What Is AI Doing To Art?

AI-generated artwork is prompting hard questions about human creativity. The history of the photograph shows why the terms of the debate are wrong.

This image was inspired by the historic 1878 photographs by Eadweard Muybridge capturing a horse in motion and by AI's ability to explore and visualize seemingly infinite creative possibilities. It was created using several AI tools, including DALL-E 2 for image generation from text, RunwayML for outpainting and Topaz Gigapixel for upscaling. The AI-generated images were cycled through themselves (the horses briefly became cows at one point!) and then manually layered on top of each other. (Kristian Glenn for Noema Magazine)

Lois Rosson is a historian of science and technology based in Los Angeles. She is currently writing a book about images of outer space and their legibility.

In 1835, William Henry Fox Talbot finally succeeded in producing a crude photograph of his country estate. He triumphantly declared that his was the first house ever known to have drawn its own picture. Fox Talbot described the calotype, his contribution to the photomechanical process, as an eradication of human intervention. In Talbot’s description, the photogenic drawing was formed “by the mere action of Light upon sensitive paper.” Photography offered nature a “pencil” with which to render herself via optical and chemical means alone. 

Fox Talbot’s self-drawing house is a useful reminder that the development of the photograph is an automation story. By the mid-nineteenth century, rendering a detailed image no longer needed to be outsourced to a draftsman because the process could be completed instantly with a camera. Proponents of the technology emphasized that not only was photography more precise than the human hand — it was faster and cheaper. 

The elimination of human fallibility was one of photography’s biggest selling points, but this prompted passionate debates about the new medium’s implications for visual culture. Could images made largely by a machine be considered art? If so, where did human creativity fit in this process? 

This image was created by the text-to-image AI tool DALLE-2 when it was given the prompt: “Calotype of nineteenth century English estate.”

The answer, negotiated after a century of messy non-consensus, established photography as a form of objective mechanical documentation and creative human expression simultaneously. It embodied two conflicting approaches. One determined that authorship of an image existed largely on the ends — in the framing of an image and its development — but the camera filled in the rest. The other maintained that photography was influenced by a human operator at every step, representing a creative process akin to drawing with light. While the second argument helped photography secure copyright protections, it was the first that made the medium compatible with an industrializing Europe and drove its commercial proliferation.

As the twenty-first century becomes increasingly automated, attempts to pinpoint where human agency exists in a technologically mediated process grow more frantic. Images generated with artificial intelligence by companies like OpenAI and Stablilty.ai are spurring questions remarkably like those that emerged with the advent of the photograph. By typing a sentence into the equivalent of Google search, users can generate “new” images compiled from images scraped across the internet, some mysteriously and others dubiously. The result has been a flood of AI-generated images in places previously exclusive to human authors. Painting competitions, commercial graphic design and the genre of portraiture have all since collided with the technology in troubling ways. 

The fine arts were thought to be a final hold-out of mortal inventiveness, but the surprising quality of AI-generated images is prompting deeper questions about the nature of human creativity. How can you automate a process that is itself indicative of human expression? A lawsuit filed early this year against Stability AI hinges on whether originality is exclusive to human creators. Adjusting copyright law to address this issue will likely have a tremendous impact on creative economies. If the history of the photograph tells us anything, it’s that the debate won’t be settled quickly, straightforwardly or by the institutions we typically associate with cultural gatekeeping. The process will, however, tell us a lot about the cultural conditions that help us make sense of emergent technologies. 

“How can you automate a process that is itself indicative of human expression?”

Photography As Both Art & Document

Trust in a camera’s ability to produce objective pictures was built up over the nineteenth century by emphasizing the technology’s externality to human subjectivity. Fox Talbot celebrated his estate’s ability to draw itself because it removed human interpretation. His French contemporaries, Louis Daguerre and Joseph Nicéphore Niépce — with equally solid claims to the invention of the medium — insisted on the elimination of the draftsman as a critical step in the “fixing” of nature’s visual expressions.

An existing literature in the history of science traces how the development of the photograph helped negotiate “objectivity” as a neutral conceptual category. In nineteenth-century science, photographs of biological and anatomical samples displaced illustrations as a more trustworthy form of visualization, even if they were less clear. Photographs could be fuzzier and more visually ambiguous than hand-made illustrations, but their mechanical nature helped them circulate as comparatively more reliable. 

Framing photography as distinctly outside the realm of human fallibility was one of the technology’s biggest selling points. The maintenance of this idea, however, prompted difficult questions about attribution. If a photograph was truly an automated form of draftsmanship, could photographers be thought of as artists? The production of a photograph certainly couldn’t happen without human operators, but could they be considered a creator more than someone using a machine in a factory? 

As with AI-generated artworks, the “automation” of draftsmanship prompted new ways of thinking about authorship. One early contender for photography’s true author was light itself, acting autonomously on behalf of the sun. An early form of the technology invented by Niépce in 1826 required a full day of exposure to the sun and was thus coined “heliography,” or “sun writing.” English art critic Elizabeth Eastlake, describing the emergent genre in the 1830s, referred to photographic tools collectively as a type of “solar pencil,” using light to draw upon the camera’s lens. This was decried — but not contradicted — during the Salon of 1859, when the interminably grumpy French critic Charles Baudelaire skewered photography as a form of fanatical sun-worship.

