Human Art Already Has So Much In Common With AI – WIRED

Despite being human-made, human-taught, and human-promoted, it’s easy to criticize AI for being fundamentally inhuman. To claim that AI models like ChatGPT and DALL-E will replace art created by people is to ignore both the ineffable qualities of the human touch and the critical flaws of these models—or so say the artists and writers.

They’re right that AI isn’t quite at the stage of completely replacing human creativity—it is biased and inaccurate, good at bullshitting without substance. It offers a simulacrum of desired output but cannot be trusted on its own. But to home in on AI’s failures underestimates the will of their developers to overcome them. And it overlooks the fact that these algorithms are able to mimic the creative process precisely because human creativity is, in many respects, just as algorithmic as the AI models that seek to outperform it. 

Writers and artists hone their craft by imitating, iterating, and integrating the work of others—further entrenching the archetypes and rhythms that humans and machines alike recognize as the trappings of art. Like AI models, we “predict” the best next word, brushstroke, camera angle, or musical note based on previously encountered work. We weigh our choices with accreted values and biases. We course-correct using feedback on proximity to desired effect. And with every iteration of this process, our ability to more precisely execute creative prompts improves. The presumed appeal of AI models is that they can do this at scale and on demand, drawing from a massive corpus of samples and feedback.

Whether an AI model will one day create a masterpiece, achieve true indistinguishability, or become a literal deus ex machina is a moot point. At face value, both humans and machines create some form of art.

But these questions over capability distract from the issues that technologies raise with every advance: If the end results are equivalent, do we value the object over the person making it? Or in this case, do we value the artist in all their messy humanity? And what does the desire to strip away that humanity reveal about the perceived purpose of art and creation being sold to us?

Certain online communities are already exposing the duplicative qualities of art by using pattern-finding as a creative medium—namely, via “web weaving.” Also known as comparatives, web weaving is a genre of Tumblr post that juxtaposes excerpts and images around a central theme. Interspersed among Kafka passages, Richard Siken verses, Renaissance paintings, street photography, and Mitski lyrics are screenshots of TikToks, tweets, and film stills. These collections of digital ephemera often cohere around motifs of yearning, alienation, and intimacy—pervasive moods of online life.

One post by Tumblr user @ohevoyev is dedicated to the Minotaur of Greek myth, joining Borges’ short story with paintings, stanzas, and a screenshot of a conversation encapsulating the through line: “the cruelty of being a cursed child” whose parents “use the curse to lose love for you.” Another has multiple users contributing multimedia excerpts in an extended meditation on homesickness. Web weaving is the social media user’s collage: The canvas of the post holds together artful arrangements of online curios and gems, finding order across a sea of content.

Another form of web weaving—fanvids, or video edits—splices together media clips to highlight thematic through lines. One popular fanvid by Twitter user @loveforcaptnswn melds sequences from 31 shows and 34 films that aired in 2022—characters walk through battlefields and gaze across burning wastelands literal and figurative, hold one another in violent and tender embrace. Each visual motif seamlessly transitions from one story to the next, interchangeable characters and immaterial backdrops swapping places like a glitch in the matrix.

As web weaving shows, art that strikes a chord doesn’t necessarily do something new so much as it reconfigures the enduring in novel contexts. Repetition is a cornerstone of art and writing—reduced to its most elemental form, all art asks variations of the same fundamental questions.

It’s why works can be organized by genre and subgenre; academics can propose grand unified theories of literature; folklorists can systematically classify thousands of recurring motifs in myths, legends, and fables around the world; and friends and algorithms can curate playlists of songs and stories just for you. And when repetition tends toward the formulaic, it’s also why AI can be used by Hollywood to “predict” a film’s performance based on its screenplay, by writers to churn out novels, and by researchers to manifest scientific papers.

But repetition isn’t automatically inferior for being predictable. Predictability allows artists to create something different via subversion. What if the Minotaur were not the monster but the victim? And what if Theseus came bearing not a sword of just retribution but of tragic deliverance? In taking the same trope a few steps to the left, inverting or casting it in light or shadow, the well-worn becomes scaffolding for a previously unheard perspective or yet unimagined possibility.

