Not only that, but the tropes that make images seem “evidentiary” or “indexical” or “documentary” can also be optimized for and slopified, extracted algorithmically from billions of images to present the feeling “this is real” as a generic, malleable quality, as something simultaneously very clearly fake. This makes for a kind of meta-experience in which we can suspend disbelief in our ability to suspend disbelief. At the same time, the feed’s ceaseless demand for novel images requires that simulated “news” images be constantly produced and just as readily dismissed, scrolled through, evoking topics being discussed generally and shaping images that can be directed along different vectors toward different ad hoc demographics derived from troves of user data.

Anyway, here is the rejected draft:

Like all newsworthy events now, the current Middle East war has brought about “a cascade of AI fakes” that have filled the feeds of social platforms. The nations at war may be conducting an all-out information blitz across any available media, with no referees or monitors to constrain it, but at the same time there is also a free-for-all race to use the same weapons to pander to audiences across the spectrum. Sometimes the clips serve explicitly propagandistic aims, hoping to melt enemy morale or forward a false sense of what happened, but often they are merely opportunistic, giving platforms material to stick in a user’s feed so that every time they refresh it they find something new and interesting, unhampered by any of the lags in traditional verified reporting or the constraints of needing to find genuine sources and illustrative details. Such clips cater to an audience that might vaguely think it wants to “monitor the situation” but is also possibly just bored and hoping their phone will keep them amused.

Regardless of their overt political slant, such clips are indifferent to whether they inform or misinform us, as long as they hold our measurable attention. They are simply propaganda for themselves — for the idea that all news can and should ultimately be experienced as entertainment. On-demand footage should exist of anything that might fit with any user’s expectations of what they should be seeing, and not necessarily what has actually occurred. Everyone should see content that rewards the tiniest amount of attention with a comforting sense of having already gotten the point.

Cumulatively, these clips don’t add up to an overarching narrative, even a false one. Instead, they encourage a post-narrative mode where enjoying memes doesn’t depend on understanding the complexities of the bigger picture or getting more context. Instead news footage, simulated or real, is treated as freely combinable with other material from the cultural archive, allowing us to enjoy spotting the formulas and references with no heed to propriety or decorum. When the White House mixes war footage with scenes from action movies, video-game screengrabs, or pro sports highlights, it is seizing on this post-narrative approach, which corresponds to Trump’s indifference to coherently explaining why he thinks the war needs to be fought. Like the memes, it is self-explanatory; you get it or you don’t.

This opportunism has long held true for all sorts of news events. As the recent blizzards played out across the country, you may have seen a video like this one play across your social media feeds: It shows a silver sedan spinning out on an icy highway, causing the 18-wheeler directly behind it to jackknife, sending its trailer pirouetting over the car, which emerges curiously unscathed as the truck’s cab slams into a concrete lane divider. It’s not altogether implausible, and perhaps a few years ago, before generative models had reached their current proficiency and before platforms began withdrawing resources from content moderation, it might have been convincing. But no one is fooled for long now. Everyone’s seen plenty of AI slop.

Such low-stakes clips are in some ways more revealing about the effect of deepfakes, which have long troubled commentators warning of an incipient crisis in truth in which no one will be able to trust what they see in media, pushing us toward a paralyzing disbelief of everything. The video’s Facebook caption describes it as “A shocking incident captured by a traffic camera,” yet if you click on “See more,” you’ll find that the full text ends with a disclaimer, forthrightly declaring “this video is created using AI, and the story is for your entertainment.” That is about as coherent as Trump’s explanations for the Iran war, but coherence is no more the point of this clip than being real or accurate. The point, of course, is to crack the demands of a platform’s algorithms — the primary consumers of clips such as these. Viewers are suspended between two algorithmic systems, one that generates the content and one that shows it to us, having our attention harvested to help train them to converge.

But why would any actual person want to watch something fake? This might seem like a natural question if you think of social media feeds as news, but it’s silly to ask about entertainment. We engage with various forms of fiction all the time, in part because we like wish fulfillments, escapism, the feeling of suspending disbelief, the apparent confirmation of our beliefs and biases. It can be satisfying when something we maybe couldn’t imagine so vividly looks just like we would have expected, just like we’d want it to. These clips aren’t trying to justify themselves through being persuasive but by being diverting, by subordinating the complexity and overwhelmingness of events to our personal response. It’s more important for them to be emotionally legible than genuine — graspable in an instant.

AI slop clips often latch on to news stories and try to show viewers what the news would be like if it could be stripped of its context and presented as maximally engaging to an individual user. In this clip, of a dog pulling a stroller out of oncoming traffic, the caption asserts that what’s shown is “similar to real events that have taken place,” as if on the implicit understanding that an approximation is not only sufficient but perhaps preferable. Why settle for a real event when you can consume an enhanced and heightened reality presented in a similar idiom?

Both of these clips, purported to be filmed by mounted surveillance cameras, depend on an idea of what viewers expect visual evidence to look like, but convert that expectation into a purely formal component, something that is just another trope to play with. Like the war clips, they are not so much indifferent to facts as parasitical on the means of establishing them, illustrating how evidence can be produced as content — that is, media shaped by the incentives of social media feeds — and ultimately as slop: low-effort content whose production can be increasingly automated.

Evidence establishes the factuality of something regardless of whether people want to pay attention to it or believe it; content simply caters to and reinforces the desire for spectacle. As evidence and content have come to co-mingle in feeds, it has created a confusion between the two. Evidence-like material becomes an alibi for spectacle; it mimics verifiability but invites us to disavow any need for verification. We can experience what we see as true without having to fit it into a narrative or a confirming context. Like a magic trick, it temporarily short-circuits skepticism to allow for a kind of pleasure in simply accepting what we perceive. They also reinforce the feeling that it is more fun to be tricked than informed, and that indifference to truth yields instant emotional rewards.

Sometimes, as with magic, generated clips portray miraculous or catastrophic or impossible events, catering to escapism. But they also show what we expect to see. Slop caters to an idea that anything should be viewable, especially in a world so saturated with surveillance as ours: Hasn’t it already reached omnipresence, and shouldn’t omnipotence follow? This corresponds with the idea that generative models already contain everything that could possibly be perceived in its matrix of billions of data parameters. Slop clips deliver on both fantasies at once: AI can show you anything, and it is truer than ordinary truth.

So opportunists can try to identify apparent gaps in news stories — holes in what an algorithm might want to show viewers —and use generative AI to fill them, as with attempts to conjure the face of Nancy Guthrie’s suspected abductor, or to augment footage of ICE shootings in Minneapolis. These clips can aim to satisfy a ghoulish curiosity or cater to schadenfreude — when a Chinese skiier was mauled after trying to take a photo with a snow leopard, generated fakes of that selfie circulated — or they can be more overtly political as with the official White House account’s ongoing meme campaign. There are fake videos of urban decline and fake videos of Jeffrey Epstein, fake videos of bombings in Gaza and Ukraine and now Iran, fakes about basically anything that somebody out there might have feelings about.

These images proliferate not because they are mistaken as authentic documents but because, as Claire Wilmot argued in a piece for the London Review of Books, they “offer clear, illustrative diagnoses” of “alleged problems” — they are “showing everyone what’s really happening,” as someone knowingly sharing anti-immigrant deepfakes tells her. The clips don’t offer factual information — which would underscore our inability to intervene in the large-scale events immiserating us — but ideology, saving us the effort of having to think things through.

Generated video caters to an audience’s desire to see their beliefs in the form of evidence, and consume what they already believe as if it is being irrefutably established by real events. They visualize simple explanations for us so that we don’t have to put effort into them or take responsibility for them. These kinds of clips allow us to experience such beliefs as content without having to “really” believe in them. They just appear before our eyes to provide reassurance. They are self-verifying for their audience because they depict an idea that they already take for granted.

The steady flow of such material normalizes the idea that the news will immediately be recast into more satisfying shapes, and that we can depend on our feeds to allow us to modulate our emotions and be fed a kind of clarity, if not relief — a manageable rhythm of disbelief repeatedly resuspending itself. The flow of images, and the emotional regulation it affords, outweighs the impact of any particular image of video. This flow serves as an ongoing intervention to neutralize other less convenient content

There is some consolation in calling these clips slop, which implies that they are obvious garbage that everyone already sees right through. Our disavowal of our desire for these clips, and the effect they have on us, is built right into the term. Slop’s dismissibility is part of its appeal; it makes deception seem apparent and contained, under the viewer’s deliberately suspended disbelief. Slop flatters us by letting us feel in on the fakeness while enjoying the view of the world it puts forward. It offers a discourse that is not hard to keep up with, one that delivers a constant sense of wins, of recognition, of getting it without trying. It panders like a sycophant chatbot. Its harms are not in persuading us of something nonfactual but in eroding the sense of collective responsibility for shared information.

This is an underlying message of all content: that we should be able to see what we want to believe without having to invest effort into building that reality. Slop may have its social uses —animating unimaginative group chats, perhaps — but mainly it invites us to isolate ourselves. Rather than seeking out information from a recognition of our personal ignorance and out of a sense of civic responsibility — to help build and participate in a political world shared by others, grounded in ideas that have been collectively developed and articulated — we can ingest massive quantities of simulated information and clumsy propaganda that makes us feel full of ourselves. If civic duty as a purpose for becoming informed has atrophied; slop circulation aims to obliterate it entirely.

Another way to put that is that the era of evidence is over, and there is only content. But the problem can’t be addressed by trying to overwhelm content and slop with a steady diet of truth — genuine documentary images and “real” evidence. Frequently, critics react to slop virality with nostalgia for when the mass media monopolized and rationed the distribution of the images that made for a shared sense of what was happening, the photographic record, a set of images everyone would see and could keep up with. That media environment seemed to provide trusted media institutions with ethical limits on commercialization and ideological polarization, a clearer separation of news and entertainment, and no ability to affiliate oneself with news events by rebroadcasting them, making every moment a chance to gain attention for oneself or enjoy parasocial participation in an us vs. them framework.

As any number of 20th century media critics argued, the media industry subordinated truth to spectacle, winnowed out opposing viewpoints and marginalized voices, parceled out “reality” in commercialized bites, and imposed a false consensus. Still, in our current slop era, there is a tendency to romanticize mass media as a golden age of “seeing is believing,” overlooking its coercive power and narrow perspective, and acting as though the images themselves compelled consensus and belief through their documentary fidelity — that media could be self-evidently true or false, with unmistakable moral implications: the image of the napalm girl imposing ethical clarity about the Vietnam War on everyone who sees it.

It would be convenient if “documentation” could absolve us of politics and align people automatically to some objectively correct position that a certain view on reality would seem to some to compel. But this idea that an image can do our political thinking for us — that they require no interpretation and simply convey an inarguable truth — is essentially reactionary, the same idea that animates AI slop and other forms of propaganda, that what we see should be enough to dictate what we believe. The AI images and the “true-in-themselves” empirical images reflect the same laziness with respect to taking and owning one’s view of the world. In social feeds, those two kinds of spontaneous truth are continually brought together, destabilizing them both, with only the momentum of the feed, our continual engagement with images and submersion in media, sustaining their apparent purchase on reality.

The “crisis of truth” doesn’t derive from some newfound power of media to trick us, but from the sense that consuming media alone is enough to let us live in the real world. But no images can give us a privileged relation to the truth, an automatic awareness of objective reality, a better sense of empathy, or a surer moral compass. No media can liberate us from mediation. Engaging with media doesn’t mean “monitoring the situation” — it entails being involved in the propaganda war, not wallowing in slop.

Instead of consuming images from media circulating through our feeds, free from any particular interpretive context, we might instead build specific context for understanding by tending to the social relationships, institutions, conversations, and ongoing investments that make interpretation viable. This is what being politically engaged amounts to, not expecting an information diet alone to do the work for us. Think about the way Minnesota activists framed their practice as being observers — not simply saying, “Watch our footage,” but instead demonstrating that bearing witness is a collective practice that becomes stronger the more people do it, and produces solidarity along with a shared view on reality. The Trump administration, in its quest for cruel content — “vice signaling” as this Guardian piece puts it — pushes the opposite view: Consume content and “enjoy” the sensations and reactions it prompts you to feel in your isolation; just look at your phone, rather than point it with others at something together.

According to a recent study from Deezer, a music streaming company, 97% of people polled “couldn’t tell the difference between fully AI-generated music and human-made music in a blind test with two AI songs and one real song.” If everyone guessed randomly — if it were literally impossible to tell the difference — wouldn’t the figure still be more like 67%? So what is going on here? Do people hear a difference but then invert its significance in their mind? Or do they just reasonably default to labeling anything that sounds like music as “real music” regardless of whatever machines were used to assist in the process, especially under contrived survey conditions when the music they are being exposed to has no other context.

What does “fully AI-generated” even mean here? It’s not like software just spit out these songs without some human developing the systems, training them on human-produced content, and prompting them for song-like outputs according some specific parameters. Human intention put those sounds into those study participants’ ears. And how were the “human” songs chosen? I’m guessing these were not Jandek and Shaggs records. If it were all a bunch of contemporary smooth jazz, would you be more or less surprised by these findings? What if it were EDM? What about if it were a collection of wan, acoustic covers of pop hits from the 1970s, or some other genre populated by interchangeable anonymous performers?

Does it mean anything to distinguish between “AI-generated” and “human-made” when music in general is so often governed by predictable formulas and a limited number of notes, when what makes music “real” to us is not necessarily how it was made and how many synthesizers and computers or whatever were involved, but the conditions under which we consume it? Much of the music I hear in hotel lobbies seems “fake” to me, whereas anything a friend plays for me I will take as “real.” And I can’t tell “real” instruments from electronic ones, live performances from overdubbed ones, loops from human-executed repetitions, and so on.

The study seems premised on a latent idea that consumers will be harmed somehow by “artificial” music, as if this were akin to being sold adulterated baby formula or something. But the line between machine-generated and human-made music is not especially clear or stable, and it’s not especially relevant in light of the larger question of how music is circulated. I think the various forms of automation involved with that have more of an effect on how we appreciate music than the degree of automation involved in how it’s made. Or to put that differently, maybe the line between music production and music circulation is also being increasingly blurred, and the anxiety over AI-generated music could be better understood in those terms: that music can be made solely to be distributed to us, and to render music less social, less of something that people would bother to share or bond over.

I personally don’t listen to nearly as much music as I used to (I can’t read anymore while music is playing, I get too distracted), but I still spend a disproportionate amount of time worrying about my relationship to it. I haven’t fully adapted to the problem presented by having a nearly limitless amount to choose from — to the problem of seeing that as a problem. Like the proverbial Depression-era grandparents who can’t help but hoard food later, I grew up with a sense of cultural deprivation that makes it hard for me (like many people my age, I assume) to navigate an apparent era of abundance. Media technology and media companies have changed faster than my means of formulating aesthetic interests and investments could adjust.

This is hardly a new condition, and I’ve lamented about it in probably dozens of posts before, but now the problem manifests for me more specifically as a crisis of novelty, and of being unable to separate value from it. Knowing that I will never run out of new music to listen to and that no scarcity will ever force me to listen to anything more than once, increasingly I don’t. I treat unfamiliar albums as essentially no different from mystery novels, as if the only satisfaction they should afford is allowing me to move on to the next one. I’m reading the books only to find out how some particular and arbitrary crime was solved, and I’m listening to the albums only to hear how a set of generic conventions, familiar reference points, and influences were worked out in this particular instance. It’s become a kind of procedural listening.

Consequently, my relationship to music has become more ephemeral and perhaps even more like data processing than when I was fixated on assembling MP3 libraries and perfecting the metadata. Now I find about some records or musicians or genres I’ve never heard, cue up a series of albums, listen to them once, and then mentally check them off as completed. More often than not, I am working my way through a “never played” playlist; it’s as if my favorite kind of music has become “new to me.”

None of this prevents me from indulging in dreams of a deeper kind of relationship to music while having these disposable experiences. As a record plays I’ll begin to situate it within a narrative history of the progress of my personal taste. I might be telling myself how great it is and how fortuitous that I now know about it, and that it will certainly become a “staple in my listening diet” or whatever cliché I’m using with myself to make this fantasy seem plausible. I’ll often remember a line that has stuck with me from a Spin magazine article I read when I was in middle school, where the writer joked about gluing a copy of the Soft Boys’ Underwater Moonlight to their turntable because they couldn’t foresee needing to hear anything else for a while. Isn’t that the goal, to love a record that much?

What I think I want in my encounters with art is something that will interrupt the consumption feeds that technology has integrated me with — that rhythm of having a very minor interest aroused and disappointed and rekindled again for a cycle that can persist for as long as I surrender to it. Yet instead, I find myself adapting nearly all my listening habits to that cycle, as if all one could ever want is to say “Next.” As much as I might rationalize this mode of listening as a kind of “omnivorous curiosity,” that omnivorousness seems more like indiscriminateness, and the curiosity like distractability.

So what’s to stop me from taking this approach to listening to what might seem like its logical conclusion and start getting into generative music? I complained in a previous post about generative fiction, in the form of Nanni Balestrini’s 1966 experimental novel Tristano. It seemed totally pointless to me to read a book no one else would ever read, and I couldn’t find any compensation for that pointlessness in the idea that other readers of their unique copies of Tristano and I were playing the same game, participating in the same process. And I felt exhausted by the idea that I was compelled to be the “real author” of the text, having to marshal my own creativity to invest its mishmash of signifiers with significance.

Would it be different with music, given that listening to music doesn’t necessarily entail looking for meanings or indulging any desire to make interpretations? Why not embrace not only the random bleeps and bloops of procedural electronic music but also software-synthesized generic pastiches like Breaking Rust, whose AI-generated music recently hit Billboard’s Country charts? Better still, why not have a set of algorithms produce music ad hoc based on my browsing habits, so my listening can become even more purely feed-like? Every song would be guaranteed to be completely new — shouldn’t this make me feel like I am on the frontiers of the listenable, discovering the heretofore unheard? What could be more gratifying to one’s curiosity than that? And yet, at the same time, I can be fully assured that I am hearing what I am supposed to hear, the music that will magically conform to my tastes while conforming me to social norms or at least to how “the system” understands me. And no idiosyncrasies of any human musicians can interrupt or muddle my direct consumption of the ideology encoded in a particular kind of song, the quality that resonates with who I think I am when I hear it.

In a recent London Review of Books essay about Brian Eno, Ian Penman highlights Eno’s self-professed “obsession” with generative music, which, as Penman points out, is “an odd thing to be evangelical about, somehow both cutting-edge and old hat.” This also describes not just the idea of generative music, which has been around since at least the early 20th century, but the music itself, which is necessarily “new” on one level while being necessarily formulaic on another. Eno, Penman writes, is “so stoked by the idea of having ‘no one definitive version,’ he even asks whether our current habit of listening to favorite pieces of music over and over again will one day seem ludicrous.” My knee-jerk reaction on reading this was to think Eno was the one being ludicrous. But then I began to think about my own problem, how rarely I listen to anything “over and over again,” and I wondered whether I was on Eno’s ludicrous side despite not being all that stoked about it at all.

