Painting of Calliope, Muse of Epic Poetry by Charles Meynier, 1798.THE CLEVELAND MUSEUM OF ART
Michael Harris’s most recent book is All We Want: Building the Life We Cannot Buy.
“It stole my job, simple as that.”
As a journalist I’m used to listening to my peers talk about a job-nabbing bogeyman called “technology.” Craigslist killed local newspapers; Twitter killed magazines. What’s new, these days, is that the bitter chorus has grown: I hear from animators, musicians, screenwriters, illustrators – all of them gulping and wondering where their value now resides. Every content creator has joined the collective chagrin. And our bogeyman has, meanwhile, emerged from the mist in his most defined form – artificial intelligence.
How real is the AI threat? Staggering advances in machine learning, memory capacity and processing speed have set the stage for products that let anyone create fully orchestrated songs, workable screenplays and photorealistic videos. It’s all made possible, though, by an old-fashioned thievery of epic proportions. Many AIs depend on Large Language Models (LLMs) that train on millions of books in order to understand the patterns in centuries of human literature. Image generators such as DALL-E are, likewise, fuelled by the whole history of our visual arts. Technology kingpins (Google, Microsoft, Meta) are insatiably hungry for the human-made content that fuels their software. They just don’t want to pay for it.
Generative AIs violate artists’ rights in two different ways. They train on material that is not licensed to the AI’s owner, and (despite the originality implied by “generative”) they often produce work suspiciously similar to copyrighted material. When DALL-E, for example, was asked to create “an Italian video game character,” it instantly drew a portrait of Mario, as though a copyrighted IP were a “free” idea that had simply occurred in DALL-E’s black-box mind.
This is not a mere nuisance or legal curiosity. It’s a looming employment cataclysm. If products like ChatGPT, Stable Diffusion and Sora can create text, images and video with a similar value to those made by humans then we creatives will go the way of the 18th-century textile workers who were pushed aside by the power loom.
What recourse do we have? Well, there are always lawsuits. Visual artists in California have sued the image generator Midjourney; several authors (including George R.R. Martin and Michael Chabon) have filed suit against OpenAI. Efforts at pushing back have muddled results, though. In Hollywood, screenwriters supposedly “triumphed” over AI in October when their union secured some guardrails, but, just weeks later, another group in Hollywood threw up their hands – hundreds of animation execs declared that “countless” jobs would be consolidated or eliminated by generative AI.
Hovering over all this acrimony is the assumption that what we do – we artists, we writers, we content creators – is fundamentally different from (and nobler than) what an AI can do. The machine’s offering (a mulchy aggregate of previous art) is deemed inhuman, drained of the novelty and individuality that makes creative work meaningful.
However, in our anger and fear, are we missing something? New technologies have a way of teaching us about ourselves, if we let them. They are always, after all, extensions of human will. And so, though the advent of generative AI may well be destructive – it could also illuminate a mystery about the origin of our creativity.
We like to believe that creative folk are a magical breed. We imagine they are gifted, special and set apart. The unique qualities in each novel or painting (signature expressions of the soul!) are bound up with the work’s validity and value. This makes it very unlikely that artists and writers will tell you the truth about their work’s dependencies. We talk about our “process” in mysterious terms – as though the work is spontaneously conceived inside of us “and I just had to express it.” And yet, while time alone is vital for any serious mental task, that solitude is necessary precisely because we have to manage a flood of input from the big, crowded world. There is a dynamic, in other words, between accrual and analysis, between conversation and reflection. The lonely artist in their garret is a pointless figure unless they have first been out at a party.
The truth is: no art emerges without enormous quantities of material borrowed or stolen from others. And, oftentimes, we are willfully unaware of those inputs, as heedless of their value as the black box of an AI that hoovers up libraries and galleries.
The notion that creativity comes from inside the artist is a fantasy well-suited to our individualist age. Artists and consumers co-create this fantasy, preferring to believe that books and songs have a precious value derived from their singular origin. But there are other theories. Consider the ancient idea of the “muse,” for example; a mysterious external force visits the artist and graces them with inspiration. This view of creative generation held sway for centuries and seems to acknowledge that no artist really goes it alone. In the first lines of Paradise Lost Milton calls upon a “Heavenly Muse” as an “aid to my adventurous song.” “Instruct me,” he begs. For Milton, poetry is not something you cobble together in solitude; it’s something you receive, take in. Plato went a step further, writing that, when we read a poet’s work, “God himself is the speaker.”