Baudelaire’s 1859 rebuke indicates that the presumed objectivity of a photograph was not yet recognized as a universal value. He mocked French aristocrats who believed that true art was an exact replication of nature, describing Daguerre as the Messiah of a “revengeful God.” “And now the faithful says to himself: ‘Since photography gives us every guarantee of exactitude that we could desire (they really believe that, the mad fools!), then photography and Art are the same thing.’” Baudelaire contested the categorization of photography as an art form but also the claim that it functioned as a perfect transcription of reality.

Numerous types of photographic technologies emerged in the nineteenth century, each with their own technical idiosyncrasies. Daguerreotypes looked different from calotypes, and all were fuzzy compared to photographs today. Which of these could be said to best represent reality? Individual cameras could consistently replicate certain types of visual information, but this was not yet true of photography as a genre. 

“If a photograph was truly an automated form of draftsmanship, could photographers be thought of as artists?”

Despite the protests of aristocratic art critics, commercial photography cemented itself as a market in France over the course of the nineteenth century. Arguments about its lack of creative merit gradually faded in the face of photography’s mounting profitability. Because of its technical novelty, however, it was unclear whether photography involved sufficient human creativity to qualify for protections under French copyright law. Erasing human influence from the photographic process was good for underscoring arguments about objectivity, but it complicated commercial viability. Ownership would need to be determined if photographs were to circulate as a new form of property. Was the true author of a photograph the camera or its human operator? 

In the mid-nineteenth century, answers to this question were hastened by the material stakes. The first legal designation of photography as a creative art form occurred in April of 1862, when the French photographers Mayer et Pierson successfully prevented the sale of retouched and altered celebrity portraits taken by their studio. Because they had taken the original photographs, Mayer et Pierson argued that the portraits were theirs alone to monetize. The court’s ruling represented a hard-won success. Earlier that year, it had rejected their suit on the grounds that photography was functionally automated — the medium was little more than a chemical process for fixing the image of external objects using a machine. 

What changed this thinking? When Mayer et Pierson appealed the decision in April of 1862, they used an argument that reintroduced human agency into the photographic process. By reframing photographs as les dessins photographiques — or photographic drawings, the plaintiffs successfully established that the development of photographs in a darkroom was part of an operator’s creative process. In addition to setting up a shot, the photographer needed to coax the image from the camera’s film in a process resembling the creative output of drawing. The camera was a pencil capable of drawing with light and photosensitive surfaces, but held and directed by a human author. Copyright protections helped photography commercialize in nineteenth-century France, but rather than clarifying what the photographic process was doing, this development codified it as both art and documentation simultaneously. 

If the court had ruled that photographs weren’t protected because the camera performed the bulk of the work, then an entirely new set of problems would have emerged with respect to the creative process. What of the painter who employed a team of apprentices in a large studio? An engraver that sold hand-made etchings based on famous paintings? Could a well-dressed portrait sitter exert some claim over the artistic process once the work was completed? Might the gardener of a meticulously maintained landscape declare authorship over a watercolorist’s portrayals? Establishing photography’s dual function as both artwork and document may not have been philosophically straightforward, but it staved off a surge of harder questions. 

Over the nineteenth-century, most western art markets established some form of copyright protection for photography, ceding that the medium involved substantive creative human input. In the popular imagination, however, photographers were still viewed largely as technicians. By 1899, Alfred Stieglitz lamented the view that “after the selection of the subjects, the posing, lighting, exposure and development, every succeeding step … require[ed] little or no thought.” Human intervention in the photographic process still appeared to happen only on the ends — in setup and then development — instead of continuously throughout the image-making process. Photography won its legal designation as an art form in the nineteenth century and spent the bulk of the twentieth convincing skeptical museum curators why.

“Arguments about its lack of creative merit gradually faded in the face of photography’s mounting profitability.”

Creativity & Commercialization In The Age Of AI

The success of photography as a medium hinged largely on early descriptions that appealed to nineteenth-century sensibilities. As European economies looked toward an industrialized future, the elevation of the photograph’s mechanical trustworthiness made it an ideologically compatible form of visual output. Separating human from camera was a necessary part of preserving the myth of the camera as an impartial form of vision. To incorporate photography into an economic landscape of creativity, however, human agency needed to ascribe to all parts of the process.

Consciously or not, proponents of AI-generated images stamp the tool with rhetoric that mirrors the democratic aspirations of the twenty-first century. Stability AI, now one of the subjects of a lawsuit filed by artists whose work appeared in their training data set, paid German nonprofit LAION to compile an open-source database with billions of images. LAION anticipated accusations of copyright infringement by invoking the spirit of democracy in descriptions of its work. The sparse amounts of information on LAION’s website emphasize its service to the public good. The “100% non-profit” and “100% free” organization is committed to the “liberat[ion]” of machine learning research. Their work facilitates “open public education,” and its recycling of existing data sets is described as “environment-friendly.”