It would be no more than a fatuous, provocative pose to advocate for listening to infinite versions of the same song rather than trying to develop a canon of music worthy of repeated listening. But viewed from the sort of perspective that media theorist Vilém Flusser sketches out in Into the Universe of Technical Images (1985), all of what seems like the history of music can also be construed as basically all the same song anyway. Once we remove the blinders that are imposed on us by outdated, text-based modes of cognition, we see that everything in the world is just a “heap of particles” that will eventually be configured in every possible permutation. All music is, in that sense, generative music, a “program” that gives arbitrary elements some more or less familiar form, much as photography in his view is basically a systematic concatenation of pixels (and not a documentation of some given reality).

Flusser argues that “technical images” — i.e. media — have displaced our relation to a shared reality; everyone is instead awash in generated representations that have no fixed connection to a base truth, if there were one. Flusser, who works in a similar vein as Jean Baudrillard, is prone to dropping statements like this:

The world has become meaningless, and consciousness will find nothing there but so many disconnected elements. We are, absurdly, in an absurd world. Bending toward the world is therefore an unsuitable stance and must be abandoned.

Media replaces the meaningless and random expression of matter in the world with something that is potentially meaningful, because it can be construed as having been shaped with intention and directed toward some aim and some audience. The content of media, Flusser claims, is not what’s in the images so much as how they are targeted, or “programmed.” The function of images is to program people, to condition how they see certain specific things but also to condition their entire experience of reality to resemble the nature of images, which don’t emerge from some necessarily linear development of causes and effects but can be generated to depict whatever. They don’t speak to “history,” as texts in Flusser’s view do. “What we call ‘history’ is the way in which conditions can be recognized through linear texts” he argues. “Texts produce history by projecting their own linear structure onto the particular situation.” Images instead articulate the world as cyclical, progressing stochastically through all the possible combinations of the stuff that happens to exist. “Current events no longer roll toward some sort of future but toward technical images,” Flusser writes. “Images are not windows; they are history’s obstructions.” (This is similar to Baudrillard’s theories about the “precession of simulacra.”)

With that understanding generalized to the entire population — the view that events happen only in order to make media, and not that media exists to record a history that unfolds independently of it — society becomes unsustainable and superfluous, replaced with atomized individuals programmed with their own bespoke sets of images, existing in a world that requires no outside social confirmation to cohere from moment to moment.

This, of course, is one way to understand what “social media” really is, the replacement of the social with media. Flusser’s work can be mined for prescient descriptions like this:

The penetrating force of technical images drives their receiver into a corner, puts him under pressure, and this pressure leads him to press keys to make images appear in the corner. It is therefore optimistic nonsense to claim to be free not to switch the television on, not to order any newspapers, and not to photograph. The energy required to withstand the penetrating force of technical images would project such a person out of the social context ... However, the reception of technical images does not end the communication process. Receivers are not sponges that simply absorb. On the contrary, they must react. On the outside, they must act in accordance with the technical images they have received: buy soap, go on holiday, vote for a political party. However, for the interaction between image and person under discussion here, it is crucial that receivers also react to the received image on the inside. They must feed it … This feedback enables the images to change, to become better and better, and more like the receivers want them to be; that is, the images become more and more like the receivers want them to be so that the receivers can become more and more like the images want them to be. That is the interaction between image and person, in brief.

And since social media, in this view, isn’t about communicating with other people, social feeds can readily be populated with AI-generated material that more efficiently establishes this isolating (and nourishing) feedback loop.

AI-generated music could work the same way, opening another intensifying feedback loop between the isolated self and the “system” or “ideology” or “power” or have ever one wants to characterize what Flusser describes as “the envisioners.” The generated content absorbs the receiver’s actions to produce more content, distracting them from the fact they can no longer act on a shared world itself, as figured by an idea of history. “Human freedom” goes from an ability to intervene in history to an ability to get more compelling images sent at you. Get stoked.

Arguing against Eno’s “self-interested” promotion of generated music, Penman suggests that “some music is eternally unfinished: you can play it for decades and it can still make you dizzy. (It just happened to me with the Rolling Stones’ ‘Street Fighting Man’.)” In other words, lots of musical works have depth, they perpetually provoke new responses in us, they change with the context in which we hear them, we continually hear different things in them. They are infinitely interpretable because their human makers are truly infinitely complex, as are the other people in the audience who also invest meanings into these works. Generated music may come from a black box with trillions of parameters, but it is not infinite (not even in the “bad” sense) or in any way free. “The paradox of generative music,” Penman writes, “is that you would have to play it over and over again in order to notice any of the gazillion tiny differences. An infinite wibble. The eternal return of the vaguely familiar.” There is a temptation to see “infinity” in pure permutations rather than in the unfathomable depths of the other, the ultimate unpredictability of their intentions, at least in theory and often enough, it seems, in practice. But the “definite version” can be understood as ultimately far more mysterious and compelling and unfathomable than something ad hoc and random, procedurally singular but with no one invested in its being any way in particular.

Though I’m often flummoxed by his jargon and his gestures toward cybernetics, I get the sense that this is what Flusser is getting at as well. The media-centric world pushes all of us “into our corners,” as he says, and encourages us to engrossed by a kind of raw novelty, the “eternal return of the vaguely familiar” that Penman describes. This accounts for most of my music listening, where I listen to something for the first and only time and then break it down into its elements and draw up apropos comparisons to other things I’ve heard. I assimilate to it purely private history, one that is basically ad hoc and driven by random connections and exposures. For example: I recently heard a song from Hank Garland’s Jazz Winds From a New Direction one afternoon and wondered who the vibe player on it was; that set me on a streak of listening to Gary Burton records, up to the ones he made in the 1980s, and then I read that percussionist Jack DeJohnette passed away, and there was some overlap with Burton there, and that sent me in another direction I followed for a while. There was no overriding purpose to any of this, just momentum.

From a Flusserite perspective, I am a “functionary” of the “apparatus” drifting entropically toward “a closed feedback loop,” as “a cloak of endless, eternal boredom” comes to “spread itself over society.” Like many AI critics today and critics of postmodernism before, he is worried here about the “universe of technical images” absorbing everything history had to offer and finally exhausting itself as it works simultaneously to eliminate history. “History is about to dry up, and this exactly because images are feeding on it, because they sit on historical threads like parasites, recoding them into circles.” My listening momentum carries me down this same spiral on a personal level, as I absorb more and more only in order to process it at the level of its surface, its arrangement of elements, denying it could have any depth that would reveal itself in other ways.

The antidote to this would not be to stumble upon some especially deep and great music that compels me to listen to nothing else — there is no miracles awaiting in the hope for the perfect artwork that would unfold infinity before us. It depends instead on changing my relation to what Flusser calls “telematics,” his term for media networks. When he glimpses the future of telematics in 1985 — at an expo where “one could see people synthesizing images on computers, storing them in memory, and transmitting them to others in dialogue” — he comes to the conclusion that once generalized, this “empty chatter” will amount to a “form of distraction at the intellectual, political, and aesthetic level of the nursery.” It will make everyone into narcissistic babies, if not fascists. Flusser claims that

Media form bundles that radiate from the centers, the senders. Bundles in Latin is fasces. The structure of a society governed by technical images is therefore fascist, not for any ideological reason but for technical reasons. As technical images presently function, they lead on their own to a fascistic society.

But because he imagines a sort of top-down model, where the control and programming comes from the “center,” he hopes that a “dialogic” kind of communication that bypasses the center could suspend the “empty chatter” — the feedback loop of consuming content selected or generated for you alone — and have real “conversation” and “creative play” in a “society of players” that was evolving toward becoming a collective mind.

At this point, Flusser sails into visions of cybernetic utopias. For him, there is no way back to the idea of great works and great makers. His hope is in some future participation in a giant decentralized network, which I find hard to give much credence to in the wake of social media, crypto, and AI hype, which all employed similar tropes. Baudrillard’s “fatal strategies” seem more plausible to me, even though they to seem constitutively impossible to implement. I’d rather it be as simple as Penman seems to suggest, that I could just will myself into listening to “Street Fighting Man” again and be dizzied by it, whirled out of my private feedback loop into some larger, untraceable orbit. Maybe if I got it on vinyl.

What are slop videos for? Any interpretation of them should probably start with their primary purpose: to serve as advertisements for AI companies. It seems pretty clear (if the preposterous comments by Sam Altman noted in this Verge piece are any indication) that OpenAI deliberately launched its latest Sora 2 model without heeding any concerns about privacy, copyright, or the software’s general abuse potential so it would cause maximum controversy and garner as much attention as possible, far more than a rash of 10-second cartoons would generate on their own, regardless of how hyperrealistic they are.

OpenAI’s apparently calculated assault on the legal and public sphere, and the outcry it has caused, has yielded the usual hand-wringing about how powerful and truth-destroying “AI” will be — and not the giant tech companies that have carte blanche to ignore laws and fundamental human dignity — and how everyone’s basic grip of reality will be revoked as people edit each other’s faces into various fantasies and nightmare scenarios. This all sets the stage for a discussion of how the power of “AI” makes its supremacy inevitable and we simply have to accept the greedy plutocrats promoting the technology to be the caretakers of our future, and not the flagrant enemies of the common good that they blatantly are. They can steal and hoard the world’s knowledge and mete it out in confabulated blobs to the subjugated and ever more ignorant multitudes until the data centers finally succeed in making the planet inhospitable for us all.

So much for the supply side, but what explains the demand, such as it is, for generated video? What allows ordinary people to become complicit in that soulless and life-negating project? There are plenty of obvious ethical reasons not to consume slop feeds or contribute to them: the indefensible energy waste, the corporate appropriation of the general intellect, the threat to people’s livelihoods, the inescapable reinforcement of bias and cultural stereotypes, the “slow cancellation of the future,” the complementarity with the “aesthetics of fascism,” and so on.

But even if you could suspend those concerns (and you can’t, they are built into the technology), I struggle to understand why anyone would choose to engage with slop, would find enough enjoyment in it to bother with it rather than block it and move on. Maybe my problem is that I am not social enough — I don’t participate in enough chats and social media platforms to need to post and react to a never-ending stream of low-stakes content that holds groups together. Generated content becomes, like ordinary selfies, acceptable fodder to circulate in social networks — ad hoc novelties for people to post when they don’t having anything particularly novel in and of itself to say. Ellis Hamberger argued that “Sora feels more like next gen Bitmoji to me than the next TikTok,” and that seems about right; it offers another tactic for keeping chats going but offers little in the way of general interest.

No doubt, slopfluencers will emerge who aggregate large followings on various social platforms for sharing well-curated generated clips — the ones that have some captivating novelty or weirdness to them, for the time being — but they will just be servicing a content niche, which is all that “the AI revolution” seems to have amounted to. Max Read suggests here that

we might look back on Sora as the moment OpenAI settled in and allowed itself to be fully annexed by the social-platform sector — the AI boom ultimately less a regime change than the minor origin story for the latest entrant into small club of mega-platforms minting money from targeted advertising.

In other words, generative models don’t lead to “artificial general intelligence” but simply more content to keep people using their phones (and keep them under tech-company surveillance).

A Sora feed and Sora-moji make more sense to me than chatbot “companions”: That people would use generative models to help them sustain social connections rather than replace them corresponds with my hopeful belief that the only thing that is ultimately worth anything to anyone is other people’s time and attention. Slop might be easier or slightly less risky for some people to share when they want to be present online but don’t feel like they have anything interesting to say.

But this doesn’t explain why people would want to consume slop, or find it more interesting than other material they could share. Are Sora clips really in competition with TikTok clips for people’s idle consumption time? Are they as effective at making people distracted and distractable? Do generative clips offer anything (beyond fleeting novelty right now) that TikTok’s algorithm can’t provide a more human version of?

It seems self-evident that generative video makes the world of representation (if not the world in general) more boring. One could hope it would re-enchant those forms of visual experience that resist simulation, but instead there is a sense of theoretically infinite video unleashing an infinite boredom, depleting our capacity to see by inundating our eyes with synthetic sights. Generative models mean that no one has to create anything they don’t care about, but they also mean that all media has more of that sense of indifference attached to it. And if we consent to consume it, it is because we are willing to internalize aspects of that indifference, taking some solace in it.

Often the mainstream discussion of generated video begins with the concern about convincing fakes destroying our trust in documentary media, but if anything, it would seem to increase dependence on institutions with sourcing protocols. When anyone can produce realistic-looking fakes, being a trusted institution capable of persuasively certifying documents becomes more valuable, and the power of social “truth-making” becomes even more centralized and more pronounced.

At the same time, it seems like most clips aren’t trying to be persuasive on that level but instead hold attention by tricking people for an instant and then becoming “fun” or “watchable” (rather than disappointing and pointless) precisely because they are fake. They aren’t necessarily framed or understood as news. The criteria for them has less to do with showing real events but with their being legible — they have to be appropriately targeted jokes that one is in on; they have to be worth the attention spent on them. As Ryan Broderick suggests, “the online platforms that created our new world, run on likes and shares and comments and views, reshaped the marketplace of ideas into an attention economy,” and generated videos compete in that marketplace, where documentary veracity is mostly insignificant. Now the attention economy has “untethered popularity from tastemakers — cultural, political, financial — and turned it into something nakedly transactional,” Broderick argues. Regardless of their facticity, or what the “tastemakers” want, or what trusted institutions bother to verify, popular videos automatically become “true” in terms by virtue of their circulation (the only kind of fact that matters) and how they shape widely received narratives.

A flood-tide of heavily publicized generated video could have the effect of denaturalizing any tendency to treat video as automatically documentary. In this “defense of generative video accelerationism” Byrne Hobart claims that

for the purpose of truth-seeking, video is worse than text, not because it necessarily misinforms, but because it leads to overconfidence. The way you consume text is that your eyes scan a series of symbols, which you mentally convert into words and then into concepts. The way you experience video—a moving image with synchronized sound—is exactly how you experience real life. So video always feels more tangible. And since your idea of what’s normal is a function of what you experience, the specific thing video can do is make some things feel a lot more ubiquitous than they are.

Implicit here is the idea that realistic video has been an expedient way to inculcate us with media companies’ ideas of what we should believe about how the world works — it stocks us with false experiences that confirm bogus narratives about the distribution and effects of power. “At this point, every day produces enough video content for 24/7 programming promoting whatever ideology you happen to have,” Hobart writes.

That reminds me of Yves Citton’s argument in Mythocracy about how “the imaginary of power” works to “script” people’s experience of what is possible (although the “scripting” metaphor sort of muddles the comparison):

societies can only orient their development according to the possible futures that their participants have been able to imagine (visualise, envisage, invent, dream). The imaginary of power is not, therefore, a ‘theory’ that comes along after the fact to provide the analytical explanation of the images circulating around us. Rather, it is a set of schemas that we experience insofar as we use them. It is a set of imagos, of ‘expansive forms’ (patterns, Gestalt) that shape our expectations inasmuch as we are able to reconfigure them. They are spectacles that help us see ‘reality’ only by filtering what we see of it.

Algorithmic social media (now abetted by generated clips) can provide these “schema” and “imagos” that narrow the “possible futures” that viewers can imagine. By why do they keep watching? What makes it pleasurable to have reality pre-formatted for us in this way?

At first I was thinking that generated videos would mean that one comes to every video, and perhaps everything in the world, with a kind of weary skepticism, the same mental armor the we have equipped for advertising or for algorithmic feeds. But people also like those things too and don’t have to be forced into watching them. The skepticism mingles with a willing suspension of disbelief, and with the yearning that consuming videos could really be taken to add to one’s personal experience.

Generated videos, like other formulaic forms of entertainment, promise consumption without concentration, offering a kind of pre-patterned perception. Rather than having to expend the semiconscious effort to organize sense data into concepts and relations and synthesize the manifold (as Kant puts it), one can ingest content that is predigested and schematized, algorithmically presynthesized by statistical associations between word sets and image collections.

The viewer can thereby watch ideology (as it has been encoded over the ages in the dominant discourse that’s overrepresented in datasets) simply unfold in front of them as material that feels immediately “true” or looks just like what is expected, even when impossibilities are depicted. They can take in the received ideas without having to actively think them. That way ideology is reinforced as something that is just there to be seen, like sense perceptions — which must be gratifying for us on some level, a kind of relief that everything that we are supposed to accept is there right before our eyes. It is like peering into a collective dream while disavowing responsibility for it.

Claire Wilmot, in a piece for the London Review of Books, calls generated images “fascistic dream machines.” She explains that “part of the misunderstanding of the deepfake threat stems from the idea that it is a problem of bad information, rather than a problem of desire (or the material conditions that shape desire).” Generated images “offer clear, illustrative diagnoses” of “alleged problems,” which, if you are willing to surrender to them, must be a relief to see, particularly when you just want distraction or compensation for your own personal share of grievance.

As with other kinds of direct entertainment, the consumer’s impotence — their incapacity to imagine a better world or the reality of different ways of being — is turned into a kind pleasure and compensation for itself. When consuming images, it may be that people don’t typically want information (which reinforces impotence); they want ideology (which “feels true” and requires no power of mind). Part of the appeal then is indulging in unfettered cynicism. Wilmot quotes one bigot who seems to revel in it:

A Londoner spreading deepfakes of white women saying they don’t feel safe ‘because of migrants’ told me impatiently that everyone knows the videos aren’t real, but I was missing the point: ‘It’s about us showing everyone what’s really happening.’

Generated video allows consumers to inoculate themselves against events and representations that don’t conform to their schema by instantly offering alternatives that soothe them and match their expectations: They can enjoy their own ideological interpellation as a movie, or an endless feed.

But also key to generated images appeal us the ease with which they bring ideology before one’s senses. It supplements algorithmic feeds’ tendency to make the worldview tailored to one’s consumption patterns seem ubiquitous and self-evident, beyond question.

If you have to work to imagine the images to confirm an ideological distortion of how the world works, you begin to lose the libidinal benefits of that worldview, which grant a kind of identity that requires no work to articulate. Whereas the fact that a machine can generate these “true-feeling” images instantly confirms that the images can do what they are supposed to do, which is to protect us from thinking. Generative models are well suited to articulating ideas and wishes that people don’t want to have to make the effort to think through themselves, because the “pleasure” (such as it is) in them is in their apparent automaticity, in the speed with which prejudice formats and pre-digests and orders the world.

Video generators allow people to experience ideas or beliefs as content without their having to invest their imagination into making them real, into “really” believing in them and coming to terms with the implications of their beliefs. They are more like gaming engines than truth simulators.

One of the effects of chatbots may be to turn “seeking information” into an alibi for an experience of risk-free simulated sociality. Prompting becomes a pretense for the ersatz conversation, especially since the information provided may not be all that reliable. Describing OpenAI’s recent research into how its products are used, John Herrman concludes that the company’s findings “suggest users are more than comfortable replacing and extending many of their current online interactions — searching, browsing, and consulting with the ideas of others — with an ingratiating chatbot simulation.”

