This post was prompted by this Note by Rachel Haywire:
Because I like Rachel, I didn’t get my nose too out of joint at the “People who disagree with me have the following reasons I ascribe to them” rhetorical formula, because let’s face it, we all do that. But I do not like AI art. I don’t like AI writing. I don’t like AI, period. I am a pre-Butlerian Jihadist when it comes to this. I don’t fill my Notes feed with it, but it’s where I stand. I will say why.
I could begin with an Ontological prescript, what is and is not Art, but I’m old enough to know that changing the definition of words creates more smugness than persuasion, and the ontology of Art is a morass thick with theory that I haven’t read, and theory is rarely more than word games. Let’s not fuck around with that just yet.
The problem I have with AI art, or AI writing, or AI creativity in general, is simple economics. Not necessarily in the sense of whether artists get paid, although such is not irrelevant. The Circus, as a class, is allowed to have legitimate class interests.
No, I’m talking about the Law of Infinitude: when X is ubiquitous, no one cares about it. When everyone is a hero, no one is. I’m talking about drowning in the supply of crisp images instantly and infinitely made. Why learn about composition, or color, proportion, how to paint, or how to draw, when you can just smash the Easy Button?
You might think this fear fanciful. You might argue, “real artists would always want to learn”. But whence comes this assumption? Right now Chat GPT and it’s ilk are making a mockery of public education. University students are smashing the Easy Button on term papers. According to Forbes:
Most college students (57%) said they didn’t intend to use or continue using AI to complete assignments or exams. However, 32% said they used it or would continue to use it in the future, while 11% preferred not to answer.
These numbers will only go up if AI is the Wave of the Future (Show Me the Blueprints) that its supporters claim. Which means the entire exercise of writing papers will become useless. And few will be sad to see them go, if you judge the writing by the result. Because the purpose of writing a term paper was never to create cutting-edge scholarship, but to give a student a means of encountering material, of thinking about it. AI alleviates the need to read, write, or think. Push button, achieve result. Repeat until good enough.
Keep in mind, college students are paying tens of thousands of dollars a year to study topics that they have themselves chosen. Unlike K-12, college is entirely voluntary. You don’t have to go. Students should want to read and learn about the subject they’ve decided to study, yes?
I’ll pull my tongue out of my cheek. College has been a joke and a scam for decades, degree mills designed to keep the white-collar middle class perpetually paying for the privilege of existing. When everyone has a college degree, college degrees lose value. So there’s no reason for the pre-debt-slaved to waste their precious drinking and reel-scrolling time on learning anything, when the machine will get them their useless degree more efficiently. Grades appear, without gatekeeping.
Now, let’s extrapolate second-order effects. If college students don’t need to read, write, or think, then they won’t learn to do so. Which means, they will be unable to perform those tasks, not just for their employers, but for any other purpose. Which means any kind of text-based work will increasingly depend on the only reliable source of it, which is always ready, and infinitely reproducible. Already the corporate world is pushing Grammarly nannies on its workers. Even if the media world didn’t want to rely on AI-writing, it might soon be forced to.
I’m not the only one who feels this way about Chat GPT et al. AI-art enthusiast Walt Bismarck does as well. Sure, he’s penned an entire defense of AI art. Why, he can make a symphony on his computer, without learning a single instrument or even how to read music (what Saliere would have given for this sorcery)! But despite his passion, he gives the game away in two separate statements:
When it comes to my AI generated songs, I usually write the lyrics entirely myself, and that is what I care most about artistically. I let the machine handle what I don’t have the technical skills to do myself, using creative and iterative prompt engineering and rigorous curation to make sure it sounds decent in the end. I’ll admit this aspect of creation puts me in the role of producer rather than artist, but the final product is a legitimate work of art and a genuine reflection of my aesthetic sensibilities.
Here we must at last confront Ontology. Why does Walt make the distinction between producer and artist? Because he didn’t actually make the music. He wrote the lyrics, like a cyberpunk Mike Lieber/Jerry Stoller, and let Robo-Scotty Moore, Android-Bill Black, and Electric-DJ Fontana do the rest.1 He recognizes that he’s not an artist, yet declares the result to be art. Which means the machine is the artist. Does the machine know this? Does the machine value what it makes? Is the machine attempting to communicate anything, or simply to mashup the presets according to the producer’s formula? Is there any possibility of the artist surprising the producer, of creating something new?
But here’s the real admission:
When it comes to my essays, the art is obviously in the prose. That I write entirely by hand, and would never dream of outsourcing to Chat GPT.
Why?
Why doesn’t he use Chat GPT? Surely, if AI art is Valid and True and Stunning and Brave, then an essay written by Chat GPT would be as well. Why doesn’t he use Chat GPT to write his song lyrics? He could have an album in an afternoon, and another tomorrow.
Surely, with a minimum of tweaking, Walt could crank out essays on a daily basis. Shit, an hourly basis. He could own Substack. So why doesn’t he do this?
There are two possibilities:
An AI-textbot could not create an essay as original and good as one Walt could actually write.
An AI-textbot absolutely could create such an essay.
The first answer means that AI is inferior, irrelevant, of no value to a discerning person, and everyone’s aesthetic objections to it are correct.
