The political function of an aesthetic sensibility

(1) The obvious “objective” reason for the widespread market embrace of AI is economic necessity. The frontiers of financialization and speculation are always useful for patching over the contradictions of capitalism, but they’re picking up now that we may or may not be in the insane desperation stage of declining US empire. Finding the next big thing is not merely a means for a few people to get rich, though of course it’s also that; speculation is a necessary engine for keeping the machine going. If not for AI, the US GDP only would've gone up this year by less than a percent. This is one aspect of the current bubble analogous to, say, the public-facing crypto hype, as manifested especially during the year or so before FTX collapsed.

But the particular psychotic passion and delusion of the moment—its subjective dimension—requires accounting for, I think. Crypto boosters were (and continue to be) delusionally hyped for the liberatory possibilities of decentralized finance and the blockchain—and the actually-existing crypto use-case, transnational crime, has been an unbelievable success—but crypto’s appeal as a public-facing product is as a speculative asset. People put their money in Bitcoin for the same reason they put money in gold or give it to a hedge fund: to turn dollars into more dollars. It stopped being about changing the world pretty fast. (For the die-hards, in fact, the insularity and strangeness of the “scene” was part of its appeal; it was conceived of as a niche, community-generating product.)

So why are we like this about AI?

(2) One reason is that it is, in some ways, genuinely unprecedented. LLM-based chatbots are capable of passing what you could call applied Turing tests, at least as performed by many people, with largely catastrophic results. (Psychosis aside, my personal count of people I know who have had relationships deeply damaged by their friends or partners using ChatGPT as a therapist has recently increased to two.)

Besides this, learning by means of active question and answer is fundamentally different than the process of looking something up in a static, pre-existing document (including a video); given that this relationship is typically predicated on a foundation of trust and authority (doctors, teachers, experts), we are proving quite bad at applying the appropriate skepticism to the output technology capable of extruding persuasively answerish language.

(3) So this stuff is weird, for sure. But it’s weird on the consumer side, and it’s weird in a way the significance of which is not pre-determined.

On the institutional side, the financial prospects for OpenAI, the market leaders and public face of the technology, are unbelievably bad, even as they claim to have over half a billion active users. (I’m not sure if this includes people using services which use OpenAI API calls, which would definitely inflate the numbers.)

So why the frenzy? One reason for the enthusiasm besides ordinary speculative delusion is that capitalists at tech companies are wildly excited about the prospect of automating white-collar workers out of existence. Silicon Valley executives, for example, are giddy at the idea that they can replace their frustrating homosexual blue-haired coders (recall that Marc Andreesen wrote, in his leaked CEO Buddies group chat, that relatively mild Black Lives Matter walkouts were akin to demanding a "cultural revolution") with something that won't raise a fuss about, say, its labor being deployed towards the ends of genocide. Given that government handouts in the form of military contracts are one of the few viable financial ways forward for AI companies, dewokifying their workforce is increasingly economically necessary.

(4) This is just context for the fact that I don't think anyone believes that truly compelling visual art or music or literature, is, in fact, generative AI's primary use-case.

(5) To be clear, their media output isn’t trivial. Chatbots and audio-generation models are interestingly OK at pastiche, though this makes sense given that they are trained on the exact works they’re pastiching. The capacity to produce fake photos and videos that look like they were taken with phones is concerning for those of us who value knowing if something that looks like it happened actually happened, as well as those of us who would prefer that less deepfaked nonconsensual pornography exist in the world rather than more. And I certainly believe that people will continue to attempt to make us accept that the art AI produces is as good as the art humans produce.

But it isn’t, and it won’t be. There’s a reason “slop” was the Mirriam-Webster word of the year; there’s a reason Satya Nadella, in a shockingly solecistic blog post, asked people to stop using the mean word to describe AI output.

I tend to think that (this is another point I think I ripped from Cory Doctorow) from the vantage of our Time magazine people of the year, AI art generation is more about the PR function than it is about actually replacing human actors or musicians. It's something between whitewashing and shock-and-awe. Isn’t it amazing that you can type something into a box and get a picture out? When you think of “AI,” don't think of the completely dysfunctional actually-existing chatbot assistants we've stuck under the place your thumb instinctively taps on every consumer-facing mobile web application. (And definitely don’t think about how the model is just riding probabilities calculated to increase your time-on-device.) Just remember how cool AI visual art is, instead. That’s AI.