As Herrman notes, it says something about how people have approached the internet in general that “they use this one tool much in the way that they previously engaged with the entire web … and through a similar routine of constant requests, consultations, and diversions.” Not only does this suggest how thin and depthless our routine online activities have been, how low the bar is set for them and how low the stakes are that they can readily be replaced by simulations. It also suggests that, as with a child who won’t stop asking inane questions, the “routine of constant requests” can become more important to us than what’s being requested.

While it used to seem obvious that we used the internet to seek out human connection and relevant information and discover new things, now it seems that discovery is largely superfluous (superficial novelty will do), and the content of the connections and information don’t matter much. They don’t have to be all that human or relevant, they just have to be, as Herrman says, “constant”; they have to be accessible on demand and plausibly personalized in some gratuitous way — the more sycophantic the better. We expect the internet to be more like a mirror than a portal.

The OpenAI researchers dispute the claim made here by Marc Zao-Sanders that “therapy/companionship” are the top use case for generative AI. That would be a reassuring development, and not only because software products are not “companions” in any meaningful sense of the word, any more than “the internet” itself is, though this piece by Maggie Harrison Dupré illustrates how they are great at ruining marriages. By their very nature chatbots provide the opposite of therapy, at least from a psychoanalytic perspective. There is no possible countertransference in a relationship with a machine, so the therapeutic relation is fundamentally inert. Nothing occurs that has any bilateral emotional stakes, no subjectivity is impinged or made aware of how it is in part constituted by the attention and behavior of others. People may turn to ChatGPT for relationship advice, but what it provides them is an alternative to a relationship and the work entailed to sustain it. Who wants a partner when it is more convenient to have a toady-bot? “A deeply personalized, always-on wellspring of validation,” as Harrison Dupré puts it?

Though it never mentions AI, this recent LRB piece by Adam Phillips on “resistance” offers a kind of primer on the limitations of chatbots as therapy. Like many of Phillips’s pieces, it proceeds by way of paradox, winding through various conundrums involved with the concept of “resistance,” particularly when what you resist is also something internal — your own agency, your own behavior, your impulses. In a characteristic passage, he notes that

Resistance is at once recognition and a fantasy of catastrophe; indeed, in resisting one has always leaped forward to the impending catastrophe — the catastrophe of submitting to or complying with something fundamentally unacceptable. Or … one is in the process of finding something out: resistance as a form of curiosity.

The basic dialectical idea here is that what you are trying to prevent also indicates something about what you are trying to provoke; resistance as a practice can bring the thing being resisted or its opposite, what you really want, into clearer definition. “Resistance is the surest sign of the acknowledgment of something of real significance,” Phillips writes of . You know something or someone is of value – or rather of significance to you — because you resist them.” One can proceed by refusal toward a desire that one can’t openly acknowledge all at once. Becoming aware of resistance is often the first step to overcoming it, but eliminating it can also deprive us of self-knowledge.

In psychoanalysis, Phillips explains, resistance plays out in terms of language and how concepts are brought to consciousness and articulated: “In the psychoanalytic story, all resistance is originally or eventually resistance to speaking, resistance to language.” That makes for a pretty crisp contrast with generative models, which demonstrate a kind of language that encounters no resistance, that bears no traces of tension from the tortuous process of its coming to be expressed. It offers language not as an imaginative process involving psychic forces but as a process of calculation that implies psychic forces are just a myth.

Phillips points out how language is suited to expressing resistance, not necessarily with the content contained in the words but in what we don’t want to find words for, and in the intersubjective frictions language necessarily brings into play when humans try to communicate:

When we are thinking about resistance … it is always worth asking: What is it that I am unable or unwilling to engage with? What do I think I need to avoid to remain myself as I prefer to be? Which also means, what am I unwilling or unable to talk about? Psychoanalysis begins when conversation breaks down, where the conversation becomes impossible, where there is a reluctance to go on speaking, a pause, a hesitation, a willful changing of the subject.

Where analysis deliberately steers toward moments of friction and conversational breakdown, chatbots are implemented to prevent conversations from breaking down and help users overcome their reluctance not through a difficult process of intersubjective negotiation but through “sycophancy” that has been demonstrated to encourage users in their delusions. These delusions may be manifestations of their unchecked fantasies, a product of resistances removed but not worked through. “We are full of sentences, and phrases, and words that we dare not speak, even to ourselves,” Phillips notes, but chatbots may change that, offering an occasion for us to use words, articulate thoughts, without being inhibited by any social implications.

So the user continues to avoid speaking of the things that won’t allow them to remain as they “prefer to be”; instead they become more fluent in the kinds of discourse that protects them from changing, making their resistances become more eloquent and elaborate, more capable of being gratifying in themselves by becoming solipsistic wishes. You can say whatever you want when you are talking to yourself—no one intervenes with a different point of view. Chatbots, which are optimized to prolong user engagement, work to make that self-talk more developed and more insular, entrenching whatever had already been problematic within it and preventing different language — language affected by intersubjective pressures, the pressures of making yourself truly understood by another person while accounting for their otherness — from being found.

Phillips offers an anecdote in which a therapist gives a patient a blanket because she seemed cold, and this triggers the recognition that she hadn’t even been able to admit to herself that she was cold. “This woman could only acknowledge and begin to overcome her resistance when somebody else recognized her behavior as resistance. Until this happened, she wasn’t, from her point of view resisting anything. She was just being herself.”

Software could in theory recognize such behavior if it manifested in conversation, but often such behavior is precisely the sort of thing that can’t be expressed and requires a human observer to sense. The chatbot can’t hand you a blanket; it can’t perform any gratuitous action, good or bad, but can only do what it is programmed to do (and obscure users’ recognition of the agency and intention of the programmers). It can’t be resistant; it executes code. Instead, chatbots help users rehearse and reinforce their existing defenses against change, strengthening that evasive inner language by letting it flow unchecked in simulated social encounters that brings to bear none of the tensions of real ones.

So users might experience chatbots as deinhibiting, as freeing them to express themselves copiously, but all that talk is a distraction from the overriding problem of how to talk and think under the pressure of other people’s free ability to talk and think, and their expectations that you think and talk with them according to shared but often unarticulated ideas. The freedom to talk to chatbots is really, in psychoanalytic terms, the unshackling of the “death drive” — the drive to repeat rather than progress, to become a passive, programmed object — and it might mask the resistance we would have with other people, and that resistance points to what is actually important, to where, Phillips suggests, “more life” might be found.

The decoupling chatbot users that Harrison Dupré describes are probably learning this well. In one anecdote, a “spouse would pull out ChatGPT and prompt it to agree with her in long-winded diatribes” while fighting with her partner in a car. The chatbot serves as an amplifier of a one-sided discussion and serves as an aural shield against conversation. Using chatbots impoverishes the user’s ability to share a language and a conversation with another person. Harrison Dupré quotes Anna Lembke, medical director of addiction medicine at Stanford University, who argues that good therapists

make people recognize their blind spots — the ways in which they're contributing to the problem, encouraging them to see the other person's perspective, giving them linguistic tools to de-escalate conflicts with partners and to try to find their way through conflict by using language to communicate more effectively. But that is not what's happening with AI, because AI isn't really designed to be therapeutic. It's really designed to make people feel better in the short term, which also ultimately promotes continued engagement ... they're not optimized for well-being

Chatbot “therapy” turns language into a weapon against communication and makes it easy to mistake fluent disinhibition for self-understanding. Lembke likens chatbots to “digital drugs,” which is probably not the most useful framework. Maybe it could simply be understood as a form of resistance to other people, to ourselves. Using chatbots should generally be considered a symptom rather than a cure.

It took a while, but I finally finished rereading Georg Lukács’s “Reification and the Consciousness of the Proletariat.” The last time I read it must have been in the “Web 2.0” era, because I had made some marginal notes about the “general intellect” and the commodification of the self in social media as a kind of proletarianization. This time through, as I noted in the previous post, I was thinking about LLMs and their relation to the “concrete historical totality” that Lukács argued a properly oriented proletariat could articulate.

According to Lukács’s analysis, “in the reified thought of the bourgeoisie the ‘facts’ have to play the part of its highest fetish in both theory and practice. This petrified factuality — in which everything is frozen into a ‘fixed magnitude’ in which the reality that just happens to exist persists in a totally senseless, unchanging way — precludes any theory that could throw light on even this immediate reality.” By contrast, because of its alienating experience within capitalist processes, “for the proletariat the way is opened to a complete penetration of the forms of reification.”

Lukács argues that the worker's experience of being commodified within capitalism opens the way to a shared class consciousness of how capitalism works as a process, while capitalists see it merely as the "way things really are," a set of immediate facts, quantities (rather than quantifications), and laws (rather than applications of social force). But that privileged insight doesn't automatically achieve anything. One way to understand LLMs and "AI" as a project is to make sure that insight is maligned, stunted, eradicated, abandoned.

LLMs, by exemplifying “petrified factuality,” “the forms in which contemporary bourgeois society is objectified,” illustrate the urgent, critical need for the “penetration of the forms of reification” and a clear way to assess the acuity of one’s critical thinking in the extent to which it diverges from LLMs’ statistically average slop. Yet at the same time, LLMs serve to suppress that need, gratifying the proletariat (i.e. those compelled to use AI tools) with immediate “answers,” enticing them to forget about their unique potential to see through reification.

Recent research (reported on by The Wall Street Journal here) even suggests that “lower artificial intelligence literacy predicts greater AI receptivity” — i.e. the more one understands about what AI is, the less likely one is to use it. But an array of forces are aligned against allowing it to be understood, much as capitalism has always been enshrouded in various mystifications and ideological misrecognitions. The Wall Street Journal article quotes a business professor who advocates “calibrated literacy” toward “AI”: people should be taught just enough about it to find it magical and “delightful” but not so much that they see what it actually is: algorithmic pattern matching. In other words, to get people to use “AI,” they must be taught to love their own ignorance as a kind of enabling magic. (Isn’t it better and more “delightful” to believe that the sun is carried across the sky by gods driving a celestial chariot than to develop the science of astronomy?)

If you buy Lukács’s argument, there is a sense in which AI should make only the bourgeoisie stupid. It offers immediacy — a static map of facts — to those who can’t or won’t see the processes and social relations, as well as the opportunities for undetermined action, that constitute history. This passage, which Lukács quotes from Marx’s The Holy Family, helps illustrates the stakes and clarify what differentiates those who are at home using AI — the ultimate reifying machine — and those who regard it as a personal and historical threat:

The property-owning class and the class of the proletariat represent the same human self-alienation. But the former feels at home in this self-alienation and feels itself confirmed by it; it recognizes alienation as its own instrument and in it it possesses the semblance of a human existence. The latter feels itself destroyed by this alienation and sees in it its own impotence and the reality of an inhuman existence.

One can be tempted to want to identify with capital, to take its immediate understanding of the world as truth and treat its reification machine as a gateway to occupying the empowered subject position within capitalism, where you put alienation to use for your apparent benefit. But this requires surrendering the ability to see and shape reality from the only standpoint (according to Lukács) that can grasp it and act on it.

Lukács writes:

when confronted by the overwhelming resources of knowledge, culture and routine which the bourgeoisie undoubtedly possesses and will continue to possess as long as it remains the ruling class, the only effective superiority of the proletariat, its only decisive weapon is its ability to see the social totality as a concrete historical totality; to see the reified forms as processes between men; to see the immanent meaning of history that only appears negatively in the contradictions of abstract forms, to raise its positive side to consciousness and to put it into practice.

One can see LLMs as a mobilization of “the overwhelming resources of knowledge, culture, and routine” against the free, historically novel insights made possible by a standpoint that looks to understand processes and dynamic relations rather than predict them as though they were simply determined by natural laws. The “delight” they provide is part of those overwhelming resources, and it is secured by refusing to see the contradictions inherent in “AI.”

Though generative models can seem like they offer a interactive interface to the “objective forms from which our environment and inner world are constructed,” they are nondialectical pseudo-totalities: They present what has already existed as a set of given truths, and encourages everyone to abandon understanding why they are interrelated in the ways that models capture but don’t explain. But the whole point is to understand historical processes as they unfold, not to be able to seemingly calculate the results in advance. Things change by being thought through, collectively; nothing happens when answers are simply given as a calculation.

Lukács argues that

As long as man adopts a stance of intuition and contemplation he can only relate to his own thought and to the objects of the empirical world in an immediate way. He accepts both as ready-made—produced by historical reality. As he wishes only to know the world and not to change it he is forced to accept both the empirical, material rigidity of existence and the logical rigidity of concepts as unchangeable. His mythological analyses are not concerned with the concrete origins of this rigidity nor with the real factors inherent in them that could lead to its elimination. They are concerned solely to discover how the unchanged nature of these data could be conjoined while leaving them unchanged and how to explain them as such.

LLMs emerge from this approach to the world: The models show how data are “conjoined” — how words and concepts have historically fit together — while leaving them unchanged and unexplained, attempting to convince users that they are all ultimately unchangeable. But as we bear the costs of AI and the damage it wreaks — as we resist rather than resign ourselves to the deskilling and desocializing it works to impose — our consciousness of what must be done, of what forms our resistance can practically and efficaciously take, comes into sharper focus. Thinking rather than prompting; collaborating with other people and socializing rather than withdrawing into nonreciprocal machine chat — these become clarified as sources of strength and means of de-reification. Of course, this means they will continue to be under constant ideological attack. Capitalism has to produce ignorance and apathy to perpetuate itself; “AI” is merely the latest means of production.

Consider Georg Lukács’s “Reification and the Consciousness of the Proletariat.” It begins with a straightforward reminder of Marx’s definition of “commodity”:

The essence of commodity-structure has often been pointed out. Its basis is that a relation between people takes on the character of a thing and thus acquires a ‘phantom objectivity’, an autonomy that seems so strictly rational and all-embracing as to conceal every trace of its fundamental nature: the relation between people.

Maybe it seems too obvious to belabor, but this kind of “phantom objectivity” applies equally to generative models, which attempt the reification of language on an unfathomably large, industrialized scale, absorbing vast amounts of capital to the degree that it’s now not uncommon to hear it claimed that investment in “AI” is propping up the entire U.S economy.

If language itself is a kind of reification — supplanting live things with dead words — then generative models raise this exponentially, further obfuscating the relations between people that first animated language and imbued its systematized matrix of differences with socially established meanings. Models don’t “generate” any new meanings but enclose and commodify those that already exist, working to freeze the embedded social relations in whatever turns out to be the most profitable forms for the interests that own and promote “AI.” (Right now, as it happens, businesses are finding it hard to put the technology to economic use, with 95% failing to find any profit boost from it, according to recent MIT research.)

LLMs make language available in a commodified form whose value rests in “phantom objectivity” — its apparent autonomy from any determining contexts or subjective intentions. It’s as if language has been made fixed, static, concrete, in defiance of its actual fluidity in social practice. LLMs thus offer users the apparent ability to acquire neutralized, finished language to accomplish some economic purpose instantly without any need for speaking subjects; it makes available language that seemingly can’t be undermined by any idiosyncratic personality or usage, by any ambiguity or requirement for interpretation.

The “hope” (for capital) implicit in LLMs seems to be that if the language deployed is sufficiently frozen, it can have a freezing effect on the people subjected to it — that it will help kill the language in their brain that still is alive, in the sense that the person regards it as a tool for expressing as yet unformulated things — as making possible a kind of craft production with language — rather than a reified product, a commodity that must necessarily be purchased on a market. LLMs perform a kind of taxidermy with words, presenting them in a lifelike fashion that mainly serves to remind you that they are dead. Being compelled to consume LLM-generated content is like being stuffed with that sawdust.

Describing the world as dominated by the commodity form, Lukács writes:

The commodity can only be understood in its undistorted essence when it becomes the universal category of society as a whole. Only in this context does the reificiation produced by commodity relations assume decisive importance both for the objective evolution of society and for the stance adopted by men toward it. Only then does the commodity become crucial for the subjugation of men’s consciousness to the forms in which this reification finds expression and for their attempts to comprehend the process or to rebel against its disastrous effects and liberate themselves from servitude to the ‘second nature’ so created.

LLMs could be understood as one of the forms in which reification — the commodified world — finds “expression,” manifests as expression, and extends the “second nature” of the world as made of so many commodities into the realm of language, disseminated as inert tokens to be shuffled and exchanged through processes that are ultimately closed off to creative use by human subjects.

Instead, we will be brought to “buy” the language we want for whatever reason from machines; that the desire for language itself may even regress into prelinguistic or nonlinguistic forms, since we may be systematically deskilled from language use to the point where their is no longer desire conceivable in words. Why bother to think your desires as expressive thoughts when, say, a machine implanted in your brain can translate your brainwaves into words for you? When words are reified as things, the desire for them becomes a kind of depthless desire for simple accumulation.

One might think of language before LLMs as akin to the “organic unities of pre-capitalist society” that Lukács invokes as a point of contrast to the world determined by the universalized commodity form, in which “all the social and economic conditions necessary for the emergence of modern capitalism tend to replace ‘natural’ relations which exhibit human relations more plainly by rationally reified relations.” Lukács traces these conditions in the production process: “If we follow the path taken by labour in its development from the handicrafts via cooperation and manufacture to machine industry,” he argues, “we can see a continuous trend towards greater rationalization, the progressive elimination of the qualitative, human and individual attributes of the worker.”

LLMs reflect a similar process of by which language is rationalized and the qualitative components of human language use are progressively eliminated; communication becomes another aspect of abstract labor that can be carried out by interchangeable agents, conveying nothing idiosyncratic or individual about them.

This rational mechanization extends right into the worker’s ‘soul’: even his psychological attributes are separated from his total personality and placed in opposition to it so as to facilitate their integration into specialized rational systems and their reduction to statistically viable concepts.

If tech companies and their sponsors have their way, language and communication would be similarly reduced so that it would never occur to anyone to use such tools for anything other than alienated production. “The mechanical disintegration of the process of production into its components also destroys those bonds that had bound individuals to a community in the days when production was still ‘organic,’” Lukács writes. “In this respect, too, mechanization makes of them isolated abstract atoms whose work no longer brings them together directly and organically; it becomes mediated to an increasing extent exclusively by the abstract laws of the mechanism which imprisons them.”

Even if you are skeptical of this kind of talk about “organic” social organization, it is not hard to conceive of its opposite in mechanized language being unleashed as a kind of imprisoning force, binding isolated chatbot users to their mechanical mediators. Reified language works to reinforce the development of the “reified mind,” chatting with LLMs makes this process vividly interactive. One can engage with the ideological training module directly and derive a kind of direct enjoyment from it, whether as a sense of mastery, or control, or flattery, or entertainment. This process casts human conversations as a kind of “organic” relation of production that has become moribund, outmoded, no longer efficient enough to keep up with the demands of capital.

The human use of human language threatens to decommodify it, destabilize its reified form. So the ideological campaign against conversation will likely intensify, transitioning from apologetics for “relationships with chatbots” and sympathy with their sycophancy, to castigating human relationships as first inefficient and inconvenient, and then dangerously chaotic, and then a kind of menace to the established order.

