I Enrolled Artificial Intelligence in Art School: The Postmodern Maestro who Taught Machines to Elevate His Classic Artworks

byWhen you read this article, it is more likely that it has already been scanned by an artificially intelligent machine. When asked about artist David Salle, large language models such as ChatGpt and Gemini may reuse some of the following words to come up with an answer: The larger the dataset, the more convincing the response. And since it first appeared in Art World Stardom in the 1980s, the monkeys have been thoroughly written. The question is whether AI can say new things about the artist and his work, or are they being criticized for producing more of the same thing?

Similar questions remain below the surface of the paintings that Saal has been doing since 2023. He is a new series just announced at Thaddaeus Ropac in London. His new idyllics were created with the help of machine learning software, but that is not immediately clear from seeing them. Each monumental canvas has a wide range of gesture strokes of oil paint that appear to be applied to the artist’s own hands. However, a thorough investigation reveals large patches of digitally printed flat patches. This is the mark of the AI ​​model that Salle trains to produce his work.

This mechanical collaboration began with the game. Sale has long been skeptical of digital painting tools; Written in 2015 “The web’s enthusiastic sprawl is the opposite of the kind of focus needed to make paintings. Nevertheless, his own paintings have a vast quality. It overlays images from a very wide range of pop and art historical references, where the eyes often don’t know where to rest. In 2021, Salle got the idea to develop a virtual game that allows players to reposition those paint elements using drag-and-drop tools.

The technology proved unrealistic, but along the way, Monkey met Danika Laszuk, software engineer at technology startup Eat__works, and Grant Davis, creator of the AI-powered SketchPad app Wand. Together they supplied the work of AI image generators by artists that Sarle of Technique considers as basic – Andy Warhol asked Edward Hopper for Color, Giorgio de Chilico for Point of View, Arthur Dub for Line – to create images based on a specific text prompt. “What I did was send the machine to art school,” says Sarle.

At first, the machine was not a model student. The creepy cartoon-like figure reminded me of an image created by Openai’s Dall-E Mini, with an unnatural shiny cartoon-like figure. “What are your fundamental complaints about this type of digital image?” Sale wondered. The answer, he felt, lived along the edge of the shape. “It’s the only pixel, so there’s no real difference between the edge of something and what’s behind it,” he explains. “There is no way to make edges meaningful. In representational paintings, these edges have this much information about the artist’s style.” So he gave the machine a scan of the gurch he was making and saw the AI ​​react to their water rim. “You can read the physicality of brush strokes,” he recalls. “It fundamentally changed the way machines think about it.”

Salle has taught at various institutions over the years, and this process felt like a feedback session at an art school. AI models are extremely fast learners and can generate images that Sale may have come up with herself over a few sessions. “A machine can synthesize things in seconds,” he says. “This evolution in the language of painting may take years.”

Born in Oklahoma but raised in Wichita, Kansas, Searle was an early model when he jumped into the New York scene in 1980. By 1987 he became one of the most acclaimed painters of his generation, and at the age of 34 he became the youngest artist to conduct a midterm survey at the Whitney Museum of Art in America. This stratospheric rise meant that Searle’s market would fall even further if the archaic paintings were no longer fashion, but the artist continued to work more ambitious series than ever, including a nasty suite of paintings at the heart of the new pastor. Now the figic paintings have returned in vengeance, and Saleh met at that moment.

To produce his latest work, Monkey trained AI with dozens of sweeping pastors completed between 1999 and 2000. It features lake idlings copied from 19th-century opera crims, which showcase landscape paintings through visual glass and rendered in Harlequin colors, covered with images from the set and distinctively different designs. These views appear to have been edited in Photoshop, but Salle rendered them perfectly similar. “The first thing they teach in their color painting classes is how to establish a palette,” he recalls. “I was really showing off. I thought, ‘I can create a painting with three independent color palettes working and it can’t stop me.’ ”

The first reviews of Idol were mixed. Some critics described them as cold and emotionless. This is often imposed on Searle. “The results seem more like a smash of ingredients than a thoroughly cooked dish,” says David Frankel I wrote it on Artforum.

But in the new pastoralism, something seems to have changed. Brush strokes are looser, faster, thicker, more abstract expressionist than anything else Searle has ever done. Meanwhile, their subject is entirely Helter Skelter, as if Searl pulsated his old paintings with a blender. The headless body enlarges and expands the frame. The object appears to be not attached to solid ground. Unsettlingly, these works appear more hand-drawn than the referent, at least until they are close to the surface. There, the paint was very thin in some areas and could only be applied on a machine.

Tharle always felt he hadn’t finished his idyllic, but he says there are other reasons why it’s a good match for his AI experiments. “I realized that those paintings would give machines the material they would understand in their own way,” he explains. “We took all these facet shapes with harmonies of different colours, and it went to town with them. But we kept the DNA of the transition, the horizon, the mountains, the water, the couple, the person, and overlay it with a brush stroke edge.”

Wand App Creator Davis sees the introduction of idyllic as a groundbreaking model moment. This is because new techniques were needed to “finally capture images and abstract content at a conceptual level to reproduce various variations.” In other words, the machine began to digest Searle’s paintings on a formal level. Pastorality may also be suitable for tasks as it resembles certain digital technologies in the production of fantastical spaces, particularly theatrical backgrounds. The new idyllic rejects the historical orders of art to create depth in flat painting.

Salle can instruct the AI ​​model to generate something similar to what is similar to her work. “Essentially, it’s a lever that runs along a continuum. One end is similar, the other end is similar,” he explains. “Depending on where the lever is placed, the results will be very close to what you started or it will be significantly distorted.”

For example, a tree, mountain and idling couple from his former pastor is still visible on the background of a red scarf, but is now drawn semi-abstractly behind a woman wearing a neckerchief. The floating stack of teacups anchors the composition in an otherwise turbulent space, like the claws that Picasso and Black painted in a cubist still life. On the other hand, many fragmented, collision bodies are much more difficult to decode. “I choose a very eccentric image and cause myself to do something I would not otherwise have,” says Searle.

Nevertheless, the artist is not worried that AI will kick him out or replace him. He sees it as another tool, like a brush or easel. “I don’t think the machines taught me at all,” declares Sarl. “I’m not rethinking how I think about the spaces of the art or the composition. I’m simply taking in what the machine has to offer after saying what I want.”