‘WI’m reading a book about Oman in a dreamy atmosphere under the night sky.” Type into the Deep Dream Generator’s Text 2 Dream feature. In less than a minute, you’ll get back an image showing what you’ve described. Welcome to the world of AI image generation. Here you can use a few text prompts to create what on the surface looks like top-notch artwork, even if you don’t really have any skill beyond drawing a stick figure.
AI image generation seems to be everywhere. On TikTok, the popular AI Manga filter shows you in Japanese comic style, and many people use it to create everything from company logos to picture books. Already used by one of his big publishers. SF publisher Tor discovered that the cover he made used a licensed image created by AI, but decided to proceed anyway “due to production constraints.”
Some of the biggest players in AI include companies like MidJourney, Stable Diffusion, and Deep Dream Generator (DDG). It’s free to use up to a point, which makes it appealing to people who just want to try it out. There’s no denying they’re fun, but a closer look at the images they produce reveals something strange. The images have a similarly sophisticated, somewhat kitschy aesthetic. And while there’s a thrill at first when you see an image appear, it’s not creatively satisfying.
The impact of AI image generation is far-reaching and could affect everything from movies to graphic novels. Children’s illustrators were quick to raise concerns about the technology on social media. Among them is author and illustrator Rob Biddulph, who says that AI-generated art is “the exact opposite of what I believe art is. Basically, art is inside you. I have always felt that it is about transforming an object into something that exists outside.Whether it is a sculpture, a musical composition, a writing, a performance or an image, true art is the final product. It’s more about the creative process than it is, and simply pushing a button to generate an image is not a creative process.”
Beyond creativity, there are deeper issues. In our online campaign #NotoAIArt, we’ve seen the artist share concerns about the legality of his AI image generator and its potential to devalue his illustration skills. To create images from prompts, the AI generator relies on existing art and text databases. They consist of billions of images scraped from the internet. The largest is his open source LAION-5B dataset used in DDG’s Text 2 Dream. DDG founder Kaloyan Chernev says the dataset consists of “mostly public domain images taken from the Internet,” but many artists and illustrators believe the database contains many images. It states that they often contain copyrighted images as well.
Harry Woodgate, author and illustrator of Waterstones 2022 Picture Book Award-winning Grandad’s Camper, said: Her illustrator, Anoosha Syed, makes a similar point. Sample everyone’s stuff and then mash it up into something else. “
Prompting image creators is very common, but they may also request an image based on another artist’s work, further blurring the ethical boundaries. Said says this can lead to the creation of images “intended to intentionally mimic my style” or images without the consent of other artists. While there is some debate that AI generators will work similarly to humans in terms of receiving, Biddulph said:
He adds: I apply them to my photographs with systematic and forensic accuracy. You begin to think, “I like the way Hockney juxtaposed the purple, green, and ocher blocks in the field paintings I saw in the National Gallery.” And then I try to add it to my photo. Inevitably, I won’t remember it and will probably end up creating in my own style something that bears a faint resemblance to what Hockney once painted. “
Syed agrees. Never move your hands like the original artist. AI cannot do the same. You can only copy. When a human artist “mimics the style or passes off the artwork as his own, it can be incredibly frowned upon and in some cases considered copyright infringement. This is fundamental. Essentially, that’s what AI art is doing.”
Chernev acknowledges “the complex ethical considerations surrounding the use of non-public domain images and their potential impact on artists like us whose work is used to train AI tools.” But there are more subtle dangers. It’s the ability to create potentially illegal images. Chernev admits that during the initial launch of Text 2 Dream, he “tried to generate images of naked children, even though no such images existed in the training dataset.”
He adds: In response, we quickly adjusted our tools to prohibit the generation of inappropriate or illegal content, including child nude images and her NSFW material. We are committed to using our image generation services responsibly and ethically. “
Chernev says DDG has reported the incident to authorities, but the artist is quick to point out that AI image generation as a whole is unregulated. Woodgate, who won Illustrator of the Year at the 2022 British Book Awards, and his Dapo Adeola want more regulation. Mr Woodgate said: “A welcome first step would be for the UK government to do away with the proposed copyright exception, allowing text and data mining for all commercial purposes, and instead advocating an opt-in license-based model. That way, future databases will be created using properly paid voluntary donations, he says.
Adeola agrees, saying, “The easiest thing to do is get permission from the artist to use their work and pay for it.” Chernev says DDG accepts requests from artists who want to be excluded from the system, but the “ask for forgiveness, not permission” model says asking for permission “should have been the first step.” does not mix well with Adeola. .
While children’s book illustrations remain largely unaffected, artists say, AI image generation has the potential to eliminate the small jobs budding artists often rely on to build their portfolios. I’m here. Syed said people could take advantage of his AI for fan art, self-published books, logos, family portraits, and more. “These clients are usually more concerned with saving money than the quality of the finished product,” she says. “They would prefer her use of AI if it means keeping costs low, so many of these small jobs will disappear.”
Adeola said the increased use of AI will also lead to a decline in the value of artists’ work. “For me,” he says. Things like this reinforce the argument that what we’re doing is easy and shouldn’t make as much money as we’d like.” Bidalf goes even further. “There is no question that AI-generated art is devaluing illustration,” he says. “Of course, people will begin to think that their ‘work’ is as valid as one created by someone who has had a career in art making. Nonsense of course. I can take great pictures of my daughters with my iPhone, but I’m not Irving Penn. “
At the moment, AI image generation is mostly used for fun, Chernev said. We believe that AI-generated content not only enhances the work of artists and designers, but also has the potential to create entirely new forms of art and expression. “
I don’t know much about artists and illustrators. “AI-generated art has a certain ‘look,’” he says Syed. “Over time, users will adjust to it and start leaving because of its dishonesty and ‘cheapness’. Also, I think we may even see a revival and appreciation of traditional media in response to AI. “
Furthermore, the illustrator firmly believes that the most honest critics and biggest fans (children and young people) are not convinced by AI art. “Children’s books are a very complex, multimodal form of communication,” he says Woodgate. “Children who read them have high expectations, not only for the stories and illustrations, but also for the people who make them.”