‍

Meandering with AI

Consumer-grade text-to-image tools make it easy to skip the arduous work of artistic practice and jump straight to a polished result. Chuck Anderson’s Imagined Wreckage shows how these tools instead can be layered into a bespoke and meandering artistic practice.
‍
Text-to-image generators seem designed to kill artistic practice. The relatively recent rise of Dall-E, Midjourney and similar tools has made generating images exceedingly easy, divorcing image-making from the artistic practice that typically precedes it. The paradox of widely available consumer AI image generators is that the easier they make it to create an image, the harder they make it to forge a way of being in the world that results in the creation of worthwhile images.‍

‍The best artworks are the ones that carry the story of their making—often circuitous and indulgent—along with the story of their maker: contradictory, inefficient and carrying a unique point of view. There are many types of AI and there is a core group of artists—Mario Klingemann, Holly Herndon and Mat Dryhurst, Memo Akten, just to name a few—who have used and built various AI systems well before consumer text-to-image models became widely available. For these artists, AI experimentation is part of a rich artistic practice. ‍This essay considers something else:
‍

If broadly available, consumer-grade text-to-image tools present the temptation to leapfrog right past the arduous aspects of artistic practice straight to a pretty result, how can artists integrate these tools into a compelling practice anyway?

‍
‍Brutally efficient, text-to-image follows direct orders and has no point of view, instead assembling Frankenstein-like images from vast heaps of indiscriminate training data. Many have written off the possibility of these tools creating compelling artworks. The filmmaker and theorist Hito Steyerl, for example, has called the output of text-to-image generators “mean images” that “replace likenesses with likelinesses,” in a double entendre that refers to both the averaging effect of these tools as well as their apparent cruelty and indifference.‍

‍But good artists find a way to work within contradictions. Text-to-image generators may be engineered to be antithetical to artistic practice but that just makes the challenge of integrating them into artistic practice that much more intriguing. Very few artists have managed this in the short history of widely available text-to-image generators but one who has pulled it off is Chuck Anderson.
‍

‍‍
‍
‍Anderson’s ongoing series Imagined Wreckage folds AI into the artist’s decades-long practice of layering, pushing and breaking digital image-making technologies.

‍Anderson is an artist, designer and creative director based in Chicago. Operating as a one-man studio under the name No Pattern, he has created personal projects and client work for dozens of brands since 2004. Working primarily with digital images, his work is dense, colorful and maximalist.‍

‍Whether he’s designing an album cover for Lupe Fiasco, apparel for Nike and Michael Jordan or the desktop for Microsoft Windows 7—Anderson has a tendency to stretch digital image-making tools beyond their capabilities. Personal art projects illustrate this tendency even more clearly. The 2023 project Infinite Pressure features abstract digital compositions that layer polygons and gradients with such manic intensity that they arrive at a kind of twenty-first-century Abstract Expressionism, like skillfully splattered digital paint built into impossibly dense compositions.‍

‍The Imagined Wreckage images have photographic qualities but they look more likely to have been captured by a scanner than a camera. Circuit boards, engine components, wires and miscellaneous plastic parts seem to be vacuum-sealed in iridescent plastic wrap. The images allude to the horror and fascination of looking at the innards of your laptop after it falls off the table and breaks in half. They evoke a kind of cool, technological gore.‍

‍Through all of Anderson’s work, two things remain constant: color and process. Every project has a color palette that manages deliberate expansiveness. The best colorists employ this trick, creating a sense that a set of images plays within the full spectrum of color. At the same time, they limit the range in a way that’s difficult to pin down but gives everything a chromatic cohesion.
‍

‍‍

‍Anderson’s other through line is process. In both personal and client work, his images clearly appear as the end result of a tremendous amount of experimentation and play. We’ve all had the experience of noodling around with various tools in Photoshop, making gradients, polygons and lines, applying filters and generally making a digital mess. Anderson’s practice is what happens when you take this kind of play and do it very intentionally for tens of thousands of hours over twenty years.
‍‍

It’s difficult to build a distinct practice with digital tools because they’re designed to simplify and speed up the image-making process, when what’s needed for artists is to relish complexity and slowness, tinkering with a material until it finally matches their vision.

Anderson has forged a practice in spite of the slickness of digital tools, not because of it.‍

‍The process of making Imagined Wreckage involves many tools and lots of iteration. Anderson typically uses Krea, Midjourney, Magnific and Luma, with final editing always happening in Photoshop. He uses custom training sets, including a careful selection of photos of found objects and textures, as well as his own past work.‍

‍The process involves generating a tremendous number of images, then sifting through them for what he calls “the needles in the haystack.” He estimates that less than 0.01% of images generated ever go on to final editing in Photoshop and eventual public display. He explains the process as a blessing and a curse.
‍‍

There’s a danger in playing with tools that can make an endless supply of images that are interesting but so few that are worthwhile or distinct.

In an email to me, Anderson described his process, noting that it’s “easy to fall into the trap of combining random things for the sake of combining random things but not actually saying anything.”‍

‍What elevates Imagined Wreckage above most artworks incorporating text-to-image is that it’s an extension of Anderson’s established practice—one built on taking tools designed to turn image-making into a frictionless process and reintroducing new kinds of friction. Good art is often the result of accumulating unnecessary processes. But text-to-image exists to eliminate processes—or at least lock them inside an algorithmic black box.‍
‍‍

‍

‍The problem remains that artworks are a sum of the decisions that went into their making. A good artistic practice means making numerous decisions in a way that only that particular artist would. Many of those decisions will be inefficient, costly and contradictory.‍
‍‍

Practice eschews efficiency; it’s meandering; it insists on taking the scenic route.

‍Text-to-image generators are built to deliver a product without a practice. This is particularly evident in the work of artists like Anderson, whose maximalist compositions keep a record of his circuitous path of additions and edits. But it’s also true for artists delivering a much more minimal and spare final product. The journey toward what can be cut and what must remain is rarely a straight line.‍

‍Chuck Anderson’s practice is distinct from what has become known as “prompt engineering.” Generative AI enthusiasts focusing on the robust and meticulous construction of text prompts are akin to mastering a video game—an impressive feat but the player always remains less compelling than the game or tool being mastered. Anderson avoids this fate with Imagined Wreckage because he layers AI into a practice defined by a stack of image-making tools that is not optimized for efficiency but instead prioritizes experimentation, play and, finally, a lot of smart editing.‍

‍Practice is a way of building a point of view; it’s a way of accumulating a unique set of personal inefficiencies. What’s interesting about an artist is how they work and what this indicates about how they—and we—can navigate the world. The particular tools used are always secondary. Practice is a way of making things harder and more complicated than they need to be; it’s about prioritizing ritual over utility. This is why making good art with AI image generators is so hard. It’s also what makes it worthwhile to find a way to do it anyway.



----



‍‍Kevin Buist is an independent critic, filmmaker and design strategist who lives in Grand Rapids, Michigan.