Back in July, I wrote about a new artificial intelligence app called DALL-E 2. How it works is a user types a text prompt into a website, and within seconds, the app produces a collection of images. There are essentially no limits to the subjects of the images or the art styles.
The capabilities of that tool are currently on display at a new art exhibit at bitforms gallery within the Minnesota Street Project (located in the Dogpatch, San Francisco’s “coolest neighborhood”). “Artificial Imagination” features eight artists from across the country, with a heavy emphasis on Los Angeles, as well as a representative from the local scene: Berkeley’s Alexander Reben.
Given the questionable ethical behavior of many tech companies, being a technological optimist is hard these days. This type of art show doesn’t make it any easier. AI has boundless possibilities; In this use case, it has the power to democratize the creation of art, breaking the limits of craft and essentially serving as an imagination translator.
Or it can just be a bulls—t generator.
The four dumbest words that can be spoken in any art gallery are “I could make that,” but this exhibit proved a rare instance in which type of artistic skepticism was justified.
The two pieces from 23-year-old LA-based artist August Kamp — a black cosmonaut experiencing heartbreak and an abstract interpretation of a synthesizer — were aesthetically appealing enough. To her credit, Kamp shared her general prompts openly. For “But You Promised,” the words she fed into DALL-E 2 were something along the lines of “cosmonaut who is experiencing heartbreak on another planet.” Anyone could type those words and get a similar result, and she stated that the lack of personal ownership was part of what appealed to her about the technology.
“I love the idea that my art is not owned. I love the idea that if somebody sees my piece and thinks, I would love that style, but for this idea of mine — take it. That’s my entire draw to this type of technology.”
The cosmonaut’s face looks like a textbook visual definition of heartbreak, but although it shows the emotion perfectly, it tells the viewer little else. When I asked Kamp how much it was for sale for, she said she didn’t know.
The next artist I spoke with was Alexander Reben, a 37-year-old based in Berkeley who studied the intersection of robotics and art at MIT before becoming a full-time artist with commissions that included a sculpture for Meta’s Menlo Park lobby. He had one digital painting in the show, as well as two sculptures he built by following a description supplied by AI text generator GPT-3.
The first thing I asked was the name of the brightly colored painting, an abstract pastiche of swirling colors and patterns with a vaguely afrofuturistic sensibility.
“I don’t remember the name of that one,” he said. “I have six prints, and that’s one of them. The name for my purely generative works is an adjective and noun that the computer makes.”
When I asked him the prompts involved in creating it, he had a very different mentality compared with Kamp. He refused to divulge them, calling the prompts his “secret sauce.”
We then walked over to one of the sculptures, a 31-by-22.5-inch canvas covered in shingles, two iron bars, a lamp light, a handcuff and a pair of keys. The text instructions are displayed on a small museum placard. “So basically, the AI is telling me what art to make,” he said.
The title of this one, which he read off the placard, is “Ceci N’est Pas Une Barriere.” When I asked Reben the price of each piece, like Kamp, he told me he wasn’t sure.
A few minutes later, I looked up the prices: A run of five prints of “nominal_quiche” sell for $2,500 each. (Kamp cosmonaut also costs $2,500.)
The sculpture is for sale for $48,500.
The second-dumbest four words you can say in an art gallery are “why is that so expensive,” but in this case, I thought the question was justified. I went back and asked Reben.
“Values are usually figured out working with the gallery, along with the prices of works I’ve sold before. Over time they go up. I’d say also things that are larger, or one-offs versus series or multiples, are priced higher. So it’s kind of a mixture of things. I think it’s also part voodoo, from working with galleries over time,” he said. “Just because it’s priced, that doesn’t mean it’s necessarily going to sell at that price.”
Although the organizers of the event couldn’t confirm whether any of these had sold, several pieces of art are on hold due to the interest of multiple collectors.
Four more words you might hear from a skeptic in an art gallery are “is that really art.” While the question could be considered condescending, in a sense, it’s also the most important one. It would require a semester-long art history course to unpack, but a short answer is simply “eye of the beholder.” One of the pieces in the “Artificial Imagination” show did feel like art to me — a beautiful 30-minute video from Turkish American designer Refik Anadol called “Machine Hallucinations Nature Dreams Study 1.” It applied proprietary machine learning to 46 million nature images, which morphed seamlessly from one landscape to another, like a modern-day version of Philip Glass’ “Koyaanisqatsi.” It’s priced at $75,000.
I don’t believe this video is worth a down payment on a home in a modest American suburb, but I’m not angered by the voodoo art world economics here. The video inspired a sense of awe and curiosity — not just about the world but about the technology that made the art possible. It didn’t look like it was made by a magic wand and a few carefully chosen words.
The rub is that if you play around with one of these AI image generators yourself, the experience does really feel like you’re casting spells. As someone who blazed through 20 trial images on Midjourney in an hour, I can say these are incredibly fun tools, and there is indeed a subtlety to achieving results. But I used the prompt “futuristic synthesizer” and received startlingly similar results to Kamp’s piece titled “new experimental version, state of the art,” down to the same shade of orange.
It brings me no joy to disrespect someone who has devoted themself to mastering a craft in the pursuit of personal expression, but nothing about most of this art looked craft-driven or personal. At best, the mission of this art was unclear. At worst, it seemed to me like a test of these artists’ ability to convince collectors that the emperor, or in this case the algorithm, is wearing clothes — and to pay obscene sums of money to own that clothing.
AI can be used as a tool of expression, but here, it felt more like contract labor, merely a hired robotic hand ordered to draw a sad astronaut. These pieces may have looked like art, but they didn’t feel like art to me. There was a cold gap between the creation and the creator, made even colder by the wildly expensive price tags. The problem wasn’t the AI; it was the lack of imagination.