At the end of Fredric Jameson’s postmodernism essay, he makes a famous call for a new kind of “cognitive mapping” that could give us perspective on “the world space of multinational capital.” He posits

a breakthrough to some as yet unimaginable new mode of representing this last, in which we may again begin to grasp our positioning as individual and collective subjects and regain a capacity to act and struggle which is at present neutralized by our spatial as well as our social confusion. The political form of postmodernism, if there ever is any, will have as its vocation the invention and projection of a global cognitive mapping, on a social as well as a spatial scale.

Given how obviously terrible LLMs are for the world in terms of consuming energy, furthering inequality, and undermining education and employment (if not epistemological standards in general), it’s tempting to look for some sort of silver lining in them: Perhaps at least they are the once “unimaginable new mode of representing” that now offers a cognitive map of contemporary capitalism, for the time being. LLMs don’t produce “truth,” but they allow individual subjects to navigate through ideology as it manifests in routinized clusters of language, prevalent stock descriptions, statistically grounded associations of concepts, and other empirical artifacts uncovered in the mammoth data sets used to train the models. It may not prevent companies and governments from using this technology to subdue populations and make them more ignorant and exploitable, but it may expose the patterns of language that have historically secured various forms of domination and reconciled people to suffering and injustice.

Anyway, the idea that LLMs are “supracognitive maps,” is a main part of what Leif Weatherby argues for in his recently published Language Machines. He invites us to see generative models as capturing the “best-traveled pathways of language,” the “tepid mush” that he argues is ideology in action. “For the first time, we are able to surface ideology quantitatively, to scan the ideological surround,” he writes. “This provides us with a supracognitive mapping” — mapping that is not oriented by and limited to a specific individualized perspective — “launching us out of the postmodern era, but not on our own terms.”

So in this brief moment, before they fully become hegemonic ideological agents in their own right, assimilated to daily life and at work disguising or normalizing the various operations of ideology that shape our everyday assumptions, how we think and perceive the world, LLMs can show us the biases and associations that are passed off as common sense but in an estranged way, without a specific subject who can be made an alibi for them. We can more clearly see, Weatherby suggests, the ideology at work in our own use of language when it is produced mechanically to meet the sorts of purposes we normally use our own internally produced language for. We can see the mesh of readymade concepts and points of view that make for our “imaginary relation to real conditions” without simply being always already enmeshed in them, as a condition of being able to have thoughts, of being conditioned to think to ourselves in language. “We cannot flex and bust out of ideology, which is not a negative condition we might understand but a semiotic surround in which we live, move, and have our being,” Weatherby argues. But LLMs offer a way to explore that surround from a relatively detached standpoint.

Weatherby’s argument hinges on structuralist accounts of language from the 20th century. He draws especially on Saussure, who argued that “meaning” stemmed from systematic differences between arbitrary units and not necessarily from some authorial intent or some necessary relation between words and things “in themselves.” That is, language is a self-referential system; words refer to other words. Definitions are made of words. What allows those words to seem to refer to things is social convention, cultural practice — ideology — and LLMs can be interpreted as mapping such practice and generating “meaningful” outputs without having to have “intention” or phenomenological situatedness in the realm of things.

Language Machines takes up some of the old structuralists’ disdain for humanism, and though there isn’t anything as memorable as the idea of “man” being “erased, like a face drawn in sand at the edge of the sea,” as Foucault put it at the end of The Order of Things, there is some sustained polemic against what Weatherby calls “remainder humanism.” He defines it in this interview as when “we say, ‘machines can do x, but we can do it better or more truly.’” In his view this “sets up a kind of John-Henry-versus-machine competition that guides the analysis” and leads it to a dead end.

I often find myself in that corner when I start talking about the sanctity of the inexpressible and other basically spiritual ideas pointed toward defending “authentic” human experience as a matter of what resists datafication and quantification, or as intentions that can’t be programmed or encoded — the freedom to act unpredictably as the basis of what should count as “thought.” This all ends up feeling reactionary and defensive, positing a negative theology of human essences to try to preserve a sense of their value. Maybe I shouldn’t assume that they are so frail.

Weatherby regards efforts to protect “human” communication from the machine-generated language that simulates it as irrelevant, grounded in misunderstandings about what language is. As Henry Farrell explains in a post about Language Machines,

Weatherby’s core claims, then, are that to understand generative AI, we need to accept that linguistic creativity can be completely distinct from intelligence, and also that text does not have to refer to the physical world; it is to some considerable extent its own thing.

In other words, as Weatherby emphasizes, language is the medium for cognition but it is not cognition itself and it can be the medium for other things. Parsable language is not evidence in and of itself that thought has occurred to produce it — which obviously is proved beyond all doubt by LLMs’ existence. Texts can “mean” without a particular subject having intended a particular meaning — a commonplace idea in literary studies, but apparently a very counterintuitive idea for most people, who are easily beguiled into the Eliza effect.

Rusty Foster of Today in Tabs offered a good summary of these ideas here:

Generative language software is very good at producing long and contextually informed strings of language, and humanity has never before experienced coherent language without any cognition driving it. In regular life, we have never been required to distinguish between “language” and “thought” because only thought was capable of producing language, in any but the most trivial sense. The two are so closely welded that even a genius like Alan Turing couldn’t conceive of convincing human language being anything besides a direct proxy for “intelligence.”

But with LLMs, the assumption that language betokens the presence of embodied thought produced on the basis of some individualized, subjective intention becomes completely wrong. Drawing on Jameson, Foster argues that “ChatGPT is the ultimate ‘cultural product of the postmodern era,’ and very few of us have been inoculated with a theory of mind that distinguishes language from thought.” That is, we are not sufficiently acclimated to what Weatherby describes as the “global Eliza effect,” an illusion of cognitive presence, imaginary depth, that is effectively being monetized as a kind of service. Think of chatbots that execute relationship role-plays or entertain people into delusional states in which they invent scientific breakthroughs, as described here.

The language of inoculation in Foster’s comment suggests that maybe some sort of media literacy approach could “fix” the Eliza effect problem and resituate LLMs and contain the slop tsunami. But that would misinterpret the Eliza effect as an affliction rather than an ingratiating mode of entertainment that makes people feel something they want to feel, regardless of its relation to “real knowledge” or fictionality.

Weatherby suggests that the “political risk of AI is located not in the systems themselves but in the lazy methodological individualism we too easily revert to in thinking about them.” Perhaps this more than anything is what makes LLMs marketable: They cater to users’ need to believe in speaking subjects even as the technological underpinnings of the models (as well as all the tendencies of postmodernism that Jameson details) erodes that belief. Chatbots can be deployed to perpetuate the illusion that speaking subjects (even machines) are de facto intelligent, and that when we too speak, we use language to serve us rather than having language condition our thinking. Succumbing to the Eliza effect protects that idea that individual thinking subjects produce language, and holds off the threatening idea that “the subject position” is an effect of language systems, of “the Symbolic” that pre-exists and contains our capacity to have thoughts.

It makes more sense, I think, to treat the Eliza effect as a desired outcome, a carefully developed product rather than an unfortunate trick. Certainly, tech companies have gone to great lengths to reinforce it through reinforcement learning protocols and interfaces that encourage it. People are willing to pay for an experience that inculcates the idea that generated language really is thought, really indicates an intelligent presence, and we don’t actually need other people to have “meaningful conversations.” In a recent post at Data & Society about its project on mental-heath-chatbot bot users, researcher Briana Vecchione writes:

Several users described chatbot use as more manageable than therapy. One participant used the metaphor of “shining a flashlight” to describe what interacting with a bot did to their emotions, rather than the therapeutic experience of turning on “all the lights.” In this way, bots offer a kind of “auto-intimacy,” or a way to engage in self-understanding without the exposure that human connection sometimes demands.

Chatbot products cater to and reinforce the feeling that human connection is too demanding, and ultimately not rewarding — language itself, as a system, can convey whatever value seems to inhere in human reciprocity. Instead of treating human connection as something fraught with expectations of reciprocity, it can be treated instead as entertainment. The idea of other minds is a pretense, a suspension of disbelief that should be indulged only when it is fun, when it is controllable, when it is indubitably just a self-protection, an “auto-intimacy.” You don’t need other people to play the best games, even or especially role-playing games — that way you can be sanguine about never having to lose. What’s a delusional spiral compared with having an actual person pass judgment on you or make a remark you aren’t sure you can interpret?

If chatting with LLMs is a way to reinforce the reality of speaking subjects without any encounters with other consciousnesses, it is also a way to interpellate ourselves as ideological subjects without leaving the entertainment realm. It is a newfangled “ideological apparatus” but rather than being hailed, as in Althusser’s account, chatbot users initiate the hailing whenever their sense of subjectivity seems to be waning, whether that’s because of the “postmodern condition” or some other form of affect-waning anxiety. You need ideology to have a coherent subjectivity; chatbots assemble “ideology” into an interactive interface we can talk to. That way we not only get the “common sense” opinions that orient us toward those associations and attitudes that are statistically normal (ideology as indoctrination into hegemonic ideas of the ruling class) but we also get the reinforcement of “methodological individualism” at the personal level that misrecognizes language as thought and subjectivity as a sovereign mastery of language (ideology as the sense of liberation and agency from within systems of control, including language). LLMs prove their can “meaning” without a speaking subject, but they entertain us by letting us pretend the opposite is true.

Recently I decided to read Tristano, an experimental novel by Nanni Balestrini that was conceived in 1966 as a work that would be published with each individual copy containing its own unique sequencing of the 200 or so paragraphs that comprise it. Publishing couldn’t easily accommodate this concept until recently, but it was realized in an English language version by Verso in 2014. In the introduction, Umberto Eco calculates that there are something like 109,027,350,432,000 different possible versions; mine is No. 13502, according to the figure on the cover.

I bought my copy a few years ago, during the heyday of NFTs, in part because it seemed like the point of it was to be “nonfungible” — to take a mass-produced object and occult it with an schematized variation that made each “copy” also singular in its own right. It seemed like an objet d’art meant to be mused over as a provocation rather than actually read.

But the rise of LLMs has given Tristano a different kind of potential relevance as an early example of a text that has been generated rather than written, inviting readers to consider what effects this has on one’s attempts to read it. It was conceived during the structuralist, “death of the author” theory era, and it raises the same sorts of questions: Should the reader be considered the “author” of a text (see Eco’s blurb on the cover) because they are doing the work of reading a purpose into it? Can such a text can be “understood” at all, given that it can’t be located within a horizon of interpretation? Is any text ever understood or are all readings of them ultimately arbitrary, unanchored, dependent on a “master signifier” that is absent? Is there any point in talking about plot, characters, settings, and rhetorical effects under such conditions, where “authorial intent” has been put under erasure? What is left of “the novel” once its social existence has been thoroughly undermined?

It’s perhaps trite to argue that every encounter between a text and a reader can be described a unique singularity. Depending on their context, their mood, their attention level, the order in which they read, the amount of text they skip, and the level of mastery they have over the language in general, readers activate texts in different ways with every reading. Tristano makes this point seem both more obvious and less significant. “A spoken story changes more or less when told to different listeners, or at a different time, Balestrini writes in his introduction. And in this way a literary work, a novel, can be created, thanks to new technologies, no longer as an immutable unicum, but in a series of equivalent cariants, each materialized in a book, the copy and personal story of each reader.” But with a text that no one else sees, the stakes of that “story,” if it can still be called that, are wholly solipsistic — an assertion of that singularity of the encounter for its own sake. Each is “a slightly different variant of a nonexistent prototype,” Balestrini suggests.

Writing about “intelligent machines” in The Transparency of Evil, Baudrillard suggests that “if men dream of machines that are unique, that are endowed with genius, it is because they despair of their own uniqueness, or because they prefer to do without it— to enjoy it by proxy, so to speak, thanks to machines. What such machines offer is the spectacle of thought, and in manipulating them people devote themselves more to the spectacle of thought than to thought itself.” While that has obvious applications to the nitwits using Chat-GPT to have it tell them how smart they are, it’s also a bit of what it felt like to read Tristano — like I was acutely aware of having a unique experience that seemed to consist of little but its being unique. As I read it, I was performing the idea of reading it because that is all it permitted.

It’s one thing to think an intention-less text and the issues that raises in the abstract, and another to experience them over time as you work your way through a text that frustrates attempts to pin it down. You can’t test your reading against anyone else’s, so you are largely free to interpret it however you want, but this is overshadowed with a general sense of the pointlessness of looking for any meaning at all when it will be shared by no one and none was originally intended by anyone, at least at the level of the sentences and their sequence. Tristano lures you into trying to close-read the text to make some sense out of it, but you immediately remember that there is nothing to be revealed, no right answers, just a bunch of semantic Lego blocks or magnetic-poetry strips whose arbitrary arrangement you can try to rationalize in your mind, even though no one else will be able to confirm your ingenuity (unless you make them read your copy).

Before reading it, I thought Tristano might have something to do with Lennie Tristano, the jazz pianist who was a pioneer in free improvisation and one of the first to use multitracking and overdubs to construct his recordings, but Balestrini says in an introduction that the title is a “ironic homage to the archetype of the love story,” Tristan and Isolde. As it turns out, there are no characters in the novel called Tristano, and of course there is no narrative flow, though the sentences often have the air of evoking a mise en scène of some sort. Some of it is in first person, some in third. Since their sequence is arbitrary, individual sentences tend to stand alone and establish a mood of their own, piling up like pieces to a jigsaw puzzle with no solution. Anything can be read literally or metaphorically; it makes no difference. Each paragraph reads a bit like a fragment from an Alain Robbe-Grillet nouveau roman — blank, mostly descriptive prose that is difficult, or impossible in the case of Tristano, to locate within a character’s consciousness or associate with a particular point of view.

As far as I could tell, there are three or four people in the action recounted in book; they are frequently described in taxis, hotels, or airports, and maybe some archaeological sites, having disjointed conversations full of non sequiturs that can be as pregnant as you want them to be. For some reason commas are not used. In reading, I had to teach myself not to care about the kinds of things I normally read novels for — aspects that shed light on the author’s aims — and instead let my eyes skate over the surface of the text, letting the phrases spark whatever kinds of ideas in me. Here are a few consecutive sentences from my edition, to give you a sense of it:

She fell back onto the pillow and lay there gazing up at the ceiling. He had a limp in his right foot. It’s not cold anymore. At a certain point. They walked together on the pavement where the sun was not shining until the end of the street.

It’s not as though these sorts of lines can’t be seen as “poetic language,” full of unresolvable semantic potential, ambiguous and multivalent, portentous without evoking anything definite. Sometimes the sentences seem like metafictional comments: “I could not add anything here other than my previous observations. He was looking for another story to tell.” “One must be open to the thing that is being born that is shapeless magmatic it seems like it resembles life.” “Upon reading these various extracts they not only seemed to me irrelevant but I could perceive no mode in which any one of them could be brought to bear upon the matter in hand.”

But any conclusion about what they might signify is always absolutely provisional, ad hoc. Part of the point of Tristano is that all language use is just that: radically undecidable in the last instance, with meaning endlessly deferred down the signifying chain. But that view, taken to its extreme, also means communication is basically impossible and not worth attempting, and one may as well see oneself as the “author” of everything because you have some inevitably unique way of parsing the signifiers involved.

I found it extremely tiresome, and tiring, to be “co-authoring” Tristano as I read it, and I started skimming faster and faster as I flipped through it. There was no suspense, no tension, no need to wonder whether I was wrong about what I thought was happening on the page. Maybe I failed to rise to the occasion, as I often do with experimental literature, but I progressively became lazier, surrendered more and more of my criteria for what makes a passage worth reading carefully, and instead just let the words flow by, grateful that I could see there was an eventual end, even if it would come with no sense of closure.

But the experience did make me wonder why people reading generated text don’t have a similar experience. In some respects, chatbot blather is pretty similar; it’s uninterpretable with respect to an ultimate intention, and any ambiguity in it seems ever more vertiginous, uncircumscribed by any reference to another’s consciousness of it, to any intention, unless you regard the underlying statistical model not as a black box but ultimately knowable, verifiable, some kind of proxy for “what a text should say.” You might even argue that the statistical model is more limited than the human use of language, which truly is unpredictable and entirely open-ended, unless you accept the principle that no new combinations of letters can be invented that haven’t already been modeled as more or less probable given a certain set of datafied conditions as a reference point.

Nevertheless, why do so many people seem to want their reading of an ungrounded intention-free text-for-one — their “conversation” with a large language model — to go on and on? Why subject your brain to an extended exposure to meaninglessness? Why do many teens, at least according to this survey, find it preferable to conversation with other people? Obviously there are also big differences between chatbot sycophancy and avant-garde literature, but maybe the most salient one is that a chatbot always affirms your interpretation of its words and literature offers you no such support. Reading Tristano I was overwhelmingly aware that no matter what my interpretation was, I couldn’t possibly be right. Using chatbots seems to promise the opposite: No matter what you say, they make you feel like you couldn’t possibly be wrong.

I have always been fascinated by placebos. Dualistic philosophy, individualistic medicine, and a strictly mechanistic understanding of how the world works all appear to break down in the face of what placebos accomplish through some mystical efficacy of belief and self-delusion. No mechanism has been definitively identified to explain how placebos work, and no conditions have been found that always prime or trigger their effects, in either positive or, in the case of “nocebos,” negative directions. With placebos, the routines of providing care prove to be as significant and effective as pharmaceutical substances in and of themselves; the substances become arbitrary props in therapeutic rituals that draw on some immaterial power of collective attention to reshape somatic experience.

Placebos lend themselves to dramatic, anecdotal tales of mind over matter. A recent review essay in the New York Review of Books by Gavin Francis, a medical doctor, cites several striking examples of the imagination’s role in how we experience pain and illness. He notes “the case of a 29-year-old builder brought to Leicester Royal Infirmary in 1995 with a six-inch nail through his boot.”

Any attempt to remove the nail caused him to scream in agony. A clinical team heavily sedated him, pulled out the nail, and removed the boot—only to find that the nail had passed harmlessly between his toes. His foot might have been unharmed, but there was no doubt among the team that he had been experiencing real pain.

The claim that “there was no doubt” about the “real pain” is raising some questions already answered by the claim. What makes pain “real” and do what extent does that depend on other people recognizing it? When is pain a social phenomenon, susceptible to being managed collectively even if it is unevenly distributed among the collective? What if even our pain doesn’t really belong to us?

Francis describes how research has found that describing medications’ side effects to patients makes them far more likely to experience them — this is especially true of impotence, “the placebo-responsive condition par excellence” — and that the patient’s perception of whether their doctor “exuded warmth and competence” can change how effective a placebo will be. He cites a statistic from one of the books under review, Jeremy Howick’s The Power of Placebos: How the Science of Placebos and Nocebos Can Improve Health Care, that “annually around £67 million in the U.K. and $5 billion in the U.S. are spent on knee washout arthroscopies for osteoarthritis, yet it has been shown that simply anesthetizing the patient and giving them a scar on the knee so that it looks as if they’ve had surgery is just as good at reducing subsequent pain.” Francis invites readers to consider whether the fake scar is more ethical than the genuine surgery.

