AI Beats Out Human Artists at Art Competition

At first glance, Jason Allen’s blue ribbon-winning work at the Colorado State Fair, titled “Théâtre D’opéra Spatial (Space Opera Theater),” looks like it could have hung on a wall of the Paris Salon. Women in expertly depicted draping gaze toward a heaven-like portal in a lavish baroque interior; a sun-washed cityscape lies outside. The work is ethereal, it feels historical, and its subject matter evokes luxury, grandeur, and 19th-century notions of exoticism.

Except that “Théâtre D’opéra Spatial” isn’t a painting, and its existence would have been impossible even five years ago. And not everyone is thrilled about that.

The image-generating artificial intelligence (AI) research lab Midjourney created the work using prompts provided by Jason Allen, the owner of a fantasy game company in Colorado. For “Théâtre D’opéra Spatial,” Allen wanted the technology to portray “a Victorian-style woman in a frilly dress wearing a space helmet.” He then spent over 80 hours fine-tuning his prompts to perfect the subject matter and lighting, creating 900 images in the meantime. When he had crafted the perfect artwork, Allen ran it through a pixel-enhancing AI, printed it on canvas, and drove it to the state fair.

“One of the first things that drew me to the work was the story aspect,” artist and writer Dagny McKinley, one of the two judges for the fair’s Digital Art/Digitally Manipulated Photography contest, told Hyperallergic in an email. “There are layers to the story — the stage, the actors facing away to another world. This work was one of those pieces that makes you want to sit with it for a while and explore the details of it.”

With only 21 entries, the Digital Art category was not the art contest’s most popular; eight categories ranging from Painting to Jewelry received a total of 596 entries, and the Colorado State Fair’s larger draws are its rodeo and livestock competitions. It was an unlikely stage for a debate over the ethics of image-generating AI and the longevity of art as a profession, but news of Allen and Midjourney’s win set the internet ablaze.

On August 30, Twitter user @GenelJumalon posted a succinct tweet that went viral: “Someone entered an art competition with an AI-generated piece and won the first prize. Yeah that’s pretty fucking shitty.” He included a screenshot of Allen’s (username Sincarnate) celebratory post about “Théâtre D’opéra Spatial” on Discord.

The Tweet garnered almost 90,000 likes and 13,000 retweets, with user @OmniMorpho chiming in: “We’re watching the death of artistry unfold right before our eyes — if creative jobs aren’t safe from machines, then even high-skilled jobs are in danger of becoming obsolete. What will we have then?”

Other comments point to art AI’s plagiarism issue. “My problem with this is not that they ‘used AI’, but that using AI means that everything here was very literally stolen off a bunch of artists’ work in a way that not even a homage or study does,” reads another popular comment by @tinytelephones.

To create content, AI generators scan millions of images in order to depict a specific subject (for example, “Victorian lady”) and thousands of images to emulate a target aesthetic (such as “Baroque”).

Allen told Hyperallergic that he created an encyclopedia to train Midjourney on the work of over 700 artists, both living and dead. He then used that training to create his blue-ribbon work.

“I wanted to find the best combination of artists that would evoke the visual language that I wanted to explore,” Allen said. After a slough of angry and fearful responses from artists online, Allen seemed to double down.

Ashley Martin’s “The Forager” won second place (photo courtesy Colorado State Fair)

“AI’s gonna keep getting stronger,” Allen said, explaining that artists should stop lamenting it. “I think it’s important that they recognize it and go through the stages of grief to accept it. You’re gonna hate AI for the rest of your life? It’s not going away.”

AI-generated art has improved exponentially in the last several years. Midjourney’s travel posters yielded impressive results this August, and last summer, artist Matteo Rattini trained a neural network to create images of contemporary sculptures with startlingly realistic results. Earlier this year, the accessible DALL-E image generator went viral with results that ranged from shockingly lifelike to downright terrifying. All users had to do was enter a prompt.

Jessica Hair’s third place winner “Pueblo West” (photo courtesy Colorado State Fair)

Allen, however, emphasized that AI “is a tool like a paintbrush” and wouldn’t exist without a creative mind driving it. McKinley agreed, comparing AI to Photoshop or filters.

The Colorado State Fair also deemed “Théâtre D’opéra Spatial” a legal entry. The fair’s director of communications, Olga Robak, told Hyperallergic that Allen and Midjourney’s work did not appear to violate any of the contest’s rules, although it “prompted internal conversations about how AI-generated art should be categorized and described.” People can submit a formal grievance, but as of September 2, no one had filed one.

“It’s really exciting to be at the forefront of the conversation of what art means, how do we judge it appropriately, and how do we categorize it fairly,” Robak said, adding that the fair began its art contests in 1887 and has added and dropped competition categories over the past 135 years.

“Programs like Midjourney open up new possibilities of what can be created,” McKinley said. “This is an exciting time in art and I’m so curious to see what comes next. Will AI learn to paint with a brush? Write a Broadway play? The possibilities are endless, but they all begin with imagination and creativity.”