“The keyword of the new [20th] century was modernity. Modernity meant believing in technology and not craft, human perfectibility, not original sin. And above all in a ceaseless consumption of things and the images of things.”
—Robert Hughes, The Shock of the New,
Episode One: “Mechanical Paradise”
In recent years, the tech industry’s tendency towards promotional mania has accelerated at a rate of speed that might have surprised even Gilles Deleuze and Félix Guattari. From crypto to NFTs to the metaverse, there have been successive waves of supposedly bold futures ushered on and off stage as if through a revolving door.
Enter AI art, stage left. From The New York Times to the Los Angeles Times, to The Atlantic and many more besides, the topic of AI art is nearly impossible to avoid in media. Unlike previous waves however, AI art shows signs that it will have a more lasting and profound impact.
In preparation for writing this essay, I did a bit of searching on Twitter (that carousel, spinning ever faster, in a burning house) for the phrase “AI Art.” Although, as a technologist and critic of the tech industry generally—and its so-called “AI” manifestation specifically—I am familiar with the mechanics of systems such as Dall-E, and products such as Microsoft Designer, I was curious to see how the application of algorithmic systems to the creation of art was being discussed and presented.
Judging from the results (and here you may assume the usual caveat about my method being “unscientific”), there are two opposed camps, shouting past each other, with clashing incentives: the artists who, rightfully, understanding that systems such as Stable Diffusion are built upon using their work without payment or attribution, are raising alarms, and the AI enthusiasts, companies, and camp followers declaring that the sci-fi future of their dreams has, at long last, arrived.
Popular
"swipe left below to view more authors"Swipe →
My own sympathies are entirely with the artists, whom I see as fellow workers and targets of exploitation. Besides, it’s difficult to have too much fellow feeling for whoever is behind efforts such as “Dream Girl AI” or “taylor swift ai art” producing an unending stream of synthetic images, each different in setting and yet identical in dull affect.
But if we can leave questions of aesthetics—and, for the moment, exploitation (or, as many artists unequivocally describe it, theft)—aside, the questions that next come to mind are: Why is this happening? And why is it happening now? Why, to put it in material terms, was Stability AI, (for example)—the organization behind Stable Diffusion—recently able to raise $ 101 million in funding? Who is looking to benefit from so-called AI art—and how?
The answer can be found in Fordism—the application of industrial methods of production and consumption named after Henry Ford—to the creation of art or, more precisely, images. This has been tried in the past: Andy Warhol’s Factory, nearly anything by Jeff Koons, and the mass reproduction of Keith Haring’s “Dancing Man” are three celebrated examples. But with the creation of systems for ingesting existing art as patterns, and then, using those patterns as the raw material for generating “new” images via text prompt, this is being industrialized on a scale Warhol’s Marilyn only hinted at.
Walter Benjamin, in his 1935 essay “The Work of Art in the Age of Mechanical Reproduction,” described the 20th-century social and technological conditions that, in time, would inform the work of these later artists as they (very deliberately) embraced the age of mechanized reproduction:
In principle a work of art has always been reproducible. Man-made artifacts could always be imitated by men. Replicas were made by pupils in practice of their craft, by masters for diffusing their works, and, finally, by third parties in the pursuit of gain. Mechanical reproduction of a work of art, however, represents something new.
Later in his essay, Benjamin defines the “newness” of mechanical reproduction as being its impact on a key aspect of art across time—“authenticity”:
The authenticity of a thing is the essence of all that is transmissible from its beginning, ranging from its substantive duration to its testimony to the history which it has experienced. Since the historical testimony rests on the authenticity, the former, too, is jeopardized by reproduction when substantive duration ceases to matter. And what is really jeopardized when the historical testimony is affected is the authority of the object.
By applying computation on a massive scale, in an updated application of Fordist methods, to the creation of art, Silicon Valley completes this break with authenticity. There is, for these systems, no artist (besides those whose body of work form the foundational data sets); there are only images.
