Unlock Editor’s Digest Lock for Free
FT editor Roula Khalaf will select your favorite stories in this weekly newsletter.
There are certain types of artists who are eager to go beyond what they have in their heads. For a long time I have wanted a truly adaptable, painting-like space. It is something that has the speed and dynamism of great abstract paintings made of images and made only from formation.
There is also a history of artists embracing innovation: Silkscreen, Video, Neon. Even oil paint was all new once. The photographs were thought to show the death of paintings, even artists as diverse as Thomas Eakins and Edgar Degus began to use them in their works. Now we have artificial intelligence. Will AI help to realize my vision of a fluid pictorial space exempt from the law of gravity?
The AI art I saw was not promised. The more futuristic the image, the more mediocre the effect. The images don’t seem to have an inner life. Was it the technology or the user’s fault, a car or driver’s fault? What does the machine “do not understand” about visual arts? For one thing, digital images should not be confused with paintings. The pixels have no surface textures and no physical presence at all. Paintings are objects of the world, no matter how much brain you are. I am a painter and not an engineer. I am in the distance of something made by one author by hand. Still, I have spent the last two years teaching computers to think like an artist.
Artificial intelligence is basically averaging the world of images taken from the Internet. The fundamentals of representational painting were not part of the machine’s neural network. Two essential things in the visual arts were missing in AI’s indiscriminate data scrapes: specificity and intentionality. Most importantly, I had no sense of how to deal with the edge, where the two shapes meet. How the painter resolves the edges will go a long way in determining his or her style. Is Edge brassy and expressive, or tight and restrained? In digital images, the edges are often approximated as the machine tries to guess the next move, but in any case, the pixels are different. It lacks intentionality and no inflections.
The aphorisms from the early “trash, trash output” of computing remained relevant. To get anywhere, you need to send your machine to the art school. Through venture advisor Tom Cohen, I was introduced to a talented young engineer named Grant Davis. Together we began designing the curriculum.
First, he trained the machine with a handful of paintings by Giorgio de Chilico, Edward Hopper and Arthur Dub. This is three 20th century masters, respectively, who illustrate the drama of perspective, the ability to convey a sense of volume, and the lyrical use of black colours. We have added these celebrities a rigorously edited selection of paintings from the 1980s and 90s. The machine’s visual IQ rose several points, but the edges were still weak.
We dialed an art history lesson and retrained the machine with my 2001 series of paintings. It is a luxurious and fragmented composition based on the background of 19th century opera. This romantic, idyllic image – the man and woman sitting on the bank were painted in harmony of colours with a lake in the center and the mountains in the distance that stunningly dazzling. As a final step, we added my paintings, which are rendered with brush strokes, with a form that is definitive and expressive.
When tweaking the model, Davis is independently developing a software program for visual artists called “Wands,” eliminating the need for language prompts. This works like a joystick. This is a lever that moves along the continuum from “similar” to “different.” We gave the machine a “pastoral” painting in intricate shades, placing the dial at the middle point between the extremes and holding our breath.
The machine has now proven to be a brilliant student. My composition melted and melted, placed in a blender, then passed through the sieve. The results were clearly derived from my work, with many familiar elements, but the reconstruction was different from what I had seen before. The machine could dismiss the rules of continuity, scale, anatomy and pictorial logic, as it didn’t know what it was doing. (The feeling of a machine that didn’t know the violation was important. Artificial intelligence is even more important.)
The image was still only on the screen pixels. I selected the most promising scenes and printed them on canvas. These prints form the background I drew like a background, layered with collages of archaic elements that appear to emerge from the swirls of the background. Male and female torso; gesturing hands; unstable, balanced brown cups are painted using large brushes of very saturated colors. The configuration is dynamic. Everything sails through the delicate Topsitabe Space. Painting speaks to the gravity of art. However, they are also linked to mysterious “future beings.” They are the Palim plague from dialogue with my previous works, like “One Duet.”
AI is a powerful tool, but still, it’s just a tool. At the moment, he is not an author. It works in the way it was taught and what it was taught. The “thinking” part of the machine, the algorithm, breaks down the photos, recombines the parts of the components, scrambling them in spatial orientation, but appears to have a drive to store and amplify the emotional subtext. What can better express this moment?
David Searle’s “Some Versions of Pastoral” will be available at ropac.net until June 8th at Thadeus Ropac, London. And his work will be in Freeze New York from May 7th to 11th.
Find out first about our latest stories – Follow FT weekends on Instagram and X and sign up to receive your FT weekend newsletter every Saturday morning