(6) Harnessing the aesthetic towards the end of advertising always requires a contempt for art. (Earnest propaganda is a related but distinct discussion.) A comparison might be Aristotle on friendship: the kind of friendship based on getting something out of someone else is inferior to the sort of feedback-loop of virtue that two great-souled people can experience when they lack ulterior motives. True friendship is pointless in the sense that it's not teleological, but it's a necessary aspect of human flourishing.

Analogously, art acts in the world and upon its receiver in a way that exceeds the extractive. In those rare instances where advertisements are aesthetically interesting, they frequently have shockingly little to do with the product in question; in fact, maximizng the part which exceeds the straightforwardly extractive is, paradoxically, one of the main formal characteristic of modern television ads; consider insurance advertising, the central joke of which is frequently that the ads are barely about the dull and unsavory insurance itself, but instead serve as an episode from a kind of Extended Insurance Company Universe centered on, say, a wry and digressive gecko. (It’s awesome what you can do with your budget when you don’t have to spend any money on R&D.) Advertising is to the arts as rhetoric is to poetry. Perhaps this is a naive way of thinking about things, but my increasing sympathy with it is ultimately what this post is about.

(7) Sometimes people frame this in terms of advertising's mind-control capacities. We could call this something like vulgar Debordianism: our desires are so ontologically warped by the pressures of capitalism on our horny lizard brains that we're fucked on some primal, unknowable level, manipulated like the hypnotized. To mention, once again, Cory Doctorow: his explanation of Shoshana Zuboff's work in How to End Surveillance Capitalism is a very lucid exploration of this.

As Doctorow has it, the internet doesn't make people anti-vaxxers through some inscrutable (to borrow his memorable term) Svengalian hypno-magic. The politicians they like say that Ivermectin cures COVID, and then they Google it and, because of the shit they read all day already, the search results say that Ivermectin cures COVID, so they think Ivermectin cures COVID. We all perform variations on this operation all of the time. It's just a lot easier to fall into stupid bullshit now, because our material conditions are wild and implausible enough that any reasonable theory of contemporary society is, on some level, a conspiracy theory, and without media literacy you’re navigating it all on vibes and “common sense.”

I think the surreal diversity of human erotic desires and impulses is an illustrative expression of the failure of mass media to control us. While it’s true that hegemonic norms play an unhappily massive role in the structure of our desires, anyone who knows anything about the internet knows that the airbrushed, conventionally-attractive, heteronormatively-appropriate celebrity is, uh, hardly the only object of modern human sexual fantasy.

(8) (To be clear, I think it’s worth separating the question of the content found on the internet from the overall ad-delivery apparatus inflicted on us by means of the smartphone and social media—which is, I think, a behavioral-psychology nightmare. Conflating the two in experience can lead to a bit of unhelpful, insufficiently materialist melodrama, I think, such as seemed to be the case with Jonathan Haidt’s book about how everything is social media’s problem, though I haven’t read it, because it’s physiologically bad for you to be too annoyed. In any case, I recommend, now as always, Natasha Dow Schull’s Addicted by Design on this. All of which is to say that absorbing your attention isn’t the same thing as making you believe something ridiculous.)

(9) Still: on principle, I hate advertising. I hate it almost more because it doesn't induce some evil trance state, but because it condescends to believe it will affect us.

But it's fucking annoying. We move through the world in a distracting and annoying fog of patronizing bullshit. It’s clutter and noise. Even something like a billboard is a grotesque imposition into the public sphere. We take this for granted, but I'm not sure we should.

(I mean, look: even as our instinctive collective loathing for American personal-injury law is in large part a result of a protracted right-wing propaganda campaign to obfuscate the fact that it frequently requires litigation to secure basic-medical-bill-level remuneration from culpable corporate actors, it doesn’t help their ambulance-chaser reputation to constantly pull my attention towards their billboards while I am literally driving a car seventy miles per hour. It’s a bad thing and a bad scene.)

