You didn't make that

The problem of authorship in the age of machine art

We live in an age of wonders. Unfathomably complex machines will, with a few words or sentences, create the most remarkable imagery for us to enjoy. “Le Voyage”, an AI generated video of a painted world, went viral on Twitter last week.

The author of this arresting machine hallucination goes by the handle bandyquantguy, and he made it using a variety of AI tools, including Google’s Veo 3, which creates short video scenes from text prompts. He’s made a lot of them — this is #28 in the “Paint World” series.

When I saw Le Voyage on twitter earlier this week, I was too quick to dismiss it, and the entire medium of AI-generated video, as “not art”. The debate over what is and isn’t “art” is ultimately too tedious to be interesting to me, and always resolves to a maximalist definition (“if it made you think or feel something it’s art”) or a subjective one (“someone thinks this is art so it is”). So, fine. AI video is an artform. I don’t think this is much of a concession given that many people would gladly hang AI-generated prints on the walls of their home. Not should it be controversial given the creeping definition of the term “art” in the modern era, which long ago expanded to encompass purely-conceptual pieces like the infamous banana taped to the wall.

This was worth $120,000 to someone

And it’s not recent. Painting a fictional signature on a urinal was gallery-worthy over 100 years ago, as Marcel Duchamp demonstrated with “Fountain.”

Many people don’t realize there isn’t one “Fountain”, it’s a purely conceptual piece of artwork that used several different urinal fixtures in different installations over many years.

So this debate didn’t originate with AI tools. Anyone familiar with the development of photography has heard the failed argument against its consideration as an artform: that the use of a tool that precisely records an image reduces the practice to mere craft, the artist to technician. But photography requires technical mastery not dissimilar to other built artforms, as well as a keen eye for one’s subject, hard work in creating or scouting tableaus to photograph, and above all, an eye for beauty — or, in other words, taste.

All of these elements are present in AI-generated imagery as well, especially taste. The term “slop” is apropos for the vast majority of images created by these tools, simply because the people creating them are so enamored of its novelty that they fail to exercise any discretion in what they share with others. It’s like the old joke: What’s the difference between a good photographer and a bad one? The bad one shows you every photo they take. The AI art we don’t dismiss as “slop” is that tiny minority that was curated by somebody with taste.

So, it’s not the case that AI art isn’t “art”. It’s that AI art has no artist.

Ancient wisdom

Suppose you find an old oil lamp in woods. You rub it and out pops a genie, who grants you three wishes. You ask it to create you a world-class oil painting of an old man cradling a newborn or something similarly schmaltzy, and the genie bobbles its head and produces an opus to rival all the renaissance masters. Who is the artist, you, or the genie?

We built that genie. “We” in this case refers to a couple distinct groups of people: first, the scientists, engineers, researchers, programmers, and technicians that developed the machine learning algorithms and then embodied them in silicon; second, and just as importantly, the countless human artists, writers, and other creatives that uploaded their creative work to the internet to be ingested by these hungry shoggoths. When you type a prompt into an AI tool, the creator of the product is, by metaphorical weight, 99.999999% not you. To the extent such work has authorship, it’s almost infinitely diffuse, spread so thin as to be meaningless. You didn’t make that, the genie did.

Asserting that your genius with prompting rises to the level of creative artistry is so delusional and wrong-headed that one struggles even to call it narcissism. It’s more like a category error. Everybody has met a guy who insists that he had the idea for a famous app like Instagram or Snapchat way before it was released. Even granting this is true (it never is), the idea is not the important part. Ideas are cheap. Execution is hard, and execution is everything. And in the case of AI art, the execution is 99.999999% the work of other people.

Last week a finance guy went viral for making the risible claim that a household income of $140,000 is the minimum to not live in poverty. There was lots of incisive criticism of this claim, but what I found most astounding was the author’s response to an accusation that AI wrote his article. Yes it did, he responded, and I’m proud of it.

This mindset is utterly alien to me as a writer, akin to a celebrity with a ghost-written memoir insisting that she did in fact write it herself since she was interviewed by the ghost-writer. It seems to view creative production as simply a means to an end, the act of creation itself dreary busywork we should be glad to offshore to less valuable entities like an LLM ghost-writer. I suppose I don’t have any real objection to treating art or writing as mere commodity with no provenance if your aim is something other than creative expression, such as creating things to be sold or trying to influence policy. But it seems to me that the typical AI art booster wants to have their cake and eat it too, to outsource the creative work onto the robot and still claim the output as a true expression of their own beautifully unique individuality. Speaking again as a writer, an LLM does not “help you write” by producing text in response to your prompt. It replaces your creative voice with its own.

If there’s anything to fear and despise about the rise of AI art, it’s not that creative types will be put out of work, although many surely will. But that’s the life of an artist since time immemorial — for most of human history only the truly exceptional artist could make a living on their work, usually with the support of a wealthy patron. Rather, the real threat this technology poses is devaluing the very act of creative expression itself by flooding the market with good-enough simulacra of creativity. Duchamp and the banana-tape guy first became famous in their local art scenes for creative genius, then laundered that fame into their bizarre conceptual art pieces using the fame as collateral. There is no shortcut to this process. An unknown concept artist painting his name on a urinal is simply a lunatic rightly ignored by everyone. The most skilled prompt “artist” is at best a technician, and their work will be swallowed by an endless ocean of similar products with nary a ripple, producing not even the traditional fifteen minutes of fame.

AI slop is harmless enough on its own, but the danger is that drowning us in slop will inevitably devalue real art in a simple equation of supply and demand.

The true test of any artform is this: can the person making the art possibly hope to get pussy as a result of their efforts? Or, to be less crass, can they use their artwork to gain social status and acclaim? If no, then they are not an artist. Prompt authors will not get pussy, therefore they aren’t artists, QED. But worse than that, the hyper-saturation of infospace with spectacular AI-created artwork, even if ultimately shallow and empty, actively suppresses the urge for human artistry. Save me the bromides about true creatives being in it for the love of the game — status-seeking is an irreducible human motivation, and when you remove that incentive it cannot but warp the outcomes in predictable ways.

Art cannot and will not ever die. If there’s a silver lining in this situation, it’s that the firehose of slop will naturally sour the public toward conceptual art, which they mostly already don’t like. Democratizing the production of Duchamps, taping a banana to every wall, will only serve to further undermine and accelerate the collapse of “idea as art.” Expect the post-AI era of art to emphasize the real and the physical, to feature unforgeable signals of labor, skill, and effort above mere conceptual novelty or beauty of form. That would be a nice outcome, and I hope it comes to pass.

As for bandyquantguy, the tweet featuring his AI-generated painted world went massively viral on Twitter for a few days, generating millions of views. But almost none of this infamy splashed onto the author’s original YouTube videos, which the tweeter helpfully duplicated and uploaded instead of embedding (although to his credit, he did provide links in a follow-up tweet). Paint World #28 currently sits at a whopping 4,100 views on YouTube, not even a muffled sneeze of attention by the standards of online virality. Considered together, these numbers may herald the arrival of a new artform, but they categorically do not signal the arrival of a new artist.

Or — if they do, his name is Veo.