Rob Horning

@robhorning.bsky.social

4734 Followers 268 Following

robhorning.substack.com

When earlier periods of skilled practice are treated as more fully human, the social order that produced them can easily appear better than what replaced it. The past, then, seems superior to the present. This is why such arguments risk becoming reactionary. When we mourn the uninvention of skills, we may also – unless we are very careful – end up mourning the loss of the social structures that surround them. That slippage is the core mechanism by which skill nostalgia feeds into a longing for a bygone world.
Prediction markets are “truth machines,” in the sense that they give people incentives to figure out what is true and bet on it. But they are also, obviously and increasingly, falsehood machines, in the sense that they give people incentives to make bets and then distort reality to make those bets pay off. Because of prediction markets, it is now lucrative to figure out what song will be the most popular, but it is even more lucrative to make a different song look most popular. Prediction markets can make it harder to know what is true.
They used to define themselves by their passions, spending years reading, practicing and mastering their fields in order to answer difficult questions. When AI can produce competent analysis in seconds, learning begins to feel unnecessary.
Artificial intelligence is an anti-community technology whose principle civic innovation is to reduce social trust and engender a greater sense of unreality in our lives.
According to Kracauer, these individuals “obtain from the totalitarian dictator ship the pleasure of being subordinated to persons embodying an overwhelming fate. These persons control him and provide his masochistic desire to obey with delectable sensations. ”41 Kracauer claims further that Fromm was right to argue that “the liberation from autonomous decision making and, with it, from doubt is one of the greatest satisfactions that the authoritarian state has to offer its fol lowers.”
The problem for the makers of lo-fi imagery is that this existential crisis, if it is indeed what motivates them, can’t be resolved at the level of aesthetics. AI can emulate most legible aesthetic styles easily enough; embedding any sort of recognizable pattern in digital content seems like one of the worst ways to prove your humanity,
As we have seen, however, Kracauer also recognizes that authority— even in a fascist society— does not rest on violence alone, which explains the crucial role of spectacle and welfare in cre ating a “mass addiction” to totalitarian propaganda. Terror also aims to have the indirect effect of convincing individuals to willingly abdicate their own ability to judge reality and to accept instead the fabricated “reality” of totalitarian pro paganda. Kracauer underscores the theatrical nature of fascist spectacle, which encourages the masses to accept fiction for reality and teaches them that “the play is more than a play, and madness is the norm.”
National Socialist agitators do not entice with intellectual proofs and arguments, but rather seduce with the “primordial energy of rhythm.”46 In fact, rhythmic formations gain the upper hand as, and to the extent that, the swarm of previous promises loses its value. Shocks are orchestrated in random material to prevent any respite from the tension; waves of propaganda are sent out and, as required, either build steadily or swiftly ebb away; and, in all cases, care is taken to coordinate tempi to good effect. These occurrences are lent propagandist significance by the rhythmic quality of the course they take: the skillful sequencing of quiet and action, of push and pull. If the proclamation of content, of substantive and attractive goals, was once decisive, now the appeal to the psycho- physical state associated with the pseudo- reality is predominantly through form— the form in which the propaganda accomplishes the requisite acts and resolutions. The further this process advances, the more the nihilism with which the propaganda handles all convictions reveals itself, of course, and the more undisguised is the eruption of the will to power, whose product is propaganda. Still, the “primordial energy of rhythm” could hardly provide lasting compensation for the destruction of the contents of propaganda if the propaganda didn’t also succeed in making people forget about its own vacuity by stamping the now transparent will to power as the content itself
But the entire internet — well, the four biggest apps we use to navigate the internet — has, quite suddenly, morphed into a horrible new version of television, seemingly programmed entirely by low-paid worker-users in the global south. Fake fans and fake shares abound, as every social media platform starts to devolve into some bleary combination of 9GAG and WorldstarHipHop.
The clearest contemporary example is the Large Language Model. An LLM is trained on the accumulated output of human writing–billions of documents, articles, conversations, books–the materialized General Intellect, scraped from the public internet and converted into a proprietary asset. The collective social brain of humanity is processed and sold back as a private product. The pattern is identical to every enclosure that came before it.
