, like it or not, the Pandora’s box of AI is open. Even when bigger companies like Stability AI and others make some retroactive moves to make amends with the humans whose work was used to train AI models, much of the damage has already been done. And with open source models in the hands of developers across the web, it’s likely we’ll see similar bots to come.
“All that we’ve been working on for so many years,” Rutkowski recently told the BBC, “has been taken from us so easily with AI.”
“It’s really hard to tell whether this will change the whole industry to the point where human artists will be obsolete,” he added. “I think my work and future are under a huge question mark.”
More on AI art: Fully AI-Generated Influencers Are Getting Thousands of Reactions Per Thirst Trap
Share This Article