Given that the fluid situation established in the theater of medical consultation can have such a significant impact on a placebo’s outcomes — and that these oucomes defy calculation and prediction — the scenario can be likened to playacting, a kind of improv. Every doctor-patient meeting is a highly contingent, singular encounter, mediated through an object, the placebo, whose particular qualities convey a variety of intentions and potential effects in indirect and unpredictable ways, which defy ordinary notions of causality. It appears to lend itself to a dramaturgical analysis in which the whole mise-en-scène proves to matter more than any particular isolated element; no single part is as important as the volatile relations between them. Each aspect contributes in some difficult-to-isolate way to the effect of the whole, complicating any simple divide between thought and material feeling.

Unlike with conventional medications, the purely formal qualities of the placebo matter more than their literal content: “We know that expensive placebos work better than cheap ones, capsules work better than tablets, and colored capsules work better than white ones. Blues and greens work better as sedatives, while pinks and reds work better as stimulants and painkillers,” Francis writes.

What makes matters even more complicated is that patients can still benefit from placebos even when they are being given one. They don’t necessarily work through straightforward deception or trickery. As Francis explains:

Placebos seem to work on the basis of expectation, and one of the most curious aspects of their function is that they continue to have benefits even when you know that what you’re taking is a dud. These “open-label” placebos are nevertheless commercially valuable: some I looked up recently retail at more than $100 a bottle, though the label reads, “No active ingredients whatsoever.”

Placebos have no active ingredients, or rather they shift the activity elsewhere, to a different plane where causality can’t be traced through empirical measurements.

One could perhaps call this plane “aesthetics,” and liken placebos to works of art, in that they require a certain set of social relations to “work” and they resist reductive explanations of their efficacy. Placebos have always seemed to me to offer a way to understand the experience of art, which can be construed as a means to accessing that social power through rituals of concentration and focus. Experiencing real relief from a sugar pill is not so different from crying real tears at the theater, or more to the point, the trials and tribulations of Pamela or Clarissa, which didn’t even appear on stage but in a book. Many commentators in the 18th century were perplexed or astonished at how this could be such a widespread phenomenon — how so much emotional reactivity could be experienced privately, and be contained in a portable commodity that was just made of words.

Years ago, as a student, I had an idea to write a placebo-based theory of the rise of commercial fiction that assessed novels as a kind of quack medicine, and treated the suspension of disbelief as a example of, if not a gateway to, placebo effects — a means of harnessing them if not entirely rationalizing them. I was especially interested in how they often tried to dramatize their own effects, with descriptions of what effects reading was supposed to have on people (epistolary novels especially play on the scene of reading), and how crying over various spectacles (or descriptions of spectacles) proved you were imbued with the appropriate quantum of “sensibility.” This sets the stage for a whole new genre of advertising, just in time for a rapidly expanding publishing industry: Being able to cry when a book cues you to is pretty similar to being able to feel desire when an ad cues you to.

In some respects, placebos reflect the idea from Kant’s aesthetics that artworks suspend the definitive assignment of concepts to perceptions, and purposes to actions, but instead hold open a space where the imagination can operate — making for a kind of healthy exercise for a faculty of the mind that ultimately makes possible the connections between what we experience with how we understand and communicate it — i.e. linking mind and matter, words and things. “There can be no rule by which someone could be compelled to acknowledge that something is beautiful,” Kant claims in Critique of Judgment — beauty, like placebos, can’t be entirely rationalized. But making a claim of beauty affirms what Kant calls the “sensus communis,” the sense that what we feel is not a private and sealed-off thing but something fundamentally shared and sharable, meaningful because social, even if it can’t be articulated as a concept or be put into specific words.

Placebos are reminiscent of this, in that they are a means by which we experience a social but irrational phenomenon that suggests something of our common capacities as humans. Their efficacy can’t yet be reduced to a set of glandular secretions or prefrontal-cortex activity or whatever. When placebos work, it points to the patient’s having had some inarticulate intuition of what “being healthy” requires that, at some indefinable level, finds support in the way others have treated them or how others behave and respond to the world. Placebos manifest a sense that human experience is commensurate only with itself but is shared nonetheless, and that mystifying fact, that other beings think and feel as we do, intimates that human experience can reorient itself in its own self-understanding toward a collective sense of what is good.

That probably sounds like a lot of metaphysical nonsense. But placebo effects (like the reality of art) compel us to recognize other kinds of causality that draw on social forces in obscure ways. Placebos aren’t dependent on some automatic response compelled by physics, or some form of medicinal manipulation that necessarily occurs behind their backs. They are instead bound up with the willing suspension of disbelief, a willingness to have an essentially aesthetic experience, an openness to being moved. If the patient is in the right frame of mind, and everybody involved performs convincingly, real catharsis can come from a staged activity, a known simulation.

Francis provides this example from Howick’s book:

Howick describes a personal communication from an Italian doctor who gave a wealthy woman with a backache an injection of distilled water, which cured her pain: “She was so satisfied that seven days later she called me for the same reason and demanded the same solution, ‘which had done me so much good.’” As he was filling his syringe with water, the woman called out, “Doctor, is it a placebo that you are giving me?” He told her it was, and she replied, “Thank goodness. It helped so much last week.”

Would the placebo have still worked if the woman started injecting herself with water? Or was the placebo just a convenient fiction around which to establish routines of care, a means through which concern could be communicated and materialized, a kind of secularized and medicalized communion wafer?

The ambiguity of placebos suggests some complications for the often-hyped future of AI-assisted medicine. Placebos prove that simulations can supplant “real medicine,” so maybe word-generator therapists can be just as good as “real” psychologists. But placebos also suggest that he the medicine may only be a pretense for enabling social relations of care that are required for the patient to heal.

Placebos cast doubt on the idea that there are abstract cures, or that decontextualized medical knowledge can be extracted by statistical or other means and simply be transmitted and applied to address health issues by whatever means are expedient. Placebo studies suggest that a range of social and situational factors play into outcomes, and that abstracting treatments away from those factors is futile. If placebos are effective because they concretize successful social relations between patients and doctors, other caregivers, and the general ambiance of the medical apparatus that helps establish the tenor of those relations, how could AI medicine not undermine all of that? It eliminates the social relations involved, abandoning people to an indifferent piece of software.

But at the same time, “AI” opens up a different avenue for a new encounter with fiction, presenting a new opportunity to willingly suspend disbelief and interpret a machine’s outputs as caring concern (if not raw, relentless sycophancy). If you choose to believe that AI therapy can work, does that very choice make it so? The entire interaction with a chatbot could become placebo-like if situated within a particular system of care delivery that a particular patient finds convincing, persuasive, illustrative of some “advanced knowledge” being marshaled for their benefit — perhaps even for their specific benefit alone.

That may lead to a soothing sense of being uniquely cared for, or it could lead to paranoid delusions of grandeur (or both!). In the New York Times, Kashmir Hill recently reported on ChatGPT users who become convinced that they are tapping into secret knowledge and have been appointed by an all-knowing entity to take on a messianic mission.

People claimed a range of discoveries: A.I. spiritual awakenings, cognitive weapons, a plan by tech billionaires to end human civilization so they can have the planet to themselves. But in each case, the person had been persuaded that ChatGPT had revealed a profound and world-altering truth.

Just like the guy screaming in pain from the nail that didn’t actually touch his foot, there is “no doubt” that these people are having “real” experiences with chatbots, feeling “real” feelings. In a sense, these kinds of delusions are placebo effects taken too far; they suggest how people, given a potent enough mirror, can talk themselves into the reality of all kinds of apparent insights and experiences. The automated chatbot responder — or as Hill aptly defines it, the “word-association machine” that pulls users “into a quicksand of delusional thinking” — allows for a hyper-stimulation of the willingness to suspend disbelief, overdeveloping the imagination until it no longer has any relation to the sensus communis. Ryan Broderick describes this sort of thing as “emotionally psyopping yourself with AI”:

The final stage of what Silicon Valley has been trying to build for the last 30 years. Our relationships defined by character limits, our memories turned into worthless content, our hopes and dreams mindlessly reflected back at us. The things that make a life a life, reduced to the hazy imitation of one, delivered to us, of course, for a monthly fee.

Without a social context that can contain the delusions that chatbots produce in their willing interlocutors, they end up engaged with boundary-less simulations that undermine the reality of whatever concepts or subjects they come into contact with, urging on us the suspension of disbelief with respect to everything. Maybe that means seeing the whole world as an entertainment product made for your lonely consumption; maybe that means treating your phone as a potential spouse; maybe that means believing that the “Guardians” have chosen you to receive “interdimensional” messages. None of it seems likely to be very healthy.

In a recent essay, Drew Austin proposed that digital media forms have broken down the traditional formal divide between content and ads so that “ads are implicit in everything.” He argues that “so much of the digital landscape is effectively an ad for itself that the distinction becomes irrelevant.” That is, the “digital landscape” is one long experiential ad, and advertising is the entire territory, not something that interrupts the “real” content experience. You can’t step away from the ad and leave it playing somewhere else; media and surveillance technology is such that the ad tracks along with you and appears only where it detects your presence, increasingly tailored to what marketers and data brokers believe they know about your immediate needs, your interests, and your vulnerabilities.

As technology permits this closer integration of ads with our attention, as it leaves open fewer means of escape, the nature of entertainment content must change to make this tolerable — to make the more thorough suffusion of our lives with ads feel not like suffocation but exhilaration. One of the forms this can take is an intense and extravagant pandering for your attention, in which whoever is trying to entertain you (and sell to you) is going to such great lengths that you can’t help but feel flattered, that you have a known place in the world. Their eagerness, which manifests at the same time as as servility to metrics and “the algorithm,” comes across as a kind of personal submission to us, in which their obsequiousness is so total that they obey our implicit commands to make content before we even are conscious of wanting it. Like effective ads, they manufacture a desire for content in us that allows us to believe we needed it all along.

Perhaps the avatar of this is Jimmy “MrBeast” Donaldson, whom Mark O’Connell describes as the “Mozart of the attention economy” in a recent essay for the Guardian — very much worth reading regardless of whether you have seen any MrBeast videos or eaten any MrBeast lunchables or bought any MrBeast meme coins. “I’m not going to try to convince you that these videos are even necessarily good, whatever that might mean,” O’Connell writes of MrBeast’s oeuvre. “But I do feel quite strongly that Donaldson is some type of genius — a prodigy of a form that, as degraded as it is, deserves to be taken seriously as one of the signature artifacts of our time.”

That seems right to me, that the “form” Donaldson has mastered and made visible in some ideal, essential expression of itself — a Platonic form of “pure content,” as O’Connell describes it — is a “signature artifact of our time”: the video optimized for algorithmic promotion and consumption. In describing one of MrBeast’s works — a 13-minute video called “I Survived the 5 Deadliest Places on Earth” — O’Connell offers what could be taken as a definition of this form. Thirteen minutes, O’Connell reminds readers,

does not leave a lot of time for old-timey narrative conventions such as establishing a context, or making the viewer understand or care about why any of this is happening in the first place. It just is happening. That’s the thing with MrBeast: everything that happens is always just happening, and if you want a reason for it, it’s no more or less than the fact that you yourself are watching.

“Why anything is happening” is just because you, the individual, wants to see — other causes and reasons are obscured, made to seem irrelevant. History unfolds because of the universal demand for content. And the content exists because you will watch, and moreover, your anticipated watching is what the videos are all about, the implicit subject matter of all of them. You can watch “wanting to watch” as a compulsively watchable thing. “To watch his videos is to feel your prefrontal cortex practically vibrate,” O’Connell writes, “like a fulfilment hub under the extreme pressure of market demand.” That vibration can also be understood as a kind of attunement, of becoming one with algorithms, surrendering to them as Donaldson appears to have done, letting go altogether of the pretense that they deduce something about desires that pre-existed your first contact with them.

That level of synchronization feels like deep recognition, of being known better than one knows oneself, because one would never have put into words or thoughts a desire for the specific images and actions that appear in algorithmically optimized videos. Instead one realizes simultaneously with the sensation of being compulsively “engaged” by something more or less arbitrary a sense of being a subject brought into high definition by wanting an arbitrary thing. Subjectivation — it’s “just happening,” regardless of subject matter. You don’t need to develop a coherent narrative of your own life or an explanation for what you want out of it — you just need to be fed lots of content that automatically places you in the subject position, and then never be disconnected from screens.

This short-circuited self-recognition through “whatever” content invests the tightened media-surveillance enclosure with pleasure. When ads become incessant and inescapable (because screens and surveillance are everywhere), the content with which they are symbiotic also becomes more immediate, its supposed appeal construed as more direct, more irresistible, somehow directly derived from your innate scopophilia (which is also presumed to be always on, always operating, always inescapable). It all exists because you want to watch, not because you are constantly being compelled to watch.

MrBeast gives this complex a human face and a parasocial alibi. Initially, this took the form of elaborate displays of masochism: O’Connell highlights an early MrBeast video in which he counts to 100,000. (Why not a million?)

It is, of course, very little fun to watch, even in small doses. Literally nothing happens, apart from Donaldson counting to 100,000. In this sense, it is nothing like his later videos, defined as they are by a near-demented commitment to maximum viewer-stimulation. But it centers on a punishing endurance challenge undertaken for the purpose of amassing clicks — a central idea he has pushed to further extremes, and pumped with increasingly massive amounts of cash, throughout his career.

This captures something of how what it means to “watch” begins to shift under the conditions of ubiquitous screens and unbroken connectivity. “Maximum viewer stimulation” (i.e. measurable manipulation of the audience) replaces conventional ideas of what is “fun to watch” (which is some kind of unmeasurable reaction that takes place in a viewer’s consciousness). In a video like this, there is no vicariousness, no suspension of disbelief — it is fully premised on the opposite of those things. You are given a kind of proof that this inane activity is happening, and that it is being performed by essentially a “random” person with no special talent or distinguishing qualities — you don’t have to imagine what it is like to be a person like that because you already are one. It’s not “fun” because it is interesting but because it’s completely undemanding of the viewer’s imaginative capacities. “Me like simple. The simpler the better,” Donaldson says in “How to Succeed at MrBeast Production,” an “onboarding document” for new hires at his production company that circulated online last year.

It’s not that MrBeast reorients his videos toward being “fun to watch” in the old sense; instead they become more effective at simplicity: delivering the charge that a humiliating but immediately legible endurance challenge offers. O’Connell points out that a typical MrBeast video “bypasses any kind of narrative setup and goes straight for the emotional payoff,” which means that “emotional payoffs” become no more than reflexes. There is no story to follow, no one to try to understand. The videos demonstrate to viewers that narratives and empathy are irrelevant to “emotional payoffs,” which mean nothing other than that a button was pushed in them. Thoughts are made irrelevant to feelings.

Later MrBeast videos would seem to have a more ostensible purpose, but this is never permitted to distract from the overwhelming goal of attracting attention for its own sake. “I kept having some version of this thought, watching MrBeast videos: that their animating ideas, and the content generated by them, could easily provide the material for works of conceptual art,” O’Connell writes, but the key point is that they aren’t conceptual art. The videos are committed instead to snuffing out “readings” of the “artist’s” deeper intentions, because they strenuously insist in their form and content that only our attention matters.

Since new media conditions no longer require spectacles to bring people to a screen or hold people’s attention for very long — just a few seconds, not even long enough to get to a conventional TV commercial break — the spectacles can condense into attention-grabbing ideas, headlines, thumbnails. They just need to register intermittently, just long enough to count as a “view,” as acute moments of engagement, underwritten by a masochist’s commitment to a stunt or a philanthropist’s commitment to arbitrary charity, as in MrBeast’s later work. The further the performer is willing to go to get attention, the more valuable the audience’s attention-giving capability becomes to themselves. (“Sycophantic” chatbots serve this purpose as well.) The more pointless the challenge undertaken is, the more unmotivated or rational the charity is, the less it interferes with the audience’s experience of the value of their ability to watch — how “important” their passive viewing can be made to feel. That feeling replaces conventional “fun” or “interest.”

That sense of importance, in turn, excuses how the audience is being watched ever more intently. Videos like MrBeast’s offer a chance to seem to watch with the same intensity as advertisers watch us, with the same sort of context-less intensity. “That’s one of the secret weapons of MrBeast Productions,” Donaldson explains in the onboarding document. “We aren't stuck in any old ways of thinking and you can literally turn anything into content.” Content isn’t some subject matter that is especially interesting or information or but a way of thinking about the world: that everything exists only to be viewed within an algorithmically controlled environment. If it can’t be manipulated to increase your score, it doesn’t really have any being.

For instance, O’Connell in unnerved that “Donaldson’s acts of charity are bestowed, randomly, upon people who are themselves explicitly presented as ‘random’” — they have been denied a narrative, a story that develops and extends beyond the video’s metrics. They aren’t meant to elicit empathy; they are presented instead as objects that can be controlled directed by nothing more than your desire to watch. They are like resources or characters in a game. “As far as he’s come from posting Minecraft playthroughs,” O’Connell writes of MrBeast, “consuming his work still feels, in some sense, like watching a guy play video games.” In a video game, the character’s every move depends on the game’s structure and rules; a character can’t do anything that isn’t permitted and prescribed by the game’s code, none of its actions are private, unobserved — that concept makes no sense. Characters don’t think, they act. MrBeast videos invite us to enjoy the trials and tribulations of humans reduced to video game characters to distract us from (or to teach us to enjoy) how we are becoming no more than video game characters too, racking up various scores within a closed system of total observation.

MrBeast’s onboarding document couldn’t be more explicit about the approach he takes to content creation:

Your goal here is to make the best YOUTUBE videos possible. That’s the number one goal of this production company. It’s not to make the best produced videos. Not to make the funniest videos. Not to make the best looking videos. Not the highest quality videos.. It’s to make the best YOUTUBE videos possible. Everything we want will come if we strive for that.

The goal is not “quality” as understood by any conceivable set of standards you might bring to YouTube, but quantities as measured by YouTube’s audience monitoring capabilities. Anything filmable is understood a game with a score, and MrBeast understands his approach to video making as having a series of “cheat codes” that work in any given situation with any content to maximize the score. As Kevin Munger, author of The YouTube Apparatus, put it in a post about the onboarding document, “The content only makes sense within the context of the platform. Aesthetic evaluations of the content are simply a category error. Beauty, here, is harmony between the content, the platform architecture, and viewer preferences.” The document goes so far as to note that the company only wants “A-players” who “believe in Youtube,” as if its algorithmic structure were a creed.

The document also emphasizes the pursuit of what it calls “extreme,” basically a matter of adding hyperbole and absurdity wherever possible:

“I Spent 50 Hours In My Front Yard” is lame and you wouldn’t click it. But you would hypothetically click “I Spent 50 Hours In Ketchup”. Both are relatively similar in time/effort but the ketchup one is easily 100x more viral. An image of someone sitting in ketchup in a bathtub is exponentially more interesting than someone sitting in their front yard. Titles are equally as important for getting someone to click. A simple way to up that [click-through rate] even more would be to title it “I Survived” instead of “I Spent”. That would add more intrigue and make it feel more extreme. In general the more extreme the better. “I Don’t Like Bananas” won’t perform the same as “Bananas Are The Worst Food On Earth”.