In the case of the real subsumption of labour under capital, all the changes in the labour process itself, analysed by us previously, actually take effect. Labour’s social powers of production are developed, and with labour on a large scale the application of science and machinery to direct production takes place. On the one hand, the capitalist mode of production, which now takes shape as a mode of production sui generis [in its own right]; changes the shape of material production. On the other hand, this alteration of production’s material shape forms the basis for the development of the capital-relation, which in its adequate shape therefore corresponds to a specific level of development of the productive powers of labour.
Socializing among friends is not nearly as profitable as convincing users to consume more and more polished media. Sociality is not as assetizable as media, which itself can only be assetized to a point
Parasocial relationships are one-sided connections, where individuals keep tabs on the lives and movements of people – like celebrities – who do not know us and feel no pressure to reciprocate
Today’s social media platforms are no longer centered around sociable activities.
In an essay on X last week, Ed Elson, host of the Prof G Markets podcast, defined the “clip” as something distinct from “short-form content,” writing, “They are specifically snippets of long-form content — things like podcasts, livestreams, and TV shows — a byproduct of something else. In other words, every clip is short-form content, but not all short-form content is a clip.” He argued in a followup post that, counterintuitively, even though a clip, according to his definition, is a slice of a larger product, they have become the main product for the majority of audiences. And the entertainment industry has noticed.
“Everyone Is Using A.I. for Everything” nowadays, a.k.a. vibe-living, and if you don’t, you’re a misfit outsider who should be stoned to death in the town square to prevent contagion, and then A.I. should resurrect you virtually from your data so you can be stoned to death in the virtual town square, for infinity.
It’s exactly this dynamic that seems to preoccupy some of our most obsessively agentic individuals. There’s a common prediction among those hoping to outstrip their ovine, mimetic peers: The A.I. models, they say, will eventually supply all of the effort, training and expertise that have historically stood between humans and our ability to simply make things happen. Once all those annoyances are pushed out of the way, the only thing separating the world’s winners from its losers will be the sheer motivation to act, the raw ambition to do a thing — pure, bold agency. You run across this notion all sorts of places, sometimes as a frantically held belief and sometimes as a sales pitch. A sufficiently agentic layperson, someone argues on X, could sit down with an A.I. and produce the same work as a Ph.D student. “In the age of A.I., becoming a super agentic individual is a superpower”; “Being a highly agentic individual is going to be so important with the rise of A.G.I.”; “In the A.I. era, become an ‘agentic individual’ to thrive! Discover how work evolves with A.I. agents. Don’t get left behind!” All that will matter is your own superior drive and will; every other aspect of achievement can be handled, as if by magic, in the guts of some colossal data center.
With so much of our world being so rapidly reshaped for such stupid reasons, it’s hard not to feel as if we’re living in one of the canonical A.I. doom scenarios without even noticing it.
In 2024, “brain rot” was selected as the Word of the Year by Oxford University Press, beating out other meme words like “lore,” “demure” and “slop.” Last year, it was “rage bait” — another term for another kind of online distraction — which beat out “aura farming,” another meme. In 2023, it was “rizz,” which beat out “beige flag” — something from TikTok. Since 2021’s “vax,” Oxford has consistently given the W.O.T.Y. to some sort of internet meme. Maybe it’s because the public votes on it, but maybe it’s because this is what our culture produces now. It almost seems like a direct challenge to the idea that language evolves to describe ever more of our world. What if it can instead degrade, or become caught in the gears of a more powerful meaning-making engine?
guy wearing a Travis Konecny shirt
A couple of months ago, Reggie James built a “situation monitor dashboard for the boys,” suggesting the business model of “making every man feel like he’s in the situation room for $50/month.” Borrowing so much of its aesthetic from sports gambling, again, the north star for situation monitoring is not just the 20th-century control room but the casino sportsbook, which truly fulfills the dream of wall-to-wall screens and their relevant metadata (a perfect setting in which to watch March Madness by the way), and the possibility that you could not just absorb it all but actually act upon it—that you, as a human, and not a computer, could take in all that extra information and then do something useful with it
We live in a world where there is more and more information, and less and less meaning
Such would be the successive phases of the image: it is the reflection of a profound reality; Disneyland is presented as imaginary in order to make us believe that the rest is real, whereas all of Los Angeles and the America that surrounds it are no longer real, but belong to the hyperreal order and to the order of simulation.
