“Woman reading a book in a dreamlike atmosphere under the night sky,” I write to Deep Dream Generator’s Text 2 Dream feature. In less than a minute, an image came back to me showing what I was talking about. Welcome to the world of AI rendering, where you can create what looks like top-notch art on the surface, using just a few text prompts, even if in reality your skills don’t go beyond drawing garbage figures.
AI rendering seems to be everywhere: On TikTok, the popular AI Manga filter shows you how you look in Japanese comics style, while people in their herd use it to create images for everything from company logos to picture books. It’s already in use by a major publisher: the sci-fi series Tor discovered that a cover he created used a licensed image created by artificial intelligence, but decided to continue anyway “due to production constraints.”
The biggest players in AI include companies like MidJourney, Stable Diffusion, and Deep Dream Generator (DDG). They are free to use, up to a point, which makes them attractive to those who just want to try. There’s no denying that they’re fun, but a closer look at the images they produce reveals quirks. The face of the woman in my picture has very strange features and seems to be holding more than one book. The images are similarly polished, with a somewhat kitsch aesthetic. And while there is the initial excitement of seeing an image appear, there is no creative satisfaction.
The effects of AI rendering are far-reaching and can affect everything from movies to graphic novels and more. Children’s illustrators were quick to voice their concerns about technology on social media. Among them, author and illustrator Rob Biddulph says artificial intelligence-generated art is “the opposite of what I believe art to be.” Basically, I’ve always felt that art is all about translating what you feel internally into something that exists externally. Whatever form it takes, whether it is a sculpture, a piece of music, a piece of writing, a performance or an image, true art is more about the creative process than the final work. And just pressing a button to create an image is not a creative process.”
Beyond creativity, there are deeper issues. An online campaign, #NotoAIArt, has seen artists share their concerns about the legality of AI renderers and how they have the potential to devalue the illustration prowess. To generate images from prompts, AI creators rely on already existing databases of images and text. These consist of billions of images scraped from the internet. The largest of these is the open source LAION-5B dataset used by DDG’s Text 2 Dream software. DDG founder Kaloyan Chernev says the dataset contains “largely publicly available images from the internet,” but many artists and illustrators say databases will often also contain large numbers of copyrighted images.
Harry Woodgate, author and illustrator of Grandad’s Camper, winner of the Waterstones 2022 picture book award, says: “These programs are entirely based on the pirated intellectual property of countless working artists, photographers, illustrators and other rights holders.” This is a point echoed by illustrator Anoosha Syed: “Artificial intelligence doesn’t look at art and create its own art. He takes samples from everyone else, then turns it into something else.”
While requests to image creators can be very general, they can also blur ethical boundaries further by requiring an image to be based on another artist’s work. Syed says this could lead to the creation of images “intentionally intended to mimic my style” or without the consent of other artists. There is some debate that AI makers don’t work differently than humans when it comes to being influenced by the work of others, but Biddulph says: “A human artist also adds emotion and nuance and memory, especially failures, into the mix.”
He adds: “If I’m painting a painting and I decide it has to be Hockney-esque, I won’t be surfing the web for millions of Hockney-esque paintings, figuring out exactly what makes these paintings Hockney-esque, then apply them to my painting systematically and with forensic accuracy.’ I’ll think, ‘I like Hockney’s juxtaposition of purple, green, and ocher blocks in a painting of a field I saw in the National Gallery.’ And then I’ll try to add that to my painting. Inevitably, I’ll misremember it, and eventually it’s probably a painting that Hockney once drew. I’m going to create something that looks little like anything but in my own way.”
Artificial intelligence does not look at art and create its own art. Samples everyone else’s – then transforms it into something else
Syed said, “Another human will never look at a painting the way the original artist did. They will never move their hands the way the original artist did. The AI doesn’t do the same thing, it can only copy it.” When a human artist “imitates a style or makes a work of art his own, it is incredibly unwelcome and in some cases can be seen as copyright infringement. That’s essentially what the art of artificial intelligence does.”
Chernev says he is aware of “the complex ethical concerns associated with the use of non-public images and the potential impact on artists whose work is used in training AI tools like ours.” But there is a more insidious danger: the ability to create images that are potentially illegal. Chernev admits that during the initial launch of Text 2 Dream, people tried to “create images of nude children, even though there was no such image in the training dataset.”
He adds: “As AI continues to evolve, there is a risk that it may synthesize images of inappropriate or illegal topics based on existing content. In response, our tools will quickly prohibit the creation of any inappropriate or illegal content, including nude images of children and NSFW materials. We are committed to ensuring the responsible and ethical use of our image creation service.”
While Chernev says DDG has reported the incidents to authorities, the artists are quick to point one thing out, as not all of the AI image production is regulated. Woodgate and Dapo Adeola, who won Illustrator of the Year at the 2022 UK book awards, would like to see more edits. “A welcome first step,” Woodgate says, “will be to remove the UK government’s proposed copyright exception, allow text and data mining for any commercial purpose, and instead advocate models based on discretionary licensing.” That way, he says, any future database will be built using appropriately paid volunteer contributions.
For a fee, Adeola agrees, saying, “The simplest thing is to get permission from artists to use their work.” Chernev says DDG considers requests from artists who want to be excluded from their system, but the “ask for forgiveness, not permission” model doesn’t quite fit Adeola, who says getting permission “should be the first step”. .
While artists say children’s book illustration won’t be greatly affected, AI rendering has the potential to eliminate smaller works that aspiring artists often rely on for portfolio building. Syed says humans can turn to AI for things like fan art, self-published books, logos and family portraits. “These customers are often very concerned about saving money on the quality of the finished product,” he says. “They will choose to use AI if it means keeping costs down. So a lot of these little jobs will go away.”
Adeola says the increased use of AI will also lead to the devaluation of artists’ work. “For me,” she says, “there is already a negative bias towards the creative industry. Something like this reinforces the argument that what we are doing is easy and that we cannot earn the money we command.” Biddulph goes even further. “There is no doubt that art produced by artificial intelligence devalues illustration,” he says. “People will of course start to think that their ‘work’ is just as valid as those created by someone who has spent a career making art. Of course it’s nonsense. I can use my iPhone to take a nice picture of my girls, but I’m not Irving Penn.”
For now, AI image generation is largely used for entertainment, but Chernev says it’s “quickly approaching a level of sophistication and complexity that would allow it to create highly realistic and nuanced images.” I believe that AI-generated content not only has the potential to enhance the work of artists and designers, but also enables the creation of entirely new forms of art and expression.”
Artists and illustrators aren’t so sure. “Art produced by AI has a distinctive ‘look’,” says Syed. “As time passes, users will become more attuned to it and begin to turn away from it for its originality and ‘cheapness’. Also, in response to AI, I think we may even see traditional media re-emerge and be appreciated.
What’s more, the illustrators believe that their most honest critics and biggest fans – kids and teenagers – won’t be convinced by AI art. “Children’s books are extremely complex, multimodal forms of communication,” Woodgate says. “Children who read them expect a lot, not just from the stories and pictures, but also from the people who made them.”