Stability AI took a similar tack, billing itself as “AI by the people, for the people,” despite turning Stable Diffusion, their text-to-image model, into a profitable asset. That the program is easy to use is another selling point. Would-be digital artists no longer need to use expensive specialized software to produce visually interesting material. 

“[They] maintain that AI art simply automates the most time-consuming parts of drawing and painting, freeing up human cognition for higher-order creativity.”

The lawsuit filed against Stable Diffusion describes the defendants’ egalitarian language as a ploy to exploit the legal gray area surrounding data sets scraped from the internet. In an interview cited by the plaintiffs, Midjourney founder Tim Holz said that to his knowledge,“every single large AI model is basically trained on stuff that’s on the internet. And that’s okay, right now. There are no laws specifically about that.” Meanwhile, communities of digital artists and their supporters claim that the reason AI-generated images are compelling at all is because they were trained with data sets that contained copyrighted material. They reject the claim that AI-generated art produces anything original and suggest it instead be thought of as a form of “twenty-first century collage.”

Because it is fast, cheap and easy to use, however, AI art continues to attract a broad user base. Lensa, an AI app that generates custom portraits for users, generated $8.2 million in the five-day period following the release of its “magic avatars” feature. The DALL-E 2 subreddit, an online forum dedicated to mastering OpenAI’s image generation platform, often echoes photography’s early attempts to be understood as a creative process. Proponents describe the process of summoning an image from the data set as “prompt engineering,” emphasizing the necessity of a human intervening by giving the AI certain prompts.

Others looking to elevate AI art’s status alongside other forms of digital art are opting for an even loftier rebrand: “synthography.” This categorization suggests a process more complex than the mechanical operation of a picture-making tool, invoking the active synthesis of disparate aesthetic elements. Like Fox Talbot and his contemporaries in the nineteenth century, “synthographers” maintain that AI art simply automates the most time-consuming parts of drawing and painting, freeing up human cognition for higher-order creativity.

Contemporary critics claim that prompt engineering and synthography aren’t emergent professions but euphemisms necessary to equate AI-generated artwork with the work of human artists. As with the development of photography as a medium, today’s debates about AI often overlook how conceptions of human creativity are themselves shaped by commercialization and labor.  

“As with photography, today’s debates about AI often overlook how conceptions of human creativity are themselves shaped by commercialization and labor.”

Economic Precarity & The Specter Of Automation

Viewing AI art as part of a broader pictorial history can temper fears that it is a prelude to a dystopian future. The problem with debates around AI-generated images that demonize the tool is that the displacement of human-made art doesn’t have to be an inevitability. Markets can be adjusted to mitigate unemployment in changing economic landscapes. As legal scholar Ewan McGaughey points out, 42% of English workers were redundant after WWII — and yet the U.K. managed to maintain full employment. In contemporary debates about automation, the real drivers of precarity often have more to do with the erosion of labor protections over the twentieth century. In the U.S., automation is an easy scapegoat for the gutting of worker protections. We look back on the development of the photograph as a technological transformation, not as one characterized by major waves of worker displacement. 

In the case of photography, we created a myth that cameras automated image-making in ways that were free from human interpretation. To make the photographic process legible to the forces of commercialization, however, we reframed it as a form of drawing, where human agents made marks using particles of light on a photosensitive surface.

An understanding of the technology as one that separates human from machine into distinct categories leaves little room for the messier ways we often fit together with our tools. AI-generated images will have a big impact on copyright law, but the cultural backlash against the “computers making art” overlooks the ways computation has already been incorporated into the arts. 

“Viewing AI art as part of a broader pictorial history can temper fears that it is a prelude to a dystopian future.”

When copyright was finally extended to photography in the mid-nineteenth century, it was partially to avoid opening other forms of artistic tools to scrutiny. Are artists using computer software on iPads to make seemingly hand-painted images engaged in a less creative process than those who produce the image by hand? We can certainly judge one as more meritorious than the other but claiming that one is more original is harder to defend.  

Art is much more than what is captured digitally on the internet, but the internet is an indispensable tool for artists attempting to earn a living. The proliferation of AI-generated images in online environments won’t eradicate human art wholesale, but it does represent a reshuffling of the market incentives that help creative economies flourish. Like the college essay, another genre of human creativity threatened by AI usurpation, creative “products” might become more about process than about art as a commodity. 

For historians of visual culture, the debate that AI-generated artwork has already sparked is as indicative of our current political moment as artistic movements of the past. Private tech companies that shape our political and economic landscape frame large open-source datasets as “democratic,” while the artists whose work is integrated advocate for greater property protections. In a moment when “truth” is a concept fraught with political partisanship, we can no longer seek solace in the apparent reality of a photograph. AI-image generators are perfectly capable of emulating the look of traditional photography, forcing us to confront the very human ways in which images have always been made.