Are artists really vessels for spirits? Do writers merely scribble down what a god whispered in their ears? If we replace “spirit” and “god” with less charged words like “life” or “experience” – then perhaps. The brain’s 86 billion neurons process enormous reams of information without bothering to involve the “self” in any conscious way. That material is absorbed, reassembled, and finally presented as “new” ideas that we happily receive. Blinking at our luck, we can only claim to be inspired. But isn’t this process more similar to an AI’s work than we’d like to admit? The unconscious mind takes in base content (every book, every conversation, every landscape we encounter) and then the mind offers a new assemblage of those same materials. We call these products “original” but aren’t they just the outputs of our own black box? And aren’t all cultural artifacts (from ballets to TikTok dances) just data sets, then, that fuel our own creations?
The line between original creation and repurposing old work is fuzzy at best. Any writer or artist who is not too full of themselves will admit it.
This admission is tricky, though – perilous, even – because ethical and legal threats hover everywhere. Did the musician “sample” that track, or did they steal it? Did the filmmaker create an “homage” or is it a rip-off? Did the author “reference” that other novel or did she plagiarize it? The distinctions are sometimes vague, misleading. When artists and writers are accused of stealing I often feel a pang on their behalf; it seems so much more likely that carelessness is their crime, and not skulduggery.
When their cancellation roars through social media, though, when the thief is caught, there’s only one moral option for other creatives. We must throw a tomato, too (or at least stand stupidly by while others excoriate them). Otherwise, the transgression could taint us. Otherwise, the myth of the unique and magical artist would be muddied. But these cases raise such ire precisely because every artist and writer knows, on a gut level, that they do copy, that they do take, that they do pass off vintage notions as their own. They attack in order to argue their own innocence.
What do we artists get to be, then – if we can’t be unicorns any more? Maybe we’re just folk with a predisposition for collecting the flotsam we come across, a fondness for quilting together what we gather. The crisis of generative AI has exposed a fundamental truth about the creative act – that creativity never did emerge from our own special souls, it never was born immaculately from our genius. It was always a cobbling of found objects; we pick up and love what others have held. Just as oil is produced by the concentration of billions of past lives, all art is a murky distillation of what came before.
And so, by consuming, digesting and repurposing stacks of content, AI does exactly what humans have always done. AI does not bypass the creative act – it reveals it.
We creatives are scared of AI offerings not because they are “cheats,” but because they are learning to do our jobs in far less time and at far less cost.
To protect human artists during this revolution, we can’t depend on vague and self-aggrandizing ideas about the specialness of creators. We must be more honest than that. By first acknowledging that we, too, are recyclers, just as prone to copying and borrowing as any generative program, we can then think more clearly about what makes the human approach truly worthwhile.
When new machinery forced artisans of the 19th century to question what was truly human about their work, it led to the flowering of the Arts & Crafts movement. It was against the backdrop of the machine’s clean cuts and rigour that the artisans’ elegant traditions came into focus. Around the same time, photography became available to the public and the game of realist painting grew less compelling. Painters did not sigh and hang up their brushes; they actually emphasized their brush strokes, and the impressionist movement was born. Each time a technology seems to “replace” the artist, it in fact forces us to re-examine our purpose, to seek again some human-specific contribution. Why should we fear that a similar creative outpouring won’t greet us again?
That higher definition of human-centred art will likely be produced in tandem with AIs. We’ll learn to use them as tools (those impressionists famously used photography in their research, by the way). The danger, then, is that a feedback loop could begin, with humans inspired by AIs, which are inspired by humans, which are inspired by AIs … If we cede too much of the process to them the chance for radical disruption and novelty could be beaten back. We may find ourselves in a queasy hall of mirrors. Meanwhile, the many transgressive roles that artists play today (story-reverser, nightmare-describer, tradition-challenger) would be subsumed by a single capitalist purpose: easy reproduction of whatever garners the most eyeballs. (i.e. the Marvel Universe swallows us all.)
We cannot know what human-AI creativity will look like tomorrow. We can imagine, however, that very soon the human role could be severely diminished, and that a livelihood in the arts may become available to a vanishingly small number. Securing whatever place we do have will begin, I think, with a kind of humbling, an honesty about our practice. We aren’t geniuses, we aren’t gifted. We are happy pirates and magpies.
All generation is regeneration. And all art is theft.