But focusing exclusively on the similarities between works loses sight of their differences, erasing the context that grants each their unique purpose and function. And it undermines the gravity of making a deliberate choice to break the pattern, leaving it all up to chance for the sake of expediency. We’ve all heard this story before, and we could stand to hear it again. The hunger for a version that honors individual needs with intentionality won’t be sated with infinite variants you can’t truly call your own.

What if repetition and predictability themselves give art its meaning and significance? The algorithmic qualities of creativity can be an asset rather than a redundancy. And no other community is as proudly and enthusiastically well versed in drawing parallels and iterating for the billionth time on the same tropes as fandom.

An abundance of fan works circle around the same plot devices, narrative alternatives, and character archetypes. Fans have set up characters from various media properties in over 33,500 coffee shop meet-cutes to date. And some find the hundredth read about a character waking up to a loved one waiting beside their hospital bed no less emotionally potent than the first.

The predictable and repetitive nature of these fan works don’t detract from their appeal to other fans—in fact, it’s through these very qualities that they become meaningful. Recursion is what sustains fan communities. In passing down the same hallowed tropes through art created for the pleasure of creation and connection, a shared love for a story becomes a common language spoken via iteration.

In conversation with cultural critic Hanif Abdurraqib, the musician L’Rain describes how she dealt with her discomfort over her voice by looping it, finding transformation through repetition. In response, Abdurraqib noted that repeatedly listening to the same sound turns it into noise, as “repetition allows the clear to become joyously indecipherable.” Interpreted in this way, repetition doesn’t render something meaningless but distills it into pure essence—smoothing the stutter of the individual until all that remains is the medium itself. Machine learning capitalizes on the fact that a trope repeated ad infinitum will so congeal in meaning that authorial source is made superfluous.

That doesn’t mean the author’s presence is no longer felt. A common critique of literary and media analysis is that “the curtains are blue” because they just happened to be blue—the author did not intentionally ascribe significance to it. But as Tumblr user @seravph noted, the curtains are blue not as a manufactured symbol for sadness, but because the author’s mother died in a room with blue curtains, and every story they’ve written since then bears that mark. In other words, it’s not a contrivance, but a haunting—an image bleeding with memory. Biographical knowledge isn’t necessarily required to perceive its significance; it’s enough to treat the work as something that bears the weight and warmth of another’s life. Every trope is made unique when run through the sieve of the individual.

If AI embodies l’art pour l’art, generating products disassociated from social function and cultural history, then fandom at its best is a celebration of the people behind the art—where the labor of creation is recognized and revered. An AI model could replicate @loveforcaptnswn’s 2022 fanvid, showcasing an even vaster panoply of media. But the experience of watching it is colored by the fact that it was hand-woven by one person as an expression of love and gratitude for her friends. 

Repetition and predictability are not what devalues art. Rather, what devalues art is the negation of the artist and community that imbue a work with significance beyond its mere existence. The value of art is what we give it—no object has inherent worth. Indistinguishable or not, to claim equivalency between AI- and human-made works is to reject what makes them worth anything more than their exchange value. And to privilege the object over the person puts to question whether these tools are as humanity-serving as their designers claim them to be.

In its repetitive qualities, art lends itself to being conceptualized as an assembly line of interchangeable parts that can be cannibalized at will. So it is that the process of creating art is often described in terms of theft and imitation—“good artists copy, great artists steal,” or so the supposed Picasso quote goes. AI users justify their means of production by claiming that all artwork is derivative: If nothing is original, why couldn’t they draw inspiration from another’s work? Some argue that AI art is ethical and morally permissible because it isn’t considered theft under current legal definitions.

But drawing inspiration manifests in varying degrees, some toeing far closer to plagiarism than others. There’s a difference between transforming a trope into something new and duplicating someone else’s iteration and calling it unique. There’s a difference between imitating someone’s style for practice or in homage and copying it to pass off as your own. And there’s a difference between sampling components of someone’s work with their acknowledgement and wholesale appropriation without due credit. 

