The creative thief: AI tools creating generated art
AI systems such as OpenAI's Dall-E, Midjourney and Stable Diffusion are used to create striking images. But it can be unclear if the images are inspired by others or stolen.
AI-generated art is painting a new path for both experienced artists and consumers. But critics say AI art systems are stealing real people's intellectual property.
Large AI models like GPT and art-generating tools such as Midjourney and OpenAI's Dall-E have enabled art amateurs to easily create striking, original-looking images, sparking a highly charged conversation about the nature of art itself and whether AI will eradicate artists' and illustrators' jobs.
Acclaimed supernatural fiction author Stephen King even joined the debate when he posted an image of "Pennywise on a bike" on his Twitter page, noting that it was "done by an AI bot."
The image of King's novelistic shape-shifting monster created by an AI system -- it's unclear which system created King's image -- looks like a masterpiece, but for one small detail. At the bottom-right corner is a signature, signaling that at least part of the artwork the AI system generated came from the million pieces of artwork that are online.
The uproar over the issue has become so intense that some photography agencies have begun scrubbing their work from the internet. Moreover, a new website, Have I Been Trained?, enables artists and consumers to check if their images are in the LAION-5B training data set -- a library of images used to train Stable Diffusion and Google Imagen, another AI-powered text-to-image system.
Also, the speed and scope of the AI text-to-image technology could lead to concerns about AI-generated art spilling over into the video editing field.
Meta, Facebook's parent company, on Sept. 29 announced its Make-A-Video text-to-video tool, which enables users to create a video by typing what they want to see.
Large AI text-to-image models might be misappropriating some real artworks.
AI fires the artist
One artist caught up in the debate about AI-generated art tools and systems is Grzegorz Rutkowski, a gaming illustrator from Poland. He has been an illustrator for games such as Dungeons & Dragons and Sony's Horizon Forbidden West.
Fans of Rutkowski told him that his pieces were being used to create new images generated by Stable Diffusion. In fact, his name has been used about 93,000 times, according to Lexica, a website that tracks images generated by Stable Diffusion from user prompts.
It was a future threat for my online portfolio.
Grzegorz RutkowskiGaming illustrator
"I realized that after a while, I can have trouble looking for my images on Google because it will be flooded with AI," Rutkowski said. "It was a future threat for my online portfolio."
AI systems like these could remove the need for mentorship between experienced artists and younger artists, Rutkowski said. Younger artists might decide to use the systems instead of gaining tutelage from those who are more experienced.
Also, the system can prompt organizations to do away with hiring junior artists and illustrators to create visuals that stimulate moods, such as landscape images, he added.
"Right now, we can see that at least part of the industry will be losing their jobs," Rutkowski said.
The role of the artist
While AI-generated art tools might cause the role of the artist to change, it's not going to erase the artist and their art, argued Cansu Canca, founder and director of the AI Ethics Lab.
"It could even become more valuable," Canca said. "You could imagine human art being this artifact that is hard to find and even highly priced, whereas computer-generated art would be not so highly valued."
The controversy over AI systems stealing work from others to create an artwork raises questions about plagiarism. But the changing forms of art from generation to generation and the custom of young artists using the works of older artists as inspiration blur the lines of traditional ideas about art, said Ricardo Baeza-Yates, director of research at the Institute for Experiential AI at Northeastern University.
"Let's say you use pieces of 1,000 images to create a new image. Is 1,000 pieces of other images plagiarism? I'm not sure," Baeza-Yates said. "If you say 10, I would say yes. If you say 1,000, I don't know."
Moreover, AI-generated art could help artists create their artwork faster, saving them time by relying on the engineers who created Dall-E and other AI art systems who took years to create their systems, Baeza-Yates continued.
'Theatre D'Opera Spatial'
Faster doesn't necessarily mean the artwork didn't require time to generate creativity.
Jason M. Allen, an artist and president of Incarnate Games whose Theatre D'Opera Spatial won first place at the Colorado State Fair on Aug. 29, said he spent 80 hours working on his piece.
The artwork was generated by Midjourney and submitted under the digital arts and digitally manipulated photography category.
Allen started using Midjourney earlier this year after being invited to participate in the beta-testing group. He said he was drawn to the program because the artwork he was seeing online felt like something that had never been created before.
"It's avant-garde," he said. "It's a new art movement."
He didn't start his award-winning project until he learned how to use the software to generate the images he wanted to see, he said.
"You have to learn how to manipulate the software because you might not get the results that you want to see," Allen said. "You constantly have to try different phrases and words and creative inputs that give you the results that you're looking for. And that's a skill."
The backlash over his winning artwork was unexpected, but understandable, Allen said.
"It comes from a place of denial," he said. Up until now, many were convinced that AI could never be as good as humans when creating art pieces, and now they've been proven wrong, he said.
"It would be better if they [didn't] deny the power available to them," Allen added.
Enterprises and AI-generated art
Meanwhile, enterprises in addition to artists and consumers are using AI technology that creates art.
Omneky is an AI marketing vendor that uses its own machine learning algorithms, GPT-3 and Dall-E to generate ads for its customers. The vendor also offers personalized prompts for image generation.
"AI is creating a freeing and flexible design process that inspires creativity and ultimately boosts performance," said Hikari Senju, founder and CEO of Omneky. AI is helping creators boost creativity and productivity, he added.
Virtuous AI, an AI ethics company that helps developers create explainable and unbiased AI systems, uses Omneky to generate ads and as a marketing tool.
While aware of the controversy of AI-generated art, Virtuous AI CEO Rory Donovan said certain applications of the tools can be useful.
"If you're, for example, displacing a person by creating the same content in the same style as them, that's definitely, in my opinion, immoral," Donovan said. "But I don't think that is really the case in these sorts of applications. We're not trying to generate Picasso paintings or anything like that. It's more of, you know, ad content."
Fighting back
However, for Rutkowski, the problem is the consequences of using AI-generated art tools and the vendors behind them.
"AI is affecting lots of fields -- science, technology, healthcare -- it's unstoppable and inevitable," he said. "We will learn to adapt to this world, but right now, as much as we can, we should treat artists and different people with respect."
For this reason, Rutkowski and other artists are pushing for regulation to protect the type of data that AI-generated art tools can access.
Artists like Rutkowski, who can show a strong similarity between their artwork and ones generated by AI systems, have a strong claim for copyright infringement, according to Michael Bennett, director of education curriculum and business lead for responsible AI at the Institute for Experiential AI.
"It would depend on how similar that generative art is," said Bennett, who has more than 15 years of experience as an intellectual property lawyer.
However, if the AI system created a piece of work that does not show a strong similarity, "there's not anything that living human artists can do," he continued.
Using existing images to inspire and help artists create art with AI tools is analogous to a young artist gaining inspiration from older works, so there isn't much harm, Bennett said.
As for the creators of the AI tools and systems, many artists including Rutkowski think they have a responsibility to other artists. In practice, this could mean removing the artwork of living artists from AI training data sets and banning it from being used, in the same way that some systems ban nude photos or images of celebrities or well-known individuals.
However, generative AI art is already proliferating quickly and spurring more creativity, with popular tools like Dall-E offering free introductory credits and affordable fees to use the system. The technology and people's use of it is sure to keep growing.
"If this technology is going to increase the rate of introduction of new works of art, increase the ability of more humans to produce art, then I think, in the end, it is a good thing," Bennett said.
