Eroticism, Interiority, and AI Art

The conversation about recent advances in machine learning-driven image generation tools (I don’t like the term “AI art,” but it’s become ubiquitous, and I am caving to the weight of convention) has—like so many other topics in the churn of cultural discourse—quickly become weird. One of the weird angles is the result of a subset of AI art enthusiasts, who are excited for the possibility of these tools to generate pornography.

There’s a particularly nihilist sub-subset of these enthusiasts, too, which is men who anticipate that the advancing state of porn-generation AI will end their reliance on women for the last thing those women have that they want.

Even setting aside the almost inconceivably dark preconditions that are necessary for this statement to be sensical, I also don’t happen to believe that it’s true, and furthermore I think it’s false in a way that’s relevant to more than just porn as we face the era of computer-generated meaning.

To state some principles up front: I am strongly in favor of human erotic experience and expression. I am both pro-erotica and pro-porn; they are ineffably not the same thing, but both are good. My stance on AI art is one of engaged curiosity and concern; I have run Stable Diffusion on my personal computer and played around with it at some length. I think there’s merit to the argument that it’s unethical to train an AI on a corpus collected without the explicit consent of the people who provided the data for that corpus. I also think that however that argument, its ethics, and the ensuing legality are resolved, AI art tools are likely to continue existing, and therefore will be used to create porn.

Furthermore, I am largely unopinionated on the matter of what people should masturbate to. Call me an erotic descriptivist. Absent other concerns, I don’t think there are bad or good ways to get off or things to get off to. In this matter, I am more interested in what people do than what they ought to.

A final caveat: I’m not interested in litigating the aesthetic merit of AI art. It seems obvious to me that it is good enough, because if it were otherwise, nobody would care. As a special case of this, I’m also not interested in judging the specific imagery the porn nihilists cite as evidence that they will soon be freed from women. It doesn’t matter whether it’s hot to me, or to you—it only matters whether it’s hot to them.

I think their optimism is misplaced. I don’t think it is, or will be, hot—not even to the porn nihilists who most want it to be. I think even the best, most perfectly specific and flawlessly generated porn will lack something even for them, and I furthermore think that what it will lack is not any technical deficiency, but rather the specific thing they’re looking to free themselves from: the participation of another sentient human.

The broadest trend of internet porn over the arc of its existence is one of disintermediation. While exceptions abound, as a general statement, there are fewer layers than ever between the people making porn and the people consuming it. I don’t think this is happenstance, either. It strikes me as an example of one of the smuggest terms of all time from a field notorious for producing smugness. The field is economics, and the term is “revealed preference.”

Revealed preference is the preference of a person in an economic system as revealed by their decisions in that system as opposed to whatever their stated priorities are. Porn is a market where revealed preference may as well be the only kind of preference that exists.

The arc of internet porn began with redistributing professionally created pre-internet porn, moved through camgirls and fansites, and has led us to Patreon and OnlyFans, where artists and models interact directly with—and are paid (nearly) directly by—the consumers of their work. In the amateur realm, it has led to PornHub and Reddit, where people post their exploits out of sheer love of the game.

In the realm of professional porn (by which I mean any porn made for money) , the revealed preference of consumers is for more interaction. In the amateur space, it’s for more legible, identifiable authenticity.

There is a reason for these preferences, and I think it’s that the interior experience of the porn creator is erotically relevant to the porn consumer.

The idea of generating porn with AI is superficially compelling -- the promise of perfect specificity with a minimum of the embarrassing disclosure required to otherwise get that specificity is a potent one. But the AI art nihilists seem excited not just by that promise but also (and very explicitly) by the removal of the cooperation of a real woman as a precondition for the existence of a pornographic image.

I find this excitment puzzling, because it’s not my impression that photos of attractive girls wearing little to no clothing are especially thin on the internet’s ground. If one’s pornographic requirements begin and end with seeing a photo of a naked women, such imagery is abundant and free. The material available and the means of its production and consumption reveal, rather, that context-free imagery of attractive sexual partners is not the essence of what people find compelling about porn.

When someone buys a subscription to an OnlyFans cosplayer, they’re not just paying to see her lewds & nudes, they’re paying for conversational access to her. They’re paying for the interaction. While all but the most naive viewers are aware that this interaction is transactional and contingent, in addition to specificity and personalization (which are now available via AI art) the mutual awareness that the interaction is taking place at all is part of why it’s worth paying for. While it may not be authentic in the sense that the performer would treat the consumer the same way without the monetary transaction, it’s inarguably real in the sense that both parties did in fact experience the interaction.

Internality is relevant is non-interactive contexts, too. Where the OnlyFans girl provides low authenticity but high interactivity, amateur porn producers provide the inverse, and amateur porn’s rising prominence makes a strong case for the potency of authenticity. The reason that poorly-lit, oddly-framed footage of a sexual encounter between people of unremarkable attractiveness can still be compelling is that it feels real. To be torturously specific, part of that real-ness is the knowledge that the people you’re looking at had the experience you’re watching them have.

There are many other modes of porn production and consumption, obviously, and some are surely less contingent on the producer’s internality than others. Likewise, the human sexual experience is wildly diverse, and people are turned on and off by and in spite of a bewildering variety of things. I’ve pretty thoroughly convinced myself, however, that estimation of the internal experience of the other is a foundational erotic element. What the other person is experiencing matters. Do they like it, do they hate it, are they repulsed, intrigued, humiliated, ecstatic? The consumer’s impression of these states is part of pornography’s erotic value, and for these states to be present, there has to be an experience-haver.

There are other ways in which AI art may prove to be of some pornographic value. I do not think, however, that a nudes-producing faucet will finally liberate the horny masses from the tyranny of the internality of the other. Such liberation is not so much impossible as it is nonsensical.