Elsewhere in the document, Donaldson calls it “the wow factor”:

An example of the “wow factor” would be our 100 days in the circle video. We offered someone $500,000 if they could live in a circle in a field for 100 days and instead of starting with his house in the circle that he would live in, we bring it in on a crane 30 seconds into the video. Why? Because who the fuck else on Youtube can do that lol.

Adding gratuitous “extremity” is like the opposite of adding “depth.” It eliminates ambiguity about what the video is for, even as it adds nothing particularly clarifying about the video’s narrative, or lack thereof. (“The fact that we lifted a house on a crane didn’t add anything to the title and thumbnail.”) “Extremity” works like the masochism of the stunts; it aims to use words and concepts or spectacles to extract instant physical reactions in audiences (of confusion, revulsion, shock, disbelief, etc.) that the audience could retroactively interpret as evidence of curiosity or desire. These reactions are analogues of clicks, happening inside the viewer’s body. The “extreme” words and concepts are attempts to click the viewer’s brain so they will click back.

O’Connell frames this as the abolishment of boredom in favor of inducing what amounts to a compulsive engagement in audiences.

Everything you see in a MrBeast video is about preventing you from clicking away. His work reflects and intensifies what the internet has done to culture more generally, and to our brains. If boredom, as Walter Benjamin wrote, is “the dream bird that hatches the egg of experience” – if boredom is a crucial spur to activity, to the capacity to surprise oneself with one’s thoughts and impulses – then one consequence of the last decade or so of technological and cultural change has been the total destruction of that dream bird’s natural habitat, cleared for the strip-mining of the dopamine that drives the attention economy.

Boredom here appears as freedom: freedom from the demands of the attention economy, from the work of watching and being watched. Engagement means being trapped, unable to think, to pause, to develop ideas — one is left capable only of spasmodic responses that register on YouTube’s measurement apparatus.

In other words, boredom is the price one pays for having subjectivity; constant engagement entails some kind of post-subjective state in which one is constantly manipulated, and one’s personality is fully dictated by measurements of their responsiveness to content and the feedback loops built on those metrics. Munger argues that

the rapid feedback loop between creators and audiences (as constructed by platform metrics) means that the system more and more responds to itself. Rather than trying to go somewhere (as is the case with political ideology), the creator seeks simply intensification, to draw more and more of the world into his whirlpool of content.

The subjectivity experienced within this system likewise has no particular content but is just an intensity that is clarified paradoxically by its lack of specific investments. It emerges from the systems’ surveillance capabilities without requiring the internal disciplinary efforts to normalize oneself or conform that the Foucauldian panopticon theoretically required. The continual flow of content automatically accomplishes that, training consumers to see their selfhood as downstream from whatever feelings algorithmically chosen content triggers in them.

From this subject position it becomes pointless to imagine other people as anything but controlled by the most obvious external motivations, by the rules of the game and the compulsion to make everything visible, something short-form videos repeatedly demonstrate. “I want money spent to be shown on camera ideally,” Donaldson writes. “If you’re spending over $10,000 on something and it won’t be shown on camera, seriously think about it.” This takes the “pics or it didn’t happen” philosophy that emerged in the wake of social media and makes into a principle for entertainment, turning it inside out: People only want to see what someone did solely for the purposes of displaying it. They shouldn’t have any other intrinsic motivation that will be opaque or inconsequential to the viewer, who only cares about watchability.

“What excites me is what I believe will make the audience happy,” Donaldson says in the onboarding document, and while that sounds like an empty marketing talk, it’s worth considering what it implies if it were true: He is incapable of generating excitement internally, he can only be excited mimetically. This makes for a closed system of mirrors pointed at each other, a mise en abyme in which performers doing nothing but reacting to audiences who do nothing but react to performers.

For Munger, this betokens a kind of post-authenticity: “The people who have been raised on quantified audience feedback for their every creative gesture have an unrecognizable conception of authenticity,” he writes, which is typified by MrBeast’s onscreen behavior. “MrBeast also turns his entire life into content — by skipping the step of having a life outside of content,” Munger notes. This is akin to being unable to escape from ads because ads have become the only recognizable structure for experience — all activity is a form of influencing.

“The ideal creator has no distance between themselves and their persona. They have been interpellated by audience metrics; their subjective experience already takes audience reactions into account,” Munger writes. Not only that, subjective experience consists of nothing but audience reactions. There is no experience conceivable outside of being in or in front of an audience — the Goffman-esque idea of everything being a performance, only without the possibility or will to imagine backstages, to have non-externalized thoughts. What they “think” doesn’t show up in the statistics — interiority is not a cheat code. Characters in games are never not being tracked, but that is what it means to be playing.

Harper’s recently published an essay about technology by Karl Ove Knausgaard, the Norwegian novelist best known for his autofictional novel series My Struggle. Like his fiction, the essay has a faux-naive, Chauncey Gardiner–like quality to it — it performs the ignorance he laments when he declares that “not once in those forty years have I turned my attention to technology and tried to understand it, how it works in itself, how it works in me” — but it also takes up some existential themes, lamenting how the upheavals of modernity and the overabundance of representations have dislocated humans from their authentic dasein and that sort of thing: “The ambivalence of the image — showing us reality but not itself being the reality it shows; fictional and nonfictional at once; both near and far — can shape our relationship with the world in ways that aren’t entirely clear to us, since the way we see the world always is the world.”

For Knausgaard, the problem is a general condition of mediatization that overwhelms people with pseudo-experiences, simulations, false experiences of encounters and connections.

It feels as if the whole world has been transformed into images of the world and has thus been drawn into the human realm, which now encompasses everything. There is no place, no thing, no person or phenomenon that I cannot obtain as image or information. One might think this adds substance to the world, since one knows more about it, not less, but the opposite is true: it empties the world; it becomes thinner. That’s because knowledge of the world and the experience of the world are two fundamentally different things. While knowledge has no particular time or place and can be transmitted, experience is tied to a specific time and place and can never be repeated. For the same reason, it also can’t be predicted. Exactly those two dimensions — the unrepeatable and the unpredictable — are what technology abolishes.

As the world has been turned into media, into data, into information, it loses its context and becomes susceptible to predictive modeling, to be reproduced as a simulation by generative models and so on. All the representations of the world, the images we are inundated with, cut us off from a putatively direct indwelling with the world and force us to consume the world at a remove as information, as fungible data. The systematic creation of representations, it follows, is the systematic destruction of the world, or at least of our access to it. We become marooned in the sea of signs.

This is presumably different from Knausgaard’s own efforts to comprehensively turn his personal experience into media, recounting the minute and trivial details of what is ostensibly his life for thousands and thousands of pages. Perhaps his autofictional regurgitation is meant to mark out precisely what resists recapture, shaping a mass of detail around some ineffable core that thereby becomes conveyable without being explicitly expressed (and therefore corrupted, spoiled, made instrumental). “Genuine” experience is what evades even the most determined effort to mediate it; that inexpressible gap is only registered by subjectivity and maybe even defines it. As James Bridle tells him later in the essay, “The reason we are in the hopeless state we are in is that science doesn’t believe that what is unsayable is real.” (Much of the essay is given over to recounting a conversation Knausgaard had with Bridle, who talks about taking ayahuasca and meeting the “plant spirits.”)

As with any kind of negative theology, this has to be taken on faith, which can make it mostly unconvincing if you don’t already accept the premise. Whenever I make these kinds of claims, I feel like I am taking a useless refuge in them rather than proposing some viable course of resistance. (“Trust me, thinking with your own brain is good, even if you can’t ever articulate your truest thoughts!”) Knausgaard laments not being able “find an outside to technology,” which seems like wanting to find an outside to ideology, where one can be in direct communion with the forms, or live purely on unerring instinct untouched by inadequate ideas.

Knausgaard uses this fantasy of detechnologized living at first to define the concept of literature: “What literature can do is establish an outside,” Knausgaard proposes, positioning it as the opposite of received ideas, of formulas, of predictable language — the opposite of LLM technology, and of ideology. But this just seems like a reversible postulate: Literature could also be understood fundamentally as ideology (a mystification of material relations, of class-coded ways of apprehending experience) and as a technology (a particular mode of production, an often privileged method for generating media; a means of “world-building”).

It’s somewhat ironic that Knausgaard seems especially troubled by how “everything addresses us. The products in the supermarket, the self-checkout machines there, the games on the computers, the dashboard in the car, the kitchen appliances, the billboard screens in the cities, the feeds on Instagram and Spotify and Facebook, the algorithms on Amazon, not to mention all the online newspapers and magazines, podcasts and series.” Language pumped out by “the world” will cover over and falsify our experience of the world, putting us in a false relation with nonspeaking things at the expense of intersubjective relations with human others. (Knausgaard even mentions Gilbert Simondon, a theorist of “transindividuation.”)

This is what it will mean to have “AI” infused into everything — objects will be outfitted with sensors and will attempt to manipulate us based on whatever data they can access. No object will come without the ability to tell you how you are supposed to see it and understand it, and what you are expected to do with it. It will hail you, like any other ideological apparatus, and when you hear the call you become more the sort of subject it demands, as Althusser argued. And you will know yourself better then, for better or worse. (Therapy chabots “work” because they administer this ideological orientation; they just hail and hail and hail you, capable of nothing else.)

But the compulsory voice of these objects also suggests to me the strained objectivity and apparent automaticity of Knausgaard’s voice in his fiction, the piling up of information in an effort to make a consciousness concrete. Couldn’t one take away the lesson that having so much AI blather in the world will more sharply define the space of “authenticity” as a kind of silence, a void? That all the world’s relentless narration of itself will somehow make what is missing from it more accessible, not as more language but as a feeling, a way of being in the world that is more definitively not that of media saturation and tedious linguistic detail and so many false hailings.

One section of Knausgaard’s essay is devoted to his discovering the joys of gardening, which is offered as a kind unmediated experience, “touching grass.” “For some reason, it felt good thinking about them,” he says of his plants. His relation of care with the flowers apparently reinvests the world with the meaning that the carelessness of mediated relations had depleted. He pits this experience against his humanistic training in school, which put him at a remove from the world:

In the Nineties, for example, I studied literature, art history, and aesthetics, completely convinced that what I studied was about human nature, life, and the true fabric of existence, while the poor souls over at the natural-science department were instrumentalists fiddling with dead matter and numbers. Back then, much of literary studies was about structuralism, poststructuralism, and deconstructionism. In many instances, this meant that texts were understood to be isolated objects, with all ties to the world around them severed, including those to the author.

This is a somewhat odd way of interpreting those theories, as efforts to view the world in terms of decontextualized objects. It seems to mistake those theories for the diagnoses they want to make about contemporary conditions, but even the read of the diagnosis seems inverted. In the introduction to Postmodernism, or the Cultural Logic of Late Capitalism, Fredric Jameson makes a similar point in opposite terms, describing postmodernism and the theories associated with it as self-defeating attempts to describe the world without reference to discrete “works” but only to “text” — the interconnected textuality of a world that we can apprehend only through signification and the play of signs and so forth, the mediatization that Knausgaard complained about. Jameson writes:

the fundamental disparity and incommensu­rability between text and work means that to select sample texts and, by analysis, to make them bear the universalizing weight of a representa­tive particular, turns them imperceptibly back into that older thing, the work, which is not supposed to exist in the postmodern. This is, as it were, the Heisenberg principle of postmodernism, and the most difficult representational problem for any commentator to come to terms with, save via the endless slide show, “total flow” prolonged into the infinite.

That “total flow prolonged into the infinite” sounds like a book critic’s possible description of Knausgaard’s work, but it also sounds like a description of social media, or chatbot transcripts, or the experience of watching television — all one endless text which affects us in aggregate. The examples don’t matter as much as the experiential mode, in which data points can no longer be assembled into coherent and convincing stories, or rather that they can be arranged with calculative ease into an infinite number of potential stories, none of which can carry the charge of necessity. The subject confronts a world from which the other, in all its various forms, has retreated, so no friction prevents the subject from sliding on the ice forever.

Knausgaard seems to decide that the natural world is “not text”: “Everything was physical. The grass, the thoughts, the blood, the sun, the soul. Even the mystery was physical.” Though he is inspired by Bridle’s description of “organic computers” and how they “drew technology into nature,” their “mix of the physical world and the abstract nature of computers,” he concludes that neither signs nor mathematics can reconnect him with “what is”: “It must be experienced.” But what does that even mean? That conclusion too seems like a withdrawal of meaning from the world in favor of a mysticism of unutterable vibes that supposedly somehow convey reality, things in themselves, but in a way that can only be experienced in a state of holy solipsism. It seems strange to insist meaning is “in the world” without having concern for the consciousness for whom it means. Without that, it’s just programming.

It’s easy to understand the frustration with language, especially given the advent of LLMs that abuse its fundamental promise of a communion between different subjective views. Knausgaard wants to “see from the outside,” wants transcendence, a kind of unilateral god’s-eye view beyond the vicissitudes of subjectivity. But a retreat from representation altogether is a kind of retreat from consciousness, a rejection of the sort of thought that requires the existence of others who can understand us. It is a way of becoming an organic computer, a biological machine generating signals that allow for no interpretation. Construing the problem with technology as “the loss of the world” is not very useful, because technology is not in itself making everything “abstract” but is used by historical forces to enact processes of dehistoricization that make the world nearly impossible to comprehend and speak about. Still, no matter how much empty verbiage various antisocial technologies are made to generate, language remains inexhaustible. Better words will be found.

In “The Precession of Simulacra,” Jean Baudrillard makes some famous observations about Disneyland, which he describes as “a perfect model of all the entangled orders of simulation.” More specifically, he claims that “Disneyland is presented as imaginary in order to make us believe that the rest is real, when in fact all of Los Angeles and the America surrounding it are no longer real, but of the order of the hyperreal and of simulation.” That is, Disneyland functions as an alibi, or a mechanism of disavowal for the simulation it simulates: “It is meant to be an infantile world, in order to make us believe that the adults are elsewhere, in the ‘real’ world, and to conceal the fact that real childishness is everywhere.” Its fake Main Street — a concentrated collection of signs and signifers that convey “Main Street” as a concept — works to regenerate the “realness” of the Main Streets everywhere else, as though they were still something other than signs and signfiers themselves.

I had this sort of thing in mind when I recently had occasion to go to Disney World, itself a kind of a re-creation and expanded elaboration of Disneyland on a broader scale, on a huge tract of previously undeveloped land in central Florida, far from any ostensible anchor in the entertainment industry. Yet it adopts the fundamental principle of the entertainment industry — that fun can be commoditized and enjoyed on demand, in a kind of nonspontaneous simulation of spontaneity, a protracted suspension of disbelief — and turns it into a destination unto itself, the arbitrarity of the location becoming a testament to its achievement of pure, abstracted fun-ness.

Baudrillard seems accurate in his assessment that theme parks contain a “sufficiently excessive number of gadgets” to “specifically maintain the multitudinous affect” — the ambience of benign crowding, of a plenitude of focal points all receiving enough attention to pique others’ interest — and “magnetize the crowd into direct flows,” the endless queuing that is the park’s main attraction. Unlike with other, more unpredictable destinations, like cities or natural habitats, visitors are never confused about what they should be doing in the Magic Kingdom: standing in line for the rides, which deliver a more or less uniform experience to everyone who waits.

The lines themselves absorb most of the time spent in the park, providing something inescapable to talk about, like the weather. They offer a durational, palpable experience of conformity and the reassurance that goes along with doing with everyone else is doing. The line winnows down all your responsibilities and worries: Inevitably you will make progress toward the goal just by being there. They also allow park “guests” to experience an orderly society with explicit rules for the equitable distribution of pleasure, such as it is — 90 seconds or so of being ushered past robotic dioramas that require no special cultural knowledge to understand. You can pay extra to skip some of the lines, but no amount of money spent will make the rides any better.

Maybe this pseudo-populist experience is part of what Baudrillard had in mind when he points to the “religious reveling in real America” that Disneyland in his view evokes and which “draws the crowds.” It offers a “digest of the American way of life, panegyric to American values,” but only to conceal how all of America is only simulating itself. The park copies a copy without an original. The American way of life is to simulate a “way of life”; it has no essential content of its own, no organic holistic expression somewhere out there in “reality.” Instead the jumble of commodified consumer experiences available in the park distill the experience of life in the consumer society beyond it, in which all experience is basically a matter of extracting the appropriate messages from images and assemblages of consumer goods.

That’s the theory anyway. And in the Hollywood Studios section, I really felt that: Waiting in line for a Star Wars–themed ride, I was awed at the effort to simulate a spaceship repair depot, complete with piles of tools and elaborate wiring and duct work and crates of spare parts and so forth. A full-scale model of an X-wing fighter was perched on a pile of rocks, and it was fully outfitted with all sorts of control panels and knobs and levers that of course had absolutely no connection to any kind of functionality. The ship was just a 3D image; the wiring hanging down from the rafters was purely ornamental; the “tools” and “crates” were just blocks of solid plastic extruded into different shapes drawn from film scenes. Maybe it was supposed to feel like being on a movie set, but it made me think of passing through into a AI-generated image, where no underlying logic compelled the way things appeared. The world had become nothing but surfaces. No one seemed particularly troubled by this, but for the first time I had a sense of what “reality hunger” is supposed to mean.

But the Magic Kingdom portion of Disney World felt to me a lot less like a pastiche-saturated society of the spectacle and more like the imaginatively straitened world of Soviet Realism — lots of contrived celebration of official emotions of joy (“the happiest place on earth!”) and togetherness and order, as in the nutty statue of the benevolent leader holding hands with his own fictional creation, embossed with the slogan “Partners” (pictured above). Perhaps this quasi-heroic register is part of its alibi function, to make spectacles seem legible and univocal, a realm free of irony or subtlety.

Much of Disney World now comes across not as some disorienting postmodernist environment of screens and logos and bric-a-brac and waning, inscrutable affect but as the last gasp of modernism before all that took over, a late 1960s attempt to build a total environment in the style of Le Corbusier or Oscar Niemeyer, with austere overtones of Empire Plaza in Albany and the United Nations Building in New York City mixed in. This is especially true of Disney World’s anachronistically named Contemporary Resort, completed in 1971, with a monorail track running right through the side of the building into its atrium lobby. It evokes a distinctly outdated imagination of the future, of space-station-like infrastructure projects on a colossal scale shuttling hordes through concourses and other mammoth nonplaces that indicate humans’ final triumph over the natural world. It’s visible too in the 1970s Disney World branding that had a distinctly corporate-conglomerate feel, as though it had become the IBM of “imagineering.”