(10) All that said: no form of advertising I've yet seen exhibits so pronounced a contempt for the human perceptual and aesthetic faculties as the corpus of images and words generated by modern AI, and I think the whole thing is worth interpreting symptomatically.

(11) Consider the AI-generated poem. Considered historically, especially following developments in post-Romantic poetics, it's regressive doggerel. Why are people so impressed by this bullshit?

(12) They like it because they don't know what good poetry is.

(13) This is claim is obviously snobby. Is it elitist? For some values of "elitist," yes. Drop the argument now if you'd like, or check out the newsletter I wrote about Warhammer 40k video games (though I don’t mention the best two, Space Marine 2 and the Owlcat Rogue Trader CRPG).

(14) What I mean, however, is not that we don't have the cognitive capacity to understand true poetry, or that elusive and difficult poetry has a monopoly on quality, or that rhyming is bad, or even that quality objectively inheres in the aesthetic object. What I mean is that, thanks to the erosion of public education and a system of production which renders leisure nearly inaccessible, the population of the USA has been deprived of the capacity to develop, among so much else, a poetic sensibility.

(15) This, then, is one of the weird takeaways of this moment. Critical thinking about the aesthetics of literary writing can now, suddenly, serve a direct social and political function: seeing through AI boosters' bullshit. The annoying joke I saw during a certain run of discourse around underemployed and overeducated (what a stupid word) English majors was, like: yeah, some people are plumbers and doctors, but after the apocalypse we’ll need someone who can analyze poetry.

But (bracketing the question of whether or not right now is “after the apocalypse”), I can firmly say that, because I have given a shit about poetry for a while, I can separate the genuine impressiveness of AI-generated poesy-slop as a tech demo from its nonexistent expressive quality; because I can do this, I do not experience generative AI poetry as aesthetically impressive; because of this, I am inoculated against a certain form of hype.

(16) In fact, I hate it. Everything about AI aesthetics is disgusting. I hate it so much. I'm sick to death of ChatGPT voice. The sycophancy makes it so fucking grotesque. I find it uniquely aesthetically repulsive. I don’t trust emojis anymore. You can't always clock it, sure, but I feel like I see it a lot (perhaps because people are starting to write and talk like it), and when I do I'm disgusted, and when I find out something is AI I fucking hate it, and even if it's not I hate that it looks like that. Glossy, uncanny, dead, trite. “Slop” is such a felicitous term for it. It's fucking puke. (And when Satya Nadella rolls out with five grammatical errors in a single paragraph, it’s like, if it’s not even correcting your grammar what’s the fucking point?) How can I help you!!!!! That’s a great idea!!!! Shut the fuck up.

(17) But okay, it's not like the Time Magazine People of the Year sat down one day and said, “We're gonna make art that looks like shit.” The unhappy fact is that a lot of art looks like shit, especially the art people with bad taste like, and so when they trained their models, they trained them on shitty art, and then people who like shitty art helped steer the models towards generating art that conforms to what they liked, and here we are.

(18) Which is to say that it looks like shit not because it's by robots, but because it comes from a world funded and ruled by people completely numb not merely to the true and the good, but also to the beautiful. Since these are the people who curate popular culture and strip our education for parts, the fact that their sensibilities suck trickles down.

(19) I remembered this article about Nashville songwriters using AI. when I heard a shitty, generic pop-country song on TV during the shitty TV New Year’s coverage. Was that song co-written by AI? I don’t know, but it fucking sucked.

(20) AI slop is a new modality of slop—the kind of slop that allowed us to conceive of “slop” as a general category—but slop as such is a product of our culture writ large.

(21) My response to LLM-generated aesthetic objects comes from my political commitments, yes, but it manifests in aesthetic terms, because our aesthetic sensibilities always express political commitments: not in the sense of "I like this because it conforms to my political beliefs," but because "political beliefs" are a component of our understanding of how the world works and what it is, an understanding in constant dialogue with our perception of art.

(22) Blah blah blah all of this is to say that, for most of the last couple of years, AI has made me want to not care about reading and writing, because our language is already so miserably evacuated of meaning, and it seemed like the final turn of the knife. But the train of thought I’ve hopefully circumscribed a bit here is nice because, to my own surprise, both tasks suddenly feel much more urgent.