Such would be the successive phases of the image: it is the reflection of a profound reality; it masks and denatures a profound reality; it masks the absence of a profound reality; it has no relation to any reality whatsoever: it is its own pure simulacrum
users — and influencers especially — are unwittingly working for the Oligarchs as entertainers without being able to articulate their own political positions because engagement itself has become the dominant (or only) ideology, reducing the entirety of their worldview to being watched and followed online.
Quizzes like the Times’ are games that L.L.M.s are designed to excel at. A.I. writing is literally optimized to be the writing most people prefer in A/B preference tests; the main thing an L.L.M. chatbot “wants” when replying is to be generating the text that its users would choose as a good answer to the prompt over all other possibilities.
People want more content than humans can supply. This does not mean that, taken as a collective whole, there is not enough content in the universe of extant media. Rather, for any given person, there is content they would be willing—even eager—to consume that is not yet being provided. However, making this content comes at a cost. The space of hyper-specialized content people would consume is vast. But for most of this niche content the audience is too small to justify its production. In economics, the balance between the benefit to potential consumers and the cost of production is called optimum product diversity
The reliably credulous New York Times described prediction markets as "infrastructure for the legitimacy of event outcomes", echoing the confidence of company founders, and went on to say, "Platforms like Polymarket and Kalshi now allow users to bet on virtually any outcome of a future event." Coverage like this cites the range of disparate possible bets (Oscar nominations, Grammy winners, how often Elon Musk will tweet this month, whether Israel will strike Iran) as proof of unlimited potential. This is despite the fact that the majority of trade volume is on sports outcomes. Non-sports events don't generate enough trade volume to be worthwhile for the platform or a user. Platforms decide what events are available to trade on, and they determine the terms of payouts. Speed of growth does not indicate long-term viability, but rather investor pressure to demonstrate proof of concept at increasing scale and attract more users every day.
he 2020’s defining feature, a statement of the real you against the horrific “better you” that was sold to you by crumbling systems.
You will start to notice a certain thematic consistency among the most bullish AI advocates: a hyper-personalized fixation on personal productivity gains to the occlusion of all else. Including factual reality, ethics, and foundational empathy.
No wonder Mr. Thomas, who is 78, often feels frustrated. He fantasizes about punching a young tech worker in face. And yet. He had ChatGPT write a speech for his wife's birthday. It was beautiful and eloquent. All of which means the future of A.I. could probably go either way.
I have argued in the past that people are overestimating the organisational level benefits of AI because they are extrapolating from individual experiences, and speeding up production behind a bottleneck doesn’t increase output (although it might reduce it). But one thing I haven’t emphasised enough is that bottlenecks are not natural obstacles – they are, in most cases, the consequence of increasing production until you hit a bottleneck. If AI removes a bunch of bottlenecks, that won’t be used to produce the same output faster and cheaper, it will be used to produce a lot more output until a new bottleneck is reached and requires human intervention. (Weirdly, there was a two week period after the announcement of DeepSeek when all the techbros were wailing at their share prices and shouting “its Jevons Paradox you idiots”, but this got really quickly forgotten).
Vectofascism does not simply produce a lie. What characterizes it is rather the production of a calculated undecidability, of a gray zone where the very status of the statement becomes indeterminable. Post-truth is not the absence of truth but its submersion in a flow of contradictory information whose sorting would require a cognitive effort exceeding available attentional capacities. This strategy exploits a fundamental asymmetry: it is always more costly in terms of cognitive resources to verify a statement than to produce it. Producing a complex lie costs a few seconds ; demystifying it can require hours of research. Vectofascism pushes this logic to the point of transforming veracity itself into a variable of algorithmic optimization. The question is no longer ‘is it true?’ but ‘what degree of veracity will maximize engagement for this specific segment?’