By using the term “Fordism” to describe AI art, I’m not merely employing a (hopefully) clever metaphor, but also recentering the role of profit incentives—always at the heart of capitalist activity—in this still unfolding story. I’m thinking of the art market and the fact that, even at its most exploitative, prior to the creation of Fordist image production via algorithm, there was still the requirement to involve and pay an actual human artist. If you wanted a reproduction of a Caravaggio fresco on the wall of your McMansion, you had no choice but to find a talented artist to paint it for you. Gallery owners needed a stream of work from both new and established artists to attract sales. Though deeply flawed in ways artists are very familiar with, the art market still demands what could be called bespoke creative labor in much the same way auto manufacturing, prior to Ford’s Taylorist innovations of the early 20th century, required skilled artisans who shaped materials into moving machines.
Whether high art or low, kitsch or avant-garde, the production of art remained safe at the point of creation (if not reproduction) from being consumed by automation—and protected, therefore, from the tech industry’s well-worn rentier tactic of imposing itself between us and the things we need. (Not to mention the things we merely want, from the ability to create and record texts to the music we wish to listen to.) The application of machine learning methods such as diffusion (which, in brief, uses a process of iterative processing—diffusion—matching output to text input) to image creation threatens to disrupt (in a real, and not simply a marketing sense) the relationship between artist and creation. Not by using machines that can match, let alone exceed, human creativity but by narrowing the definition of art to fit within the confines of what image synthesis systems can do.
To get a better understanding of what I mean by the term “Fordist image production,” consider the system, DALL-E, produced by OpenAI. Like other such systems, DALL-E (and it’s still-in-development successor, DALL-E 2) can produce visuals based on text prompts. You can, for example, type as input, “a dog, playing with a ball in the style of Picasso” and the system, using its combined corpuses of text and images, will output a synthetic result that more or less meets the criteria (more information about how DALL-E works is here).
As part of the DALL-E 2 development program called “Extending Creativity” (the use of “extending” here is intriguing, suggesting something in need of amplification or modification—much the way binoculars extend the range of vision), OpenAI stated that it enlisted the help of “more than 3,000 artists from more than 118 countries [who] have incorporated DALL·E into their creative workflows.” One effort was OpenAI”s collaboration with Austrian artist, writer, and curator Stefan Kutzenberger which OpenAI describes as
a project conceived by Austrian artist Stefan Kutzenberger and Clara Blume, Head of the Open Austria Art + Tech Lab in San Francisco, DALL·E was used to bring the poetry of revolutionary painter Egon Schiele into the visual world. Schiele died at 28, but Kutzenberger—a curator at the Leopold Museum in Vienna, which houses the world”s largest collection of Schiele’s works—believes that DALL·E gives the world a glimpse of what Schiele’s later work might have been like if he had had a chance to keep painting. The DALL·E works will be exhibited alongside Schiele’s collection in the Leopold Museum in the coming months.
The project uses DALL-E to create new works in the style of an artist who died over a hundred years ago. One of Kutzenberger’s prompts was “A painting of tall trees walking along a road, with chirping and trembling birds in front of a white sky in them in the style of Austrian expressionist Egon Schiele.” Time has, sadly but quite naturally, deprived us of Schiele himself. But through the use of DALL-E as an image production assembly line, the relationship between artist and image is deconstructed and used as raw material—like parts in a Ford assembly plan—for the manufacture of supposedly new images. Yet the resulting images, forever dependent on the past—on Schiele’s existing work—are actually old, trapping the viewer in a time loop of kitsch, presented as brilliantly new.
In 1928, construction of the Ford Motor Company’s sprawling River Rogue facility was completed. As an integrated manufacturing complex, River Rogue combined the ingestion and processing of raw materials with the production, via assembly line, of automobiles. So-called “AI art” systems such as DALL-E are also integrated manufacturing complexes (data centers instead of auto plants) ingesting our text and images as raw material to create a method for removing the artist from the art. And unlike River Rouge, whose workers’ efforts to organize in 1936, under the slogan “Unionism, Not Fordism,” were met with brutal beatings—a public relations disaster for the company that helped lead, within just a few years, to the recognition of the United Auto Workers—DALL-E’s employers never have to worry about strikes. Which perhaps is also the point.
