Software's Culture Problem
Machine learning models are cultural artifacts. They encode and reproduce textual, audio, and visual media; they participate in human conversations and spaces; and their interfaces make them easy to anthropomorphize. We lack appropriate cultural scripts for these kinds of machines, and will have to develop this knowledge over the next few decades.
The US (and likely much of the world) lacks an appropriate mythos for what “AI” actually is. This matters: myths drive use, interpretation, and regulation of technology. Inappropriate myths lead to inappropriate decisions, like mandating Copilot use at work, or trusting LLM summaries of clinical visits.
Consider the broadly available myths for AI. There are machines that essentially act human with a twist, like Star Wars’ droids, Spielberg’s A.I., or Spike Jonze’s Her. These are poor models for LLMs, whose protean character and incoherent behavior differentiates them from most humans. Sometimes the AIs are deranged, like M3gan or Resident Evil’s Red Queen—a closer analogue, but one suggesting a degree of efficacy and motivation that seems altogether lacking from LLMs. There are logical, affectually flat AIs like Star Trek’s Data, and efficient killers like Terminator; LLMs are the opposite, producing highly emotional text while being terrible at logical reasoning. Hyper-competent gods like Iain M. Banks’ Culture novels are obviously not this: LLMs are, as previously mentioned, idiots.
Most people have essentially no cultural scripts for what LLMs turned out to be: sophisticated generators of text that suggest intelligent, emotional, self-aware origins—while the LLMs themselves are nothing of the sort. LLMs are highly unpredictable relative to humans. They use a vastly different internal representation of the world than us; their behavior is at once familiar and utterly alien.
A few better myths exist for today’s “AI”. Searle’s Chinese room, Chalmers’ philosophical zombie, and Peter Watts’ Blindsight—which asks what happens when humans meet unconscious intelligence—all come to mind. The closest analogue for LLM behavior might be Blindsight’s Rorschach. Most people seem concerned with conscious, motivated threats: AIs could realize they are better off without people and kill us. The real concern is that ML systems could ruin our lives without realizing anything at all.
Authors and screenwriters have a new niche to explore. Any day now we might see a trailer featuring a villain who speaks in the register of ChatGPT: “You’re absolutely right, Kayleigh. I did drown little Tamothy, and I’m truly sorry about that. Here’s the breakdown of what happened…”
Beyond Static Text
The movable-type press and subsequent improvements in efficiency catalyzed broad cultural shifts across Europe. Books became accessible to more people, the university system expanded, memorization became less important, and intensive reading declined in favor of comparative reading. The press also enabled new media, like the broadside and the newspaper, and the interlinked technologies of hypertext and the web created new media as well.
People are excited about using LLMs to understand and produce text. Reports and books that used to be written by hand will be produced with AI; people will use LLMs to write emails to colleagues, and recipients will use LLMs to summarize them.
This sounds inefficient, confusing, and corrosive to the human soul, but it is also probably not looking far enough ahead. The printing press was never going to remain a tool for mass-producing Bibles. If LLMs did get good, there is a future in which the static written word is no longer the dominant form of information transmission—in which a few massive ML services like ChatGPT become the medium, and publishers put content through them.
One can envision a world in which OpenAI pays chefs to cook while ChatGPT watches—narrating their thought process, tasting the dishes, and describing the results. This information feeds general-purpose training, but it might also be packaged as a “book”, “course”, or “partner” users could ask for. A famous chef, their voice and likeness simulated by ChatGPT, would appear on a screen in your kitchen, walk you through a dish, and advise when the sauce fails. OpenAI takes a subscription fee and dribbles out (presumably small) royalties to the human “authors” of these works.
Alternatively, we may train purpose-built models and share them directly. Instead of writing a book on gardening with native plants, one might spend a year walking through gardens while a nascent model watches—showing it plants and insects, interviewing ecologists while it listens, then asking it to perform additional research and “editing” it by asking questions and correcting errors. These models could be sold or given away like open-source software.
Corporations might train specific LLMs as public representatives. One can only imagine children learning to induce the Charmin Bear on their iPads to emit six hours of profanity, or to tell them where to find matches. Artists could train Weird LLMs as personality art installations. People might download licensed or bootleg imitations of popular personalities and set them loose in home “AI terraria” where they’d live out ever-novel Real Housewives plotlines.
What is the role of fixed, long-form human writing in such a world? At the extreme, one might imagine an oral or interactive-text culture in which knowledge is primarily transmitted through ML models. Writing books becomes an avocation like memorizing Homeric epics. Writing will survive in some form, but information transmission does change over time. How often does one read aloud today, or read a work communally?
Content Moderation at Scale
With new media comes new forms of power. Network effects and training costs might centralize LLMs, leaving most people dependent on a few big players to interact with these works. That raises questions about the values those corporations hold, and their influence—inadvertent or intended—on our lives. Facebook suppressed native names, YouTube’s demonetization algorithms limit queer video, and Mastercard’s adult-content policies marginalize sex workers. Big ML companies will likely wield increasing influence over public expression.
