AJ Chronicles: It’s Getting Expensive to Prove You’re a Human Artist

Artists now have an increasingly expensive new toll gate to pass through. Last week the Commonwealth Foundation cleared this year’s short story prize winners of “cheating” after allegations that some winning entries were written by AI. The Foundation went back to writers and asked them to hand over drafts, outlines, manuscripts, and other evidence of their creative process, then spent a month poring over the evidence. Then they exonerated the winners. But Granta, the prize’s publishing partner, quit anyway — “for the sake of our own editorial integrity,” it said, ending the partnership.

A troubling story that says a lot about where our culture is right now. We’re now focused on “doping tests” to determine if artists have cheated. Rather than pee in a bottle, however, we’re depending on AI detector tools and documentary proof of human creation. Did the Foundation double down to ask what’s the best writing? No. They cared more about how it was made. Perhaps that’s important. Of course it is. But, in a way, it’s now an impossible question. Moreover, it may ultimately be the wrong question.

The AI detector tools are being widely proven as inaccurate. And artist work documentation can be incredibly inconsistent and onerous to produce. The new requirements impose an uneven and time-consuming tax on artists. Some artists are feverish documenters of their process. Others work organically, without notes. So if you’re an artist consumed in your work and you neglect to document, is your work now invalid? If you don’t have the resources to employ approved documentation tools, are you uncertifiable? This is a mess.

Surveys of artists across fields show they’re using AI in their work extensively. And why not? If it helps make work better, it’s hard to resist. In some creative fields, AI is only a small step from technology already in use for years. Movie CGI special effects. Music auto-tune. Photoshop processing of images. Each of these technologies in the hands of amateurs can look and sound amateurish. In the hands of artists they make imagination sing.

In the analog art world, technology is also hard to separate. In many significant ways, the story of art is one of technological progress. The modern symphony orchestra, for example, owes its sound to improving technologies over the course of centuries. A significant reason music from three centuries ago sounds different from a 21st Century orchestra is improvement in instrument technology, which allowed composers and musicians to expand what they were capable of.

AI that critiques and challenges your ideas can make you better. So where is the line? Pre-production AI use that made an idea better is surely undetectable. But is that okay? AI is expert at iterating hundreds, thousands of versions of an idea while an artist works through a concept. Is that cheating?

We’re building mechanisms for certifying that art was made by humans, and it is getting expensive. Contests now demand a documented chain of custody for inspiration. Insurers want proof of human-authorship before they’ll cover productions. The legal ground shifted when the courts held in the Thaler case a few months ago that AI-created work can’t be copyrighted. Artists used to be presumed human. Now they’re asked to document and prove it.

Here’s a problem: increasingly the proof doesn’t mean anything. Detection tools flag human writing and music as machine-made often enough that students and professionals are being falsely accused. And research keeps confirming that people — including experts — can’t reliably tell AI music or prose from the human kind. Spotify just beat a class action alleging billions of bot-inflated listener streams. Meanwhile synthetic AI artists and avatars rack up real plays from real listeners who don’t know or don’t care.

The incentives when it comes to the human/AI creation debate are all also aimed one way: studies report that when audiences learn a song or a story is synthetic, they value it less. So the rational move for an artist using AI is to hide rather than declare it. We’ve built a certification system that punishes honesty but also can’t reliably catch the dishonest.

Missing the Point?

This is all a mess. And yet, I think it ultimately misses the point. The backlash against AI is huge right now. For all sorts of very understandable reasons. I think this is a good thing: it’s forcing us — for the first time in a long time — to debate the basic values of art. What we expect art to do. How art should be created. What are the qualities that determine good art. Who can create art. Most of these questions aren’t new, but the prospect of non-humans making things has forced us to be more explicit about, really, defining what art is and what it does.

And sorting out what art is made by machine and what by human has a purpose, I guess. We want to know, as we have this debate about values, so we can figure out what’s important.

But I go back to the fights we had as journalism entered the digital age. So much of the debate about digital journalism was about the journalists, the need to save jobs, the importance of the news gathering process. These were all very important. But ultimately all the market cared about was the journalism itself, not the journalists.

The argument seemed almost always about preserving the making: the newsroom jobs, the trained journalists, the credentialed process. It was rarely about whether the work was getting better at the things audiences actually wanted from it. Newspapers defended the institution of reporting while readers drifted toward something the institutions dismissed. Convenience. Voice. Point of view. Authenticity rather than institutional voice.

When YouTube arrived, professional videographers were appalled — bad lighting, shaky cameras, no craft. They were right about the craft and wrong about what mattered. What audiences wanted, it turned out, was not polish but authenticity. The shaky camera and bad lighting screamed authentic. Ultimately, the technology redefined “good video” and the language of good video changed.

How a thing was made has been a thing, but rarely the main thing. People care about what the thing does to them. How it makes them feel. So finally we’re having the serious discussion: What is art for? What do we mean by good? The verification panic is a reaction to the realization the culture’s old definition of excellence is up for grabs.

Clearly we want and need artists, and thus the ferocity of the debate. It’s worth really figuring out why. And that’s certainly a reason that has changed many times in human history. This week there was a fascinating story in The Conversation about research into how listeners in Bach’s time felt music in their innards. And another argument in The New York Times from John McWhorter that “music you can see” is the future because images now mediate everything. Not who made it, not how. Does the art make you feel something? Or not? That’s the question that ultimately matters.

Also Worth Your Attention

The “theft” economy is quietly becoming a licensing economy. Two deals last week demonstrate the trend. Getty Images cut a deal to put its licensed images inside ChatGPT, and Google put $75 million into A24 to build AI filmmaking tools. Meanwhile The Atlantic mapped the staggering catalog of music scraped without permission into the AI music generators. The fight over whether training data is theft is real but watch where the money is going: toward making the inputs licensable and the tools native inside the studios. The new AI creative economy infrastructure is being built not by legislation or in the courts, but by corporate America making deals that benefit themselves. Arts leaders should assume the “is it theft” question gets settled by contract, not principle — and plan for which side of that contract they want to be on.

Is the first AI museum art, or a very expensive lava lamp? Refik Anadol’s Dataland opened in LA to mixed reaction — Artnet calls it an “ushering into our new contemporary art world,” while the skeptics ask whether sensory overload is the same as meaning or as art. It’s the definitional fight in miniature: the argument isn’t about who made it but whether it does anything to the people standing inside it. Whatever Dataland turns out to be, it’s being judged on its effect, its value as an experience, and not on how it was made, which is exactly where I think debates over using AI in art is headed.

Editor’s Note: These weekly essays are meant to connect stories from the week to larger trends and ideas across the arts world. Want to support our work? Subscribe to ArtsJournal’s free newsletters. Or better yet, support us with a premium ArtsJournal subscription at $5/week or $52/year. Much appreciated.