Last February, the writer Coral Hart launched an experiment. She started using artificial intelligence programs to quickly churn out romance novels. Over the next eight months, she created 21 different pen names and published dozens of novels. In the process, she discovered the limitations of using chatbots to write about sex and love.
Our suggestions for alternatives to anthropomorphizing language are presented in Table 3. For two categories, Emotion and Human role analogy, we found that there were no simple replacements. Rather, when faced with such anthropomorphizing language, the best bet was a larger reframing of the matter at hand. For the others, we could often find simple edits within a sentence, as exemplified in the Table 3.
If internet firms are defined by their fastest-growing monetized products, well, Meta is basically a Reels company, one that successfully chased TikTok into continued relevance, allowing Mark Zuckerberg to throw money at his next big chase (into generative AI). This isn’t just a formal change from the News Feed to Stories to predictively recommended vertical videos, though. It’s a long (and nearly complete) process of platform desocialization. Platforms originally defined by keeping up with people you know, or have at least heard of, become something fundamentally different.
Creativity is a wonderful solution to boredom, but you can’t hope that it works the other way around. So the paper from 2014 that everybody cites that says boredom makes you creative: There’s loads and loads of problems with that paper. We tried to replicate it and found the opposite: The more bored we made people, the less creative they were.
While many mentioned positive effects of the bots — like helping patients understand their diagnoses — they also said the conversations deepened their patients’ feelings of isolation or anxiety.
All those who have power over workers and consumers are going to need to use it to make them start accepting and using the Generative AI tools our industry has collectively invested trillions in.
At a glance, existing companion products seem good enough to replace human boyfriends
Unlike the fibre-optic cables of the dot-com era or the railways of the nineteenth century, AI chips must be replaced frequently as their performance fades and technology improves. If, due to profitability concerns, investment suddenly seizes up, a shrinkage in AI availability relative to its current abundance is a material possibility. Theoretically, if the scale-back of capital expenditure outweighed cost reductions from improvements in AI processes, the legacy of the AI boom would not last long, and available computing power for ordinary AI queries could decline.
This is where the limits of technical reasoning become clear. Determination of the optimal solution set is a problem for technical reasoning to solve, but choice among the options in that set is not, since from a technical perspective, these options are equivalent in rank, no matter how different they are in content. Hopes that advances in computational power might overcome any need for social and ecological choice—making it possible to digitally plan the economy—fundamentally mischaracterize the problem, once we recognize the existence of multiple, non-compensatory social goals. The challenge is not simply processing more data. It is deciding what matters—negotiating trade-offs and choosing among possible futures. Reason, experience and collective deliberation will inform these choices but never entirely resolve them.
One idea I’ve been circling around is that “slop” is that which is “fully optimized” to its domain to the point of texturelessness or characterlessness. “Slop” in this sense is anything designed to be as easy as possible to produce, sell, and consume, but it’s particularly slop at the point where all or most other players in the same space adopt the same strategies
The idea of AI produced art or music is in some sense of culmination of the dead weight of existing intellectual production over the living. One of the striking things of the streaming model, of movies, television, and music, is that its ability to cheaply deliver content is almost matched by its inability to produce anything new. Spotify can make the entire discography of dead artists available at the click of a button, but it cannot pay or support living artists; Netflix can buy up Warner Brothers, Disney can buy everything else, but the cost of this is an undermining the entire film industry by shifting distribution to streaming. In some sense AI solves a problem that the streaming services has created. Spotify may turn out to be better at creating the basis for AI generated songs than promoting living artists.
Petter Ruddwall knows the idea of AIs becoming sentient and seeking to get high with code-based “drugs” seems “stupid.” But the Swedish creative director couldn’t get it out of his head.
In two large peer-reviewed studies, AI chatbots shifted voters’ views by a substantial margin, far more than traditional political advertising tends to do.