In July 2022, outcry regarding a painting of a Black cowboy by artist Gala Knörr brought attention to its similarity to an uncredited film still by the artist dayday. Knörr eventually apologized, stating that her work is “based on many images and media” and her excluding credit was “not ill-intentioned.” The Guggenheim Bilbao ultimately exhibited the pieces side by side, showing and linking the source of inspiration to, ironically, foster reflection on Black erasure.

All artists base their work on many sources—it’s why many can rattle off a list of their influences and inspirations on command. Far from diminishing, identifying these echoes enriches an oeuvre by putting it in conversation with the creative genealogy that made it possible. Imitation is how we learn to create, and incorporation is how we learn to create well. The central issue plagiarism raises isn’t a philosophical question about what constitutes originality or gatekeeping—it’s whether we value the people behind a source of inspiration, without which all other works based upon it would not exist. Nor does the legality of an act mean it’s ethical—it only makes plain what those in power deem worthy of protection.

Some AI users express reservations about crediting their sources. One Reddit user based a custom model on artist Deb JJ Lee‘s work but named it after animation studio Kurzgesagt because they “didn’t know what to call it.” The account was deleted after Lee discovered their art was used without permission or acknowledgement.

By obfuscating sources, users render transparent the mode of engagement AI inherently advances: one founded on mass consumption, where quantity itself is quality. Where intent is obsolete, because the end product justifies the means. Where credit is deprecated, because everyone is an interchangeable part. Where collaboration is eliminated, because the act of creation is estranged into a black box. Where the profession of artist is obviated—save for the elect few—because the product is ultimately privileged over the person.

A fair number of stories on AI and the future of creativity end with the author describing or transcribing their experience with a model. Often this is done with an air of morbid curiosity, wary apprehension, or dismissive ridicule—expressing both unease and self-reassurance over the future of their profession.

But why give the algorithm any more space than it already has? Why not give credit where credit is due?

This essay is itself a form of web weaving, composed of threads both self-evident and subsumed. Hidden within its warp and weft lie, among others: an elementary school teacher’s basic essay structure; Yuan and Yu’s Can’t Help Myself; Dafen Village; a high school teacher’s lesson on strategic paragraph breaks; Benjamin’s aura; Lil Nas X avenging Mr. Van Gogh; a friend’s long memory; Gaiman’s The Sandman; Deleuze’s multiplicity; an author who writes fanfic at 3 am; an editor’s deft eye; Carson’s conception of eros; Weil’s notion of love; years of thoughts and experiences tangential, parallel, and asymptotic to this subject.

I drew from a trove of sources far smaller than what an algorithm has access to, and I took far longer than what an algorithm would require to compose an essay of this length. But I carry forward those who taught and inspire me, transforming and consciously honoring their contributions. As US poet laureate Ada Limón noted, everything has been said before and better than you can—but despite and because of the weight of all that preceded you, your vision will be “tugged beautifully out of the collective past and into the singular present.”

Creating good art is a time-consuming, inherently repetitive, often taxing, and sometimes costly process. It’s frustrating to have a brilliant idea and be unable to execute it exactly as you imagined, if at all. Craft takes years to master, and even the most vaunted of artists can struggle to wield a medium to the specifications they desire. It’s just part of being human.

But artists make art because there is pleasure in chasing an idea that may never be exactly realized. There is joy in parsing the raw materials at your disposal to make something worthwhile. There is love in creating an homage with every work you make. When you strip away the filters of profit, power, and prestige, all that’s left is something you’ve dreamed of bringing into existence, and the time, effort, and care you’ve dedicated to carving it free with your own weathered hands. In the end, as artist Jackie Liu wrote, what else is it all for?

Everyone has the capacity for mastery in art—given the time and resources to practice. So why give it all up to a machine? Why not give people the means to own the efforts of their own creation?

Why not save the pleasure of reaching towards that elusive masterpiece for yourself?