The second answer means that Walt’s essays are irrelevant, a buggy whip modality of prose production, doomed to be overtaken by the overwhelming quantity of a machine that improves every time it’s used. And therefore, Walt is a bigoted Luddite gatekeeping chud (his words) for not using Chat GPT. Or he soon will be.
The common response to this argument is something on the order of “anyone who loses to a machine deserves it!”2 Which, in my observation, is something people who think AI won’t affect their chosen art say. Walt predicts doom on furry-porn artists, but thinks legitimate Art, his Art, will make it through the maelstrom of Infinite Content unscathed. It’s really a fortunate position, to be able to create instant, democratized music without having to trouble yourself with those pesky gatekeeping musicians, while knowing that his writing is in no danger of being similarly democratized. That Gate is a sturdy as those of Vienna.
The history of the modern age, of course, says that quantity overwhelms quality, that noise drowns signal, that automation destroys craft, and that no human activity is immune from this. Some people think the productivity gains will be glorious, and only the unworthy will suffer. They have the mentality of Wall Street brokers and C-Suite executives, whose aesthetic instincts have yielded bananas taped to walls and Disney live-action reboots. For them, art is already product, already nothing but an Intellectual Property to be mass-reproduced with minor tweaks over and over again as long as ROI holds. Am I really to believe that the coming age of “Hey Siri, make me a {Thing} in the style of {something I know}” will alleviate this situation? Does the Fortune 500 fear AI, or are they all in on it?
It may be that my fears are overblown, and that the last film I ever see, sometime in the 2050’s, will not be the Chat GPT-written, Bing AI-animated Star Wars Episode MCMLXXVI: The Jedi Empire Forces Revenge, starring Mark Hamill’s digital clone Hark Mamill. But so far, no one’s bothered trying to persuade me. No one among the AI-boosters even seems interested in considering the possible downside of technology, which at this point in the Modern Age is unbelievably myopic.
So go ahead: make your case. Convince me that in a world where all creativity is done through the machine, by the machine, the vanishing of skill sets we’ve had since the Paleolithic will not be corrosive to the human spirit in any meaningful way. Convince me that generations soon to be born will not regard the idea of playing an instrument or writing words as quaint as plowing a field or churning butter. Convince me that Full Automation World will have a place for human activity beyond consumption of product.
But please, don’t pretend unintended consequences don’t exist. That’s not how any of this works.
If you get it, you get it.
Have these people ever heard the song “John Henry’s Hammer”? Do they recognize themselves in it?
This is well written. Of course, we still hoe our garden by hand, though we have an automatic tiller (not worth it for raised beds.) Art sits in a weird place; some necessities are amenable to being mass produced with little loss, but others, not. Art is in a way a necessity, but one that sits pretty far on the end of lossiness in any mass production. However, what sometimes happens, as has happened with *some* food production and a fair amount of clothing production, is that the method becomes so ubiquitous that the results get imagined in terms of what the method can produce alone. This is not a sinister process as some might suggest, when you have a set of tools, you learn them well so you know what they can or can't do. Many do not understand this, though, and are sold all sorts of fantasies about what they can make with this or that tool (or framework) -- when in reality what would matter is knowing what it *can* do and using that as a frame when imagining what you'll produce.
In musical terms, there is a subtle degradation between the person who knows how to play a variety of instruments using software emulations and one who does not. There becomes a sense of freedom to create, without really knowing the true limitations you've taken on with an emulation. A generation that has never played a string instrument would be surprised at the sounds you can make them produce. Perhaps they might so boldly say that if those things mattered they'd emulate them!
My main chief concern is that "democratization" of "high art" styles just mass produces kitsch. If I had the ability to produce wine on command, I would not be putting out anything above a 20 dollar bottle -- honestly, I had great Sicilian wines for 3-6 bucks. I recognize that my taste in the matter doesn't adequately represent the richness of wine, and only jokingly share my preferences, knowing that they are sort of philistine. But if the average wine enjoyer determined the wines that would flood the market, there would be quite a dilemma; one that the general quality and variety of wine would absolutely drop to bare satisfaction, and two that it would be an impossible problem to solve since I can't really tell the difference between a six and thousand dollar bottle of wine. Even if we set aside mere cost, since European wines are often not as expensive as hoity toity American ones, I could not tell the difference between a great wine and an acceptable one. I do enjoy Port!
What happens here though is a game-theoretical defection process, something like how oligarchy ruins a nation; yes AI "art" is bad in a number of ways, but at least positionally, a lot of people could stand to benefit _personally_ from it. They might not benefit in the long term, but benefitting now while it still matters might make sense. Naturally (and such is often the naive attitude of computer science folks) this is misunderstood under the starry-eyed guise of democratization. Sam Altman has been known to push dangerous nonsense in a very optimistic fashion; so it goes.
There is very little to convince me to listen to anyone's AI-generated "music." There is too much real music by real musicians I haven't listened to yet.
This is the dilemma that confronts fans of the arts in the 21st century; we are confronted with surfeit. I am not sure I want gatekeeping, exactly, but such as it is, it's not working. Just an endless wall of undifferentiable schlock.
20 or 30 years ago, we all knew That Dude who buttonholed you with his crazy eyes. "Hey man, you gotta listen to my band's new record! CD is only $5!" Now he wants you to listen to his AI music. We really have come to live in Geocities.