Ammaar Reshi thought of it as just a fun, creative idea: Use artificial intelligence tools to write and illustrate a children’s book that he had always wanted to make for a friend’s daughter. He gave himself only a weekend to do it.
Using ChatGPT and Midjourney, Reshi generated drafts of text and illustrations that would stitch together a story that would show the magic of AI to children, as he put it. Both programs, free for at least a trial period, require the user to type prompts that then refine them by regenerating images or text.
The end result is impressive to anyone unfamiliar with AI but often far from perfect: Images tend to appear with strange anomalies — in Reshi’s case, crooked eyes and 12 fingers — and text created by ChatGPT can have quirks and errors that remind us that AI is not quite human. Reshi spent hours refining prompts and editing text generated for the book, and he rejects the criticism that all he had to do was “hit a button.”
He has sold more than 900 copies since he put his book, “Alice and Sparkle,” on Amazon in early December. But a look at the reviews — 60 percent 5 stars and 40 percent 1 star — as well as his Twitter mentions suggest a growing divide over these tools as the public considers whether they’ll starve the starving artist, or if they’re ethical at all.
“The man who made [this] isn’t an ‘author,’ nor is he an ‘illustrator,’ yet in his bio above he claims that he ‘writes,’” one Amazon reviewer wrote. “Our world is turning into a joke.”
Reshi doesn’t hate the technology, but he understands why some would be worried.
“With any kind of new tech that is incredibly powerful, it’s somewhat threatening to people,” he said, adding: “You see people wondering, ‘Will this replace my job?’ … That concern — we shouldn’t pretend like it isn’t a serious one.”
One of the main complaints about AI art, for instance, is that some tools appear to have learned from data sets of art created by real people — with real copyright protections — to provide the fodder for its computer-generated creations.
Reshi doesn’t have an answer for that: “People say, ‘Well, if this model is trained on my artwork, and my artwork is copyrighted, is this exactly fair or legal?’ But then I think you’re going to get into this philosophical debate, which is, how is this different than a human learning [about] their favorite artist or someone drawing Batman fan art? One could argue that the computer is doing the same thing here.” He adds, “I don’t have a very concrete stance here yet.”
Already, AI has made its way into the creative world. Last summer, a Colorado man won the state fair’s art competition with an image generated on Midjourney. In November, the Lensa app debuted a new feature that sent AI selfies flooding into social media feeds. A comedy robot created by an Oregon State University professor has begun learning how to gauge the crowd in how it times and tells its pre-written jokes. Shudu, the “world’s first digital supermodel,” was created through artificial intelligence and has been used in a Louis Vuitton ad.
Some high-profile creators have made their disdain for this technology clear. The Australian singer Nick Cave recently called ChatGPT an exercise in “replication as travesty” — and a song it wrote in his style “a grotesque mockery of what it is to be human.” During a presentation on artificial intelligence, the famed animator and Studio Ghibli co-founder Hayao Miyazaki referred to the technology as “an insult to life itself.”
Online, artists have also banded together to stage a digital protest of AI-generated art. Last month, many railed against the platform ArtStation after AI-generated images appeared on its site. One protest image implored AI users to “pick up a crayon like the rest of us did.”
Earlier this week, a U.S. law firm announced a class-action lawsuit against Midjourney, Stability AI and DeviantArt, alleging that “billions of copyright images” were used in a data set “without compensation or consent from the artists.”
“AI image products are not just an infringement of artists’ rights; whether they aim to or not, these products will eliminate ‘artist’ as a viable career path,” a release from the Joseph Saveri law firm stated. It added, “If streaming music can be accomplished within the law, so can AI products.” The law firm did not respond to interview requests from The Washington Post.
Nik Thompson, an expert in human-computer interaction at Curtin University in Australia, said that he has heard of cases where a real artist’s signature has appeared in an AI-generated images, and that creators “are quite rightly upset about this.”
“The thing is: The cat’s out of the bag and there’s no going back, so I don’t think litigation is going to stop these platforms from continuously developing and gathering up as much data as they can,” he said. “It’s going to keep happening.”
Thompson believes that many are overestimating the present level of sophistication in AI programs like ChatGPT or Midjourney, both of which were released in the past year. Artificial intelligence is actually just “a simulation of intelligence,” he said — it cannot think like a real human.
“Over time, we’re going to realize it’s not as fantastic as it might seem,” he said. He added: “I would like to believe that the discerning consumer who appreciates art and creation will still be able to notice the difference and gravitate toward the work of creators.”
After an explosive backlash on Twitter, Reshi “braced himself” before sharing his latest personal project with the public — a fictional, animated Batman video he put together using an edited version of a script he generated on ChatGPT. He generated images on Midjourney, scaled them to larger resolutions using AI functions in Pixelmator, and then recorded himself doing a voice-over that he edited using an Adobe AI tool. He edited the video on the phone app Motionleap.
“I saw claims that this is going to replace storyboard artists,” he said. “I actually don’t agree with that take.”
Though he acknowledges he may be too optimistic, he said he hopes professional creators can also find a use for these tools. Storyboarding artists or illustrators could test their ideas by generating them with AI and then use their hard-earned skills to create a more refined product, he said. Amateur creators might also use these AI tools to help their visions come to fruition, as he did with his Batman video, he said.
As it stands, some amateur video game developers have begun looking to Midjourney to generate game assets and graphics, while others have used the program to brainstorm visuals for an indie board game.
“A lot of people see this as empowering a new set of creators — the kids who couldn’t illustrate or write as good of a story. Now they might get a head start or a jump on this,” he said. “I view this as an equalizer in many ways.”