We think of social media platforms as distribution networks, but they are also, in large part, moderation services—the platform weighs in on every idea their millions of users could express. By offering machines that generate a staggering array of content, OpenAI et al have placed themselves in the same position: they must weigh in on every possible utterance their bullshit machines could extrude. Meta, for example, had to decide how much to let its LLMs flirt with children, and whether they can say sentences like “Black people are dumber than White people.” The broader consequence is that general-purpose ML companies are intrinsically tasked with encoding, formalizing, and adjudicating essentially all cultural norms, at unprecedented scale. That affects everyone who interacts with ML content, as well as the human moderators who curate it.
Generative AI and the Future of Erotic Culture
Fantasies don’t need to be accurate or consistent—they only need to be compelling. That makes machine learning a natural fit for generating sexual content. Some of the earliest use cases for Character.ai involved erotic role-play, and platforms like Chub.ai now host a wide range of NSFW chatbot characters. Social media and adult sites are flooded with AI-generated images and video, both entirely synthetic creations and altered versions of real people.
The current moment offers unprecedented possibilities for online sexual expression. Niche communities with highly specific visual interests—say, fans of giant anthropomorphic animals—previously had to settle for illustrations, crude image edits, or 3D renders. Today, a text prompt describing an elaborate scenario with a tall vampire noblewoman can yield surprisingly compelling results.
But pornography is also an industry, and human attention is finite. AI-generated content will likely absorb some demand that previously went to commercial studios and independent creators. The economic effects are already visible: OnlyFans personalities and the contractors who handle their erotic messaging are facing competition from language models that can perform similar work at scale. This won't necessarily eliminate the market for amateur or professional erotic art—drawing and photographing remain enjoyable activities in their own right—but the economics will shift.
There are also concerns about how AI-generated imagery shapes self-perception. People with body image issues already struggle with curated social media; an endless stream of idealized—or simply bizarre—AI-generated bodies could intensify unrealistic comparisons. Tools that let people enhance their own dating-profile photos or create deceptive personas will only become more common.
On the positive side, generative AI may enable new kinds of liberatory erotic experience. Virtual reality already lets partners perceive each other through avatars; real-time image synthesis could extend this to hyper-realistic transformations. People might explore body configurations, genders, or identities that aren't otherwise accessible, turning private fantasies into vivid, shared experiences.
Media has always seeded new erotic subcultures. Westerns and midcentury biker films shaped the leather-bar aesthetic of the 1970s; superhero bondage fantasies have their own extensive canon; the blueberry transformation kink traces directly to Willy Wonka. Furry fandom, with roots in early animation, exploded after films like the 1973 Robin Hood. Generative AI will inevitably produce its own distinctive kinks and fetish communities.
Drone fetishists are an early example. This kink centers on the erasure of human individuality in favor of machines, hive minds, or alien intelligences—and its practitioners have embraced AI image generation enthusiastically. Groups like the SERVE Hive, the Golden Army, and Unity all rely on ML-generated visuals to enact their fantasies, and the slightly uncanny quality of the imagery enhances the experience. The aesthetic of the medium reinforces the theme of the fetish.
There's also speculation that people will start fantasizing about being an AI themselves. Robot kink is well established, so erotic narratives about having one's personality replaced by a language model, or hypno tracks about having a small context window, seem like a natural extension. Queer theory will have plenty of material to work with.
Companies like OpenAI have been cautious about allowing explicit sexual content, which creates its own dynamics. Motivated users will attempt to jailbreak filtered models, and sexuality becomes a useful test case for identifying AI systems. In practice, writing deliberately transgressive prompts in email exchanges can reliably trip up an LLM at the other end.
The Emerging Aesthetic of AI Slop
AI image generators produce recognizable visual styles: hyper-detailed fantasy realism, distorted hands, glossy pornographic imagery, and surreal Facebook clickbait. These patterns constitute a family of aesthetics that are becoming cultural signifiers.
Visual styles have always carried meaning. The Nagel look defined the upscale hair salon in the 1980s; Tuscan-themed home decor and the greige palette of HGTV mark particular eras and social classes. Eurostile Bold Extended signals retro-futurism, while certain typefaces announce gentrification. It's inevitable that ML aesthetics will acquire their own set of connotations—the question is what they will signify.
One troubling answer is fascism. Marc Andreessen's Techno-Optimist Manifesto borrows from Marinetti's Futurist Manifesto, which in turn fed directly into Italian fascism. Andreessen's political alignment and Elon Musk's embrace of AI-generated propaganda imagery suggest one vector. But slop is not politically univalent: ML imagery is used across the political spectrum, from AI-generated leftist cartoons to gay party promoters' promotional materials.
A more practical connotation may be that of cheapness and inauthenticity. Where corporate marketing teams have professional artists, smaller businesses that generate their own visuals with AI may initially benefit from the "polished" look—but audiences may eventually prefer hand-drawn signs and authentic imperfection as more trustworthy.
Any aesthetic is eventually appropriated for irony and nostalgia. Extremely online teenagers are likely to reconstruct, subvert, and romanticize today's AI slop, much as vaporwave reclaimed the corporate aesthetics of millennial computing. Expect future generations to celebrate the too-many-fingers look and share garbled AI text—a profound cultural shift that will nonetheless have its own appeal. The aesthetic movements of the past, revived as false memories, can still produce works of genuine beauty.