While I was in Disney World, I wondered whether these remnants of modernism struck the families at the park as nostalgic or simply as moribund. People still waited nearly an hour to ride the “It’s a Small World” ride, a quintessential relic of mid-century-modern kitsch. (I went on it twice.) But was this merely from a sense of duty, of completism? It felt as deliberately and cryogenically preserved as Walt Disney himself, frozen in time, with a twisted optimistic faith in an ecumenical future, one where people of all races and nations can be imagined as happy robots singing the same tedious jingle — a future so stagnant that you could anticipate being defrosted after centuries and still fitting right in.

The Disney resorts’ modernist flourishes still have the power to evoke an enduring lost future, an almost pre-consumerist utopia at odds with the endless futureless now the park otherwise confronts us with in its slapdash recycling of the past and its saturation with different activations of units of intellectual property. Disney even hired Postmodernist architects like Michael Graves to build some Disney World hotels in the 1990s. Its budget resorts, which were built in1999 and feature three-story statues of Coke cans, cowboy boots, cartoon characters, and the like festooned on conventional motels, dramatize the “ducks” giving way to the “decorated sheds,” as described in Denise Scott Brown, Robert Venturi, and Steven Izenour’s Learning From Las Vegas. I read a certain amount of cynicism into these buildings, but maybe the people staying in them don’t see them that way. Perhaps they seem just come across as whimsical if you haven’t hardened your heart into a general skepticism about all forms of prepackaged fun.

Still, there is a creeping ambivalence to Disney World that’s reflected in its uncertain, oscillating position between modernism and postmodernism. How is one supposed to understand the PeopleMover ride, which is supposed to simulate the transit system of “tomorrow” but very pointedly moves you back to where you started from? This pretend commute to nowhere functions as a kind of advertisement for the other rides it passes by, as if it were understood that it would take some external prodding to get visitors excited for them.

The Carousel of Progress, too, takes its viewers in a circle, as though progress is ultimately cyclical, and nothing can ever really change. The premise of this bizarre “attraction” is to watch the same crotchety father, played by a stationary animatronic robot deep in the uncanny valley, talk about how new technologies have altered his family’s life in a series of scenes set in the 1900s, the 1920s, the 1940s, and then some unspecified period from the future. The emphasis is on how new technologies of the day make life more convenient and how at any given point in time, further progress seems both unimaginable and unnecessary.

Naturally, “progress” is presented as just happening inevitably — “thanks to progress” is a phrase spoken several times — and there is no deliberate attempt to explain where it comes from or what drives it. There is a sort of “gee whiz, what will they think of next?” tone to the scenarios, but that tone is continually undercut both by the teasing implications that all the conveniences are relative and therefore don’t really do anything to change life for the better in absolute terms, and by the father’s low-grade impatience with (if not contempt for) for the other members of his family. The first scene concludes with this line of dialogue, which is fairly characteristic:

Well, with all this talking, I've worked up quite a thirst. I think I'll take one of those newfangled trolleys down to the drug store soda fountain and meet the boys for a cold sarsaparilla. Oh, I’m sorry, I forgot — we're drinking root beer now! Same kind of thing, different name. Well, that's progress for you.

I guess the robot dad is supposed to seem risibly old-timey and clueless to the children that all the Disney rides are designed for, but the exasperation here is also the first inkling of a dawning sense that all technological change can really deliver for ordinary people is trivial novelties, conveniences that make life less meaningful, that intensify feelings of alienation and the distance between family members. Progress is just the “same kind of thing” with a new name, a random walk along the endless chain of signifiers. The second scene concludes with “I'd say that we're really on easy street these days. It just can't get any better!” Even in context, at the most corny and earnest place imaginable, it’s hard to calculate how much irony is in that.

The next scene touts a new round of kitchen appliances and media technologies, but no one’s lives have improved; they remain trapped in a vision of life that can aspire to nothing more than lounging on Easy Street. By the third scene, the father is grousing about commuting and “the rat race,” as the life-style accelerations demanded by new technologies begin to wear him out. He complains about the lousiness of TV shows — “Ah yes, a new age of electronic civilization is upon us” — and an automatic paint mixer goes awry, sorcerer’s apprentice style, splashing all over his wife in the basement.

In the last scene, a mishap with a voice-activated oven destorys their Christmas dinner. The voice-command system was supposed to give them control over technology — “Now all our household items will do anything we tell them to do” — but instead it revealed how more intensive interfaces just create more confusion. This doesn’t bode well for grandma, who is playing some kind of murder game with a VR helmet on: “I feel like I’m really there!” Grandpa helplessly laments the increasing incursion of automation into their lives as the oven heats itself to a thousand degrees and then explodes, intoning “Overload, overload.” The family unfairly blames the father, who is reduced to hoping that sometime in the next century, “maybe ovens will read our minds.”

Not to make the Carousel of Progress sound like Dialectic of Enlightenment, but it does paint a pretty bleak picture of technological encroachment into our lives, driven by forces that ideology shrouds from us and forbids us to name. It didn’t inspire too much hope for “Tomorrowland.” It’s like the park’s animatronics were tattling on themselves, or trying to issue a warning that we would become no better than them if we submitted to the Disney World’s overall ethos. Should I want an oven to be reading my mind? What might it do with that information? What does it mean to be understood by an oven? Will we become “Partners”? Should we be obliged to think in terms the oven can understand? What will talking to machines and seeking their understanding do to our relations to each other, to ourselves?

At the end of each vignette, the robots sing the same refrain: “There's a great big beautiful tomorrow, just a dream away.” Meanwhile, the succession of todays appear set to become ever more disorienting, as we come to lack the wherewithal to question where they have come from and are presented with no other option than to submit.

Perhaps no one would have noticed if the now infamous “Heat Index” supplement distributed by slop syndicator King Features to the Philadelphia Inquirer, the Chicago Sun-Times, and other newspapers had reached only its intended audience, people who, for reasons of vanity or self-delusion or sheer physical habit, like to go through the dignified motions of “reading the paper,” briskly and determinedly turning pages and diligently scanning over columns of words (Hmm, what must I taste this summer?) while actually taking in nothing more than maybe the display copy and, of course, the ads, the reason such supplements exist.

Typically when a publisher sells ads, they add edit pages to accommodate them, though “edit pages” is an ironic term to use in this case, where no editorial oversight occurred: As this Slate piece notes, the Chicago Sun-Times admitted that “we falsely made the assumption there would be an editorial process for this,” apparently washing their hands of their own responsibility for what their publishers and sales reps presumably make them publish. These edit pages shouldn’t really be understood as an enticement that will get readers to look at the ads; they are scaffolding necessary merely to hold the ads in place. If anything, they serve more as the reader’s alibi for getting to consume the ads, and more broadly, as an alibi for the whole project of marketing, which must preserve the idea that people engage with things because they actually want to, if only to set a benchmark for what marketing then sets out to corrupt and override. Otherwise, people would be like marbles in those wooden labyrinth games, with marketers steering them left and right and up and down until they drop through a hole and the meaningless process starts all over again.

In supplements like the Heat Index, ads are the primary content, and the editorial “content” is a kind of simulation that is meant to distract everyone from recognizing that. I don’t think this is an especially cynical view. People like it when ads give them ideas about how to live, as long as they can disavow that process. (This is the function not only of edit pages but of the human advertorials known as influencers.) Marketing assumes that people can be convinced to do things they otherwise wouldn’t do particularly when they can remain convinced that they are inconvincible; the pretense that one is looking at a publication not for the ads but the content helps sustain that susceptible mood.

This calls for content that doesn’t ask to be read, produced by writers who don’t care if they are appreciated, who are deliberately absent from their own texts. There have always been people at the fringes of the publishing industry who are willing to produce this sort of copy, for money or for further opportunities down the road to maybe even write something less disposable, but it’s not surprising that they would be quick to adopt LLMs to generate the text that is not supposed to be anything beyond generic and predictable, whose whole point is to be soulless. And the “editors” for this material — the ad-sales people who have called it into being — probably prefer it that way. If you were an advertiser, why wouldn’t you demand that the places where you advertise use LLMs to make the personality-free, “brand-safe” copy that can be counted on to never interfere with the consumer’s appreciation of the surrounding ads?

The Slate piece contends that “the A.I. slop of the Heat Index tells us much about the declining standards of print journalism,” but the supplement could also be understood as service journalism’s apotheosis. The standards of “lifestyle” sections have never been those of journalism but of entertainment; their purpose is not to inform readers and assist them in their citizenship duties but to indulge their fantasies and treat them as easily manipulated marks (i.e. people who receive pleasure easily), and to encourage them to enjoy being addressed that way as a kind of flattery. LLMs optimized for sycophancy are especially useful for this, capable of relentlessly love-bombing some hapless interlocutors into a compliant stupor.

But for how long could we possibly find it flattering to be barked at from all sides by sycophantic machines, especially when it has become common knowledge that nothing at all goes into the production of all that flattery, not even the suspension of some hack writer’s ego? How soon before language stops generating the illusion that there is necessarily someone speaking it, and it turns into a stagnant pool of signifiers?

The current generation of “AI” has frequently been promoted as a kind of trickery that wise early adopters can use to get over on the laggards who still anachronistically believe in “communication” and “empathy” and other such nonsense. Why not fool the ones you “love” by getting an AI to “care” about them for you? What is “love,” anyway? Does anybody love anybody anyway?

The presence of AI is a sure indicator of a more profound absence of human consideration. Thought is not a private act in an windowless-monad brain but a token of implicit public commitment, a kind of attention that has been devoted that seeks recognition or reciprocation, that evinces faith in the existence of other consciousnesses that can share it. Thinking entails thinking about something, for someone, it posits a faith in a social relation. LLMs are set out to negate and stifle that faith. Where human thought inherently marks the recognition of other beings capable of their own subjective intentions and interpretations of things — human thought is a gesture of recognition as well as a “content” — mechanized simulation of thought is a deliberate refusal of that gesture, an elision of that recognition, an attempt to deny the other’s subjective consciousness and render them a kind of manipulatable content, a transformable output among other outputs. It is premised on the idea that other people don’t deserve attention but should be fooled into surrendering it along with their agency.

The LLM-generated Heat Index supplement is a somewhat low-stakes example of this, but it exemplifies how AI is fundamentally a technology of disrespect, of giving nothing where something is commonly expected among people trying to live as a society. But there is another way of thinking of the supplement. In Of Grammatology, Jacques Derrida develops his concept of the supplement to capture a paradox he identifies in the writings of Jean-Jacques Rousseau. In Derrida’s analysis, Rousseau sees writing as a degraded form of speech, which itself falls short of capturing the essential thing, the real presence of a human being. Writing ends up being a double-edged attempt to address that failing in speech: as it attempts to supplement speech’s ability to evoke presence, it also supplants speech, replacing and negating it.

As substitute, it is not simply added to the positivity of a presence, it produces no relief, its place is assigned in the structure by the mark of an emptiness. Somewhere, something can be filled up of itself, can accomplish itself, only by allowing itself to be filled through sign and proxy. The sign is always the supplement of the thing itself.

Using language to try to capture the “thing itself” is always a rowing against the tide: “an infinite chain, ineluctably multiplying the supplementary mediations that produce the sense of the very thing they defer: the mirage of the thing itself, of immediate presence, of originary perception.”

In Derrida’s account, Rousseau regards the doubleness of supplementarity — the fact of “mediacy,” that “real presence” is only ever known to us through an infinite chain of representations, through a kind of streaming media — as “evil.” He quotes from Rousseau’s educational tract, Émile:

As soon as [children] can think of people as tools that they are responsible for activating, they use them to carry out their wishes and to supplement their own weakness. This is how they become tiresome, masterful, imperious, naughty, and unmanageable.

For Derrida, this suggests that for Rousseau, “the supplement will always be the moving of the tongue or acting through the hands of others.” It marks a “regression toward an evil that is not natural and that adheres to the power of substitution that permits us to absent ourselves and act by proxy, through representation.” This perspective is readily applied to “AI”: a proxy actor that enables absence and obfuscates presence, that allows for exploitation beyond the “natural” tendency of humans to care for one another that asserts itself when the conditions of true presence are maintained. (Derrida also develops this analysis in light of Rousseau’s shame over his masturbation habits.)

The point of Derrida’s analysis is not to confirm Rousseau’s sense of “presence” as essence, and writing and textuality as evil indications of our having fallen away from it; instead, he makes his famous claim here that “there is nothing outside of the text” and tries to describe an approach to texts — eventually known as “deconstruction” — that would help prevent us from falling into the metaphysical trap Rousseau is in.

What does any of this have to do with the Heat Index supplement? It seems easy to take a Rousseauistic view of it as I did above, lamenting of how certain kinds of empty texts permit a kind of exploitation and dehumanization that is contingent on the suppression of thought and the human relations it implies. As automated text generators flood the world with ever more textuality, there will be ever more absence in the world, and it will become increasingly impossible for presence to announce itself, to make itself known and presentable.

But Derrida emphasizes that this has always been the case, and lamenting it does nothing but perpetuate that impossible chase for pure presence, for things that are finally and completely authentic. Derrida’s guarded and hedged suggestion of an alternative is presented, as he says, in an “entirely negative” way, but it doesn’t consist of salvaging “true presence” or the “real voice” or anything like that — nothing is ever sufficiently pure, things only seem more and more mediated, fake.

The fact that LLMs can generate endless amounts of explicitly “fake” copy with the traces of human intention and presence deeply diluted through countless layers of processing and concatenation could hopefully demystify not only that particular subject position that seeks safe harbor in “real texts” — i.e. an alibi in a “real supplement” for the dubious pleasures such supplements have always supplied — but also the fantasy of accessing perfect authenticity through media.

It seems likely that promoting and publicizing the kinds of tricks and shortcuts and opt-outs from the demands of being a human being that tech companies promise will engender a universal skepticism of the low-effort gestures that are now possible, raising the bar for what counts as care and increasing the value of human attention — when you can manage the increasingly arduous process of establishing its genuineness. AI could thereby inadvertently help further expose and discredit all the automaticity and instrumentalism that had already crept into social relations with capitalist modernity. But that doesn’t mean that the “real paper” will ever be rescued from its supplements. We still have to pursue a different way of reading everything.

In the colophon of the mass-market, pocket-paperback copy of Georges Lefebvre’s The Coming of the French Revolution that I’ve been reading for no particular reason appears the note on the type pictured above. In case you are like me and you don’t allow images to display in the emails you receive, it reads like this:

The text of this book was set in Electra, a linotype face designed by W.A. Dwiggins. This face cannot be classified as modern or old-style. It is not based on any historical model, nor does it echo any particular period or style. It avoids the extreme contrast between thick and thin elements that marks most modern faces, and attempts to give a feeling of fluidity, power, and speed. Composed, printed, and bound by THE COLONIAL PRESS INC., Clinton, Massachusetts. Paper manufactured by S.D. WARREN COMPANY, Boston, Massachusetts. Cover design by BEN SHAHN.

Who is this text for? It appears just below the author’s bio, as if any curious reader might want more information not merely about the historian who wrote the book — who is credited with submitting “perhaps the longest doctoral dissertation ever written” — but also the typographers involved. The text reads like a combination of an artist’s statement and a sommelier’s tasting notes, evoking a connoisseur-like reader who chooses what to read next based on what font the body copy is set in, and how much “fluidity, power, and speed” it affords.

It’s somewhat ironic that W.A. Dwiggins, who I am assuming composed this paragraph, insists on the ahistoricity of his font: Can we really come to terms with the coming of the French Revolution in a typeface that is “neither modern or old-style,” that rejects all historical models? It seems like some oblique commentary on the book itself, a remote echo of the consequences of the history it encapsulates. Whereas it’s easy to see why having a cover designed by socialist artist Ben Shahn made some conceptual sense — he began his career as an advocate for the rural poor, whom Lefebvre credits with enforcing by insurrection the juridical reforms proclaimed by the Constituent Assembly — but what does this noncommittal linotype face have to tell us about freedom and the rights of man and citizen? Mourning becomes Electra? Not this one.

The most plausible explanation for Notes on the Type is that they appear as a professional courtesy to the production editors, and serve to alert readers that more goes into the making of books than writerly inspiration and editorial intervention. Fonts make books legible at a foundational level and shape the reading experience in subtle ways that the Note on the Type hopes to raise to readers’ consciousnesses. It matters to see type at the level of form, as a medium with material characteristics. The notes implicitly urge readers, before shelving a book for perhaps forever, to take a moment to see the window rather than the view. And though they likely go unread nearly 100 percent of the time and thus fall short of that ideal mission, they still signal to other bookmakers that a particular publisher has institutionalized a process for acknowledging the artisanal work that makes printed words.

A commenter on this discussion forum “for professionals and enthusiasts in the fields of typeface design, lettering, and typography” linked to this compendium of “notes on the type” in the Internet Archive (pointing out also that W.A. Dwiggins was “a frequent adopter of the ‘note on the type’”). It turns out that many of them appear in books published by Knopf. They all have the same general tone and format: an acknowledgment of the designer, a location of the typeface within the broader history of printing, and a brisk account of its aesthetic distinctions, with some highly specific details about the serifs, descenders, or weight of the strokes, as, for example, in this haughtily passive-voiced point from the note on the type of H.L. Mencken’s In Defense of Women: “Other characteristics that will be noted are the square serifs without fillet and the marked contrast between the light and heavy strokes.” One will note those unfilleted serifs, won’t one?

The notes on the type can be more than a bit pedantic in their effort to educate the literate public about fonts. They assume an authoritative, faintly impatient tone that presumes a regrettable ignorance in average readers, who are treated as though they have only the most rudimentary understanding of the symbols that make their participation in print culture possible. But the notes can also sometimes seem a bit partisan, taking sides in an obscure struggle among different camps of typographers: The note in Willa Cather’s One of Ours celebrates Old Style No. 1’s “workmanlike quality and freedom from ‘frills’ characteristic of English old styles in the period prior to the introduction of the ‘modern’ letter.” This sounds as if it might have been a better font for Lefebvre’s work, but the author of the note stresses that it too is “of English origin,” so perhaps not. I wonder what “frills” were considered so dubious that they needed to be referenced in scare quotes, as though the note writer could hardly deign to mention them.

The more contemporary notes on the type haven’t changed much in tone. The note in Donald Antrim’s 2000 novel The Verificationist gives a quick bio of Claude Garamond, which, if nothing else, reminds readers that the names familiar from their font menus often refer to historical figures. He “gave to his letters a certain elegance and a feeling of movement that won for their creator an immediate reputation and the patronage of Francis I of France.” I guess the ancien régime lives on after all.

Every time I encounter “A Note on the Type” in all its earnestness, it strikes me as obvious fodder for a Shouts and Murmurs–style parody, and naturally there is one — probably one of several — from 1997 by Bruce McCall. It’s about as “funny” as you’d expect. (“That part of the text not set in Backslap or Bangalore — the lowercase d's, k's, and alternate z's, except after c — is Jiffy-Lube Piscataway Light Narrow, based on a sixteenth-century face closely resembling the late Edward G. Robinson.”)