In the coming years, we will see the rise of AI that can personalize arguments, test what works, and quietly reshape political views at scale. That shift—from imitation to active persuasion—should worry us deeply.
Photographic images become part of an infinite, continuous, and largely ahistorical space of synchronously available visual information, fed by billions of digitally available images from online sources – photographic as well as non-photographic, historical, and current, canonized and marginal. In this virtual archive of potential images, there are no clear boundaries between media, genres, artistic techniques, and visual clichés. Everything becomes a “style,” a reproducible esthetic look, including photography itself.
In twentieth-century theory, however, the accidental, unintended, seemingly insignificant photographic detail has become a defining feature of the photographic as an esthetic phenomenon, most notably for Barthes (Citation1981). The famous ‘punctum’, the almost accidental detail that cannot easily be subsumed under the more obvious culturally coded meaning of the photograph, is analyzed by Barthes as a kind of excess of contingency that transcends any legible caption, written description and ultimately any intended meaning. As visual details that were never prompted, the strange distortions in AI-generated images also create an excess of visual contingency beyond the reproducible patterns, stereotypes and clichés that the model was trained to reproduce, thus transcending any intended meaning. In contrast to Barthes’s punctum, however, the contingency manifested in the impossible and distorted spaces of AI image generation is not the contingency of the material world, but the contingency of the statistical process that generates these images.
much of what we dislike about A.I.-generated text from a formal perspective--it’s generally cautious, inoffensive, anodyne, predictable, neutral, unmemorable and goes down smooth--is a product not of some inherent L.L.M. “voice” but of the training and fine-tuning processes imposed by A.I. companies, which are incentivized to make their chatbots sound as annoying and bland as possible.
Being digitally alone, then, is not merely a social condition but a method of relating to oneself. In conversation with a chatbot, loneliness becomes visible as this existential condition is translated into text and archived, interaction after interaction. For some, this aids in the unfurling of patterns they could not otherwise name; it might surface feelings they would rather keep private. Technology mediates a fundamental, familiar act: talking to oneself via another. And what begins as a private exchange soon reveals itself as a social system, reflecting both the self as well as the architectures of isolation that shape us all.
What I have tried to put into words here is both a feverishly involved and a passionate state of mind, something like a synthesis of what absorbs people in artistic and scientific creativity, in political activism, in revolutionary proclamations, in chess and roulette, in the stock market, and in libidinous dreams. It is a state of mind that does not intensify and then fall away, as in an orgasm, but that maintains itself at its orgiastic climax without interruption through a lifetime. For this state of mind is not physical but cerebral. Images are steering the telematic society in this direction: toward a continuous cerebral orgasm.
the “male suckerfication crisis”: The process by which young men are continuously cultivated--by politicians, influencers, podcasters, sportscasters, and the very structure of social networks--as dupes and suckers, both for the enrichment and empowerment of Trumpist politicians and movements but also in service of the profits an ever-expanding roster of gambling and gambling-adjacent industries, among them app-enabled retail day-trading, crypto speculation, predictions markets, and of course, actual sports betting.
One way of thinking about Twitter, especially in the Elon Musk era, is as the latest and most sophisticated version of the mailing lists that have powered the conservative movement, and the various scams and cons and schemes that constitute it, since the 1960s--in other words, as a huge and immediately accessible collection of suckers, waiting to have their avidity and paranoia exploited for cash. The predictions markets here are happy to play the same role with heavy X.com users as crypto scams and meme-stock hustlers did on the same site a few years ago, or as cash-for-gold ads did with Fox News viewers a decade before that, or as direct-mail get-rich-quick schemes did with Goldwater voters deep in prehistory.¹
it was a super-chill community where anyone could ask literally any question and there were 15 people there to help you.
he’s watched the group spiral into the depths of AI paranoia as images of phony crochet projects and impossible-to-execute patterns flooded the feed
This kind of AI-generated irony functions as an unconscious without a subject, carrying out a significant labor on which ruthless political powers can cash in. It is a work that pulls us all into the orbit of a generated déjà vu, where everything is possible, but nothing can happen anymore.