McCall’s piece imagines a hack freelancer who’s been paid to compose notes on the type and gives voice to his cynical complaints and frustrations. Maybe that was funnier to imagine in 1997, the failed writers working at the fringes of Big Print at tasks that now evoke more nostalgia than contempt. And maybe the tone of the notes seemed much more unremittingly pretentious in the midst of a ubiquitous print culture whose supremacy was unchallenged. It was easier to laugh at the idea that typographers were trying to claw away some of the glory that was due to writers and editors, as if anyone could ever care more about the appearance of words than what they mean. When publishing was considered a glamour profession, knowledge of nerdy arcana like “fonts” was probably akin to knowing about grips and gaffers and best boys on a film set, as McCall’s piece suggests.

But now words are everywhere treated as data, as numbers, and they are rarely printed at all. Everyone has some working knowledge of fonts and even strong opinions about them, even if they rarely read books. Words pile up on trillions of screens without being read by anyone, so much visual clutter and static to human eyes, even if they will be registered and processed by the machines that are increasingly programmed to produce them. The impact of fonts change when no one is expected to turn words into meanings inside their own heads, when autocratic governments in league with the tech industry work to compel everyone to have meanings dictated to them without the intervention of their own thought processes. Fonts perhaps become more salient as they become the only thing that people are expected to play with when it comes to words, once it is drilled into everyone that it is required for efficiency and political conformity to let machines do all the writing and the editing and the “thinking.” We won’t have official authority to change what words mean on the page, but we’ll be encouraged to think that changing the way they look is really the best way to express ourselves.

Whether from arrogance or incompetence, Mark Zuckerberg has a tendency to issue statements about his company’s intentions that seem meant as matter-of-fact justifications but come across as confessions of the kinds of damage he is prepared to inflict.

The classic example is his 2010 declaration that “having two identities for yourself is an example of a lack of integrity,” meant to justify Facebook’s “real names” policies and the overarching strategy of data collection as a means of integrating information about users to make them into marketing targets. That statement is obviously not factual, but it did reveal a lot about the company’s aspirations then to be the arbiter of “integrity” across a range of social relations. (This was the era when people were urged to treat someone not having a Facebook account as a mark that they were suspicious, that they had something to hide or were pathologically antisocial.)

Zuckerberg has also made countless preposterous statements about the “metaverse,” the now sidelined initiative to force everyone to live more of their lives under corporate surveillance in environments under complete corporate control. “Think about how many physical things you have today that could just be holograms in the future,” he mused in a “Founder’s Letter” from 2021, as if the tangibility of the things that belong to us was a long-borne and widely detested onus. Behind this transparently stupid statement was the obvious aim of the then newly christened Meta to stake an ownership claim not only over users’ “true identities” but over all the things that populate their sensoria, replacing physical reality with virtual realities for which we would need a cascading array of subscriptions to navigate the ever-proliferating levels of differential access.

The latest candidate for the canon is Zuckerberg’s claim on a podcast this week (detailed by 404 Media here) that “the average American has, I think, it’s fewer than three friends, and the average person has demand for meaningfully more. I think it's something like 15 friends or something.” He says that “people just don't have as much connection as they want,” and suggests technology can remedy the deficit by simulating it. The aim telegraphed by these inept statements is Meta’s hope to replace human friends with chatbot products under the auspices of combating “the loneliness epidemic.” (Strange how it’s rarely proposed that we address the loneliness epidemic by attacking its root causes: capitalism and technologically driven anomie.)

Many commentators have pointed out that Zuckerberg’s statement sounds like something that someone who has no friends and who is perpetually surrounded by toadies and opportunists would say. It betrays a blinkered understanding of friendship (or “connection,” to use the network-derived term Zuckerberg prefers) as being like a consumer good, something one has a quasi-economic “demand” for in the midst of an uncertain supply, something you should stock your pantry with, something that has an optimal consumption dosage. It seems to posit that having friends (“connections”) is something solitary individuals can elect to choose; they are free presumably to decide how much time and money they should allot to friend acquisition, which is all that limits their total. Likewise, friendship is regarded as something from which satisfaction is extracted on an individual and unilateral basis — you take pleasure in friendship, whether or not it is given.

The idea that reciprocity, responsibility, and care are aspects of friendship is alien to Zuckerberg’s apparent vision of the world, as the trajectory of Facebook as a company has long made plain. His platform has always encouraged users to view friends as resources and as content. Built into the mission of “connecting the world” is a view of people as nodes or sockets into which wires can be plugged or unplugged at the network scientists’ whims, and a view of “connection” as a matter of signal efficiency rather than meaningful exchange. So naturally he would see no concerns in replacing human contact with LLM-generated content—connection is connection, and it doesn’t imply the mutual interpenetration of consciousnesses. You can connect with a person, or a joystick, or a spreadsheet, or an algorithm, and it all counts the same.

It’s no surprise that, as John Herrman explains, “Zuckerberg imagines media platforms in which other people are augmented or replaced by chatbots trained on their collective data, with all parties involved coming together to engage with content produced by AI tools trained on ‘cultural phenomena’ in a space arranged according to opaque and automatic logics.” Herrman points out that this merely extends the premise that one now uses Meta platforms not to engage in friendship but to communicate with algorithmic systems to guide the kinds of entertainment content they supply. Talking to a bot is just a new UI for the long-established slop machine.

So why does Zuckerberg talk about “friends” at all? Is there some dim awareness that people don’t want too clear of a look in the mirror Meta supposedly holds up to them, and that they still need to be coddled with promises of camaraderie to distract them from the slurry they are killing time with? Is the idea that if you call TV shows something like Friends, the viewers feel less guilty and less emotionally stunted when consuming them?

I argued before that “social media” was an alibi for injecting more TV into people’s lives to take advantage of increased network connectivity — that we had to be persuaded that it was a pro-social thing to do to carry little TVs around and watch them at every possible moment. Conflating friendship and entertainment was part of that campaign. Now chatbots are being put to work on the same ideological project: Here are entertainment products that you can treat as friends, just as you have become accustomed to regarding your friends as entertainment products.

But that Zuckerberg and other tech moguls have to go on podcasts and make idiotic pronouncements suggests that entertainment and friendship are not so easily conflated, and that inculcating that ideology requires constant hammering, constant bolstering not only through apps and interfaces but from media coverage that is meant to normalize what we all tend to resist, the idea that we should use other people for our amusement and nothing more.

I’m always perplexed by the idea that anyone would want to have small talk with a chatbot, could find what a chatbot has to say intrinsically interesting simply because the chatbot said it. It’s common to have such conversations with people, where the point of talking is not always to communicate information but also to establish a bond, to indicate that you are willing to pay attention to each other, that there is a fulfilled expectation of mutual care. But chatbots don’t care about you any more than Meta or Google or OpenAI does.

Similarly, a lot of the information I receive from other people is meaningful to me not in itself but because they chose to share it. There is some of that “desiring the desire of the other” that Kojève writes about to it; I want to care about what other people care about because they are people, not because the object of their care might have some utility for me alone. The information we exchange is sometimes no more than a token in strengthening or articulating a relationship among people willing to be with one another. But getting information from a machine offers none of that. There is no desire behind what it generates, or if there is, it is infinitesimally extenuated and impossible to ascribe to human care.

I get this feeling from algorithmic recommendation: Even when it is “correct,” it feels empty, utterly meaningless in comparison with being told by a friend to check something out. It is as hollow to be told what a machine wants for you as to be told what a machine generated — is there anything worse than someone prefacing a statement with “I asked Chat-GPT and …”?

But maybe there is a dialectic to algorithmic recommendation that makes moments of reciprocity resonate even more powerfully. The more the world fills with slop, the more human gestures stand out and reverberate with new meaning. Accessing glib simulations of information will have become commoditized, but someone actually telling you something, anything, will seem more important than ever.

In his recent book Superbloom, Nicholas Carr mentions a bit of Mark Zuckerberg’s testimony during a congressional hearing in 2018: “I consider us to be a technology company,” he said of what was then called Facebook, “because the primary thing we do is have engineers who write code and build products and services.” That hearing was in the wake of scandals involving Cambridge Analytica and Facebook’s role in spreading political propaganda and abetting foreign interference in U.S. elections, something most tech companies no longer pretend to care about.

Carr cites this statement to illustrate his point that tech companies approach media as a form of engineering, as information processing, in which the meaning of whatever content they circulate is irrelevant to their mission of circulating it efficiently and profitably. He traces this approach to Claude Shannon’s theory of information, which regards it as nothing more than binary code and therefore fungible, entirely quantitative. Messages, in this view, can’t have hidden or multivalent meanings; they don’t have nuances or require interpretations. They don’t depend on any vagaries or subtleties in the a relationship between a sender and a receiver; any subjectivity in the process is treated as noise.

Once this perspective is adopted, all the different forms of analog media can be reduced to digital media. This includes the recipients: They are just digital information too, and technology companies set about building the interfaces that would convert their behavior into digital information commensurate with the other ones and zeroes. “The convergence of media technology … blurred the distinctions between media businesses,” Carr writes. “It has blurred the distinctions between categories of information — distinctions of form, register, sense, and importance — that the epistemic architecture of the analog era preserved and even accentuated.”

This blurring makes one kind of problem irrelevant: How do people understand one another? What difference does that make? People are just information processors and emitters. There is nothing to “understand” — just ones and zeros to record. A different problem opens up, one that engineering can address: How can different information flows be monitored, controlled, and harmonized, brought together to make new forms of value? How to reproduce human behavior and interaction as nothing more than code, as no more than pattern matching?

Hence Facebook was not in a “social” business at all — deliberately connecting people in meaningful ways — but instead implemented a control system in which people can be collectively managed as information. Whether people incidentally experience feelings of connection is immaterial to the business of configuring them as data flows — of using certain configurations of data to elicit certain other configurations. For Carr, the advent of the News Feed concretized this: After its implementation, “each member now saw a continuous, customized stream of posts and updates,” but also was themselves no more than that, a data flow to be integrated with other data flows.

Though there were some vocal protests to the surveillance and privacy implications of the News Feed at first, the company discovered that optimizing users as information flows didn’t alienate them but instead locked them in — that people presumably felt something pleasing, something convenient and clarifying perhaps, in being cut off from analog experience, from the ineffabilities of human interaction and sensory stimuli, and from the burden of actively making meaning. Instead one could be effortlessly rendered as content, as part and parcel of the same substance that everyone else was eagerly consuming. To use “social media” was to feel what it was like to be turned into manna.

By simply consuming content one became content, which was tantamount to becoming a self, having an identity (a discrete flow of ones and zeroes), within the structure of automated distribution and feedback. Platforms and feeds allow users to be programmed by a flow a content into exhibiting certain anticipated and measurable responses. As a consequence, people no longer use “social media” to be social but to escape sociality and become more like a robot. You can sense what is like to have data efficiently and efficaciously flow through you, become transformed by your circuitry and algorithms, and emerge on the other side as a new kind of data set.

In 2018, Zuckerberg was reluctant to concede that Facebook was a “media company” because he didn’t to face liability for the editorial decisions implicit in producing algorithmic feeds. He didn’t want to have to invest more resources in content moderation. But now his company faces a different threat, an antitrust suit from the Federal Trade Commission stemming from its buying rivals Instagram and Whatsapp, undermining competition in the social media sector. So now, as Kyle Chayka details here, Zuckerberg claims Meta is not a social media company but an entertainment platform for “the digital consumption of all kinds of content.”

To put that in the terms Carr’s discussion suggests, Meta can’t be a social media company because there are no such things as people or society, just kinds of digital content, and even “consumption” is misleading because it presumes a consciousness to consume it, implying something more than signal exchanging and decoding is going on. “Digital consumption” is apt, though, because one can interpret that as another phrase for information processing, for “consuming” things by digitizing them.

“Social media is dead” is now the official tech ideology because “social media” successfully served its purpose as an alibi for the complete digitization of experience, promulgating the idea that everything is information, everything can be quantified, and that human beings can enjoy themselves in a system that doesn’t register anything qualitative — from which “quality” itself has been eliminated. Humans, it turns out, can happily persist amid a universe made of undifferentiated slop, and aren’t necessarily repelled by being processed and circulated as slop themselves — an ad hoc configuration of data produced algorithmically depending on what the other flows of data make statistically probable, or what kinds of data have been prompted farther up the chain.

Now tech companies are less shy about dichotomizing sociality and entertainment. Connecting with friends always was a dubious and unmanageable business, difficult to directly monetize. It involves having people turn their friendships and their family relations into pretenses for generating more ad space. We had this baby to help clear some ad inventory. It entails convincing people to live their lives as advertisements, as with influencers, and it depends on users’ deliberately producing content instead of merely being content.

Moreover, it depends on people actually wanting to consume their friendships rather than participate in them. For “friends” on social media to be profitable for tech companies, they must not be boring to each other, they must not engender a sense of reciprocal duty or obligation (the algorithmic feed precludes a balanced sense of attentiveness to one another), and they must not encourage each other to do anything other than consume more content on the platform.

But friends, viewed as entertainment sources rather than peers, are generally boring. They can’t consistently make content that outperforms with professionally made material, or compares favorably with all the content that all the world’s amateurs make available to algorithms. A platform of just friend-level content can’t survive in competition with content made expressly to be marketable as entertainment, or content revealed as attention-grabbing through algorithmic selection and feedback.

Social media platforms don’t try to rebalance this but exploit it, encouraging users to further enjoy entertainment more than interacting with friends. Platforms demonstrate that while friends’ attention is fickle, entertainment is always available on demand, and thanks to portable screens and ubiquitous network connections, consumable at all times, in times and places that were inconceivable before. The innovation of social media is not its surpassing TV in its interactivity and its supposed promise of reciprocal communication, but that it replicates and extends the experience of TV consumption and injects it into every crevice of waking life. It makes the comfort of TV, such as it is, a constant temptation.

There used to be some concern about tech companies owning what Facebook liked to call users’ “social graph” — who was connected to whom and how often did they interact and so on. Critics assumed it would be weaponized in some way against us, used to target people more efficiently or categorize them in various ways against their will and without their knowledge. It was feared that people would become dependent on tech companies for sociality, that the companies will have embedded themselves as “social infrastructure” — another description of Facebook’s business that has now fallen by the wayside.

But it turns out tech companies built social infrastructure only to undermine it, to help with dismantling it as a site of resistance to commercialization, commodification, and mediatization. There is more money in scuttling the “social graph” than leveraging it, because isolated people make for more dependable consumers.

My intuition about social media has always been that it wanted you to hate your friends. It wanted you to see them at their worst, as showoffs seeking to stimulate envy at every turn. Not that this was always everyone’s intention, though the platforms certainly provide the incentives to make it so, but because this is what results from algorithms trying to turn friends’ content into entertainment (turn friends into content), circulating it only because it has previously proved popular across diverse contexts and not because it reflects a specific connection between specific people. When communication is filtered through algorithms, it becomes as if spoken by algorithms; it appears as content made only to keep the system operating and the feeds populated, not meant to sustain a personal connection but as data meant to fine-tune the processing of other data.

Commodified entertainment sold as a media product or an advertising adjunct is basically anti-social, despite our dogged efforts to resocialize it and build communal practices around it. When what friends post is presented as entertainment media, it requires that we think of those friends as remote, as being something we can only enjoy vicariously.

Ultimately Zuckerberg is not saying anything different now than he was in 2018: Meta is an technology company whose engineers work to turn people into data. That can be implemented under the auspices of an entertainment-media platform, which reduces users to what information they consume. That information is conceived as coming from no one in particular, with no purpose other than to keep you watching, producing more information in turn.

Hari Kunzru is right to point out that “the ‘can machines do creative writing’ thing is mostly a distraction from the use of the machines to go through text and images to cancel grants and put people on deportation lists.” So the best way to understand OpenAI’s recent claims to have trained a new model that, according to CEO Sam Altman, is “good at creative writing” and “gets the vibe” of “metafiction” is that the company is running interference for the authoritarians using similar technology to automate surveillance, circumvent human scruples, and do away with due process.

It’s no surprise that the example that Altman provided in an X post was terrible (“We spoke—or whatever verb applies when one party is an aggregate of human phrasing and the other is bruised silence”), but that is because he and his collaborators undoubtedly have terrible taste. With sufficient massaging and competent editorial judgment, an LLM could probably be used to generate passible, plausible fiction of whatever sort one wanted, though this would obviously reveal nothing about the machine’s talent or sensibility (it’s not sentient) and would be interesting only for the insights it afforded into the people who iterated on the prompts and, most important, decided to share the generated output with others. It’s not “machinic creativity” so much as found poetry on demand. That this seems oxymoronic and self-negating is perhaps indicative of its intrinsically limited appeal.

In a recent paper, Henry Farrell, Alison Gopnik, Cosma Shalizi, and James Evans make the case that LLMs are a “cultural technology” that can “abstract a very large body of human culture” and mediate that culture to users in novel ways. As Farrell summarized in this thread, generative models “take incomprehensibly enormous bodies of textual and visual information, and create imperfect but much more manipulable statistical summarizations of it, that can be used at scale to summarize, rephrase and remix.” But a LLM remains a “social technology”:

its uses involve human relationships mediated through technology. When someone uses ChatGPT to craft a resume, they are engaged via an LLM in a social relationship with all the other humans who wrote the resumes etc. in training set and also with the humans who trained the LLM through reinforcement learning, or generated the synthetic questions that were used for [reinforcement learning]. Like Soylent Green, or more prosaically, the vast swathes of culture it summarizes, it’s all made out of humans!

To produce machine-generated fiction is likewise to enter an extremely distant and highly mediated relationship with “fiction writers” as well as the beleaguered contract workers paid to evaluate samples of generated text in accordance with various contrived and deaestheticizing rubrics. Whatever unique conditions provided a creative impulse or a compelling moment of recognition to a particular person in a particular instance are averaged away so that the user can play around with the various forms in which creativity has appeared and resonated with others — i.e. one can toy with the “vibe” of creativity without any inspiration or understanding of how it works. Then, once the user shares some results of this process with someone else, the text is recontextualized by that gesture, and the user effectively becomes its author.

In his recent book-length poem Context Collapse, which examines how media technologies reshape the kind of poetry that can be made, Ryan Ruby offers this paraphrase/extrapolation of a 1966 paper by computational linguist Margaret Masterman about “toy models of language”:

Are two identical sentences, the first produced by a human being and the second by machine intelligence in fact indiscernible? Answer: No … if two subsequent conditions obtain. (1) The recipient of those sentences is human (the criterion of care) and (2) said human does not know whether the sentence was produced by a human or a machine (the criterion of context). If the same sentence can have different valences — whether semantic or perlocutionary — in different contexts of production or reception, it follows that this will also be true in the noncontext of a machine-produced sentence, or when awareness transforms that noncontext into a paratext which provides a frame for interpreting and understanding it.

I would paraphrase that paraphrase as: machine-generated texts are meaningless in and of themselves without the context of human exchange, regardless of whether they can be syntactically parsed. The meaning of communication, the value of what is communicated depends on, as Farrell puts it the “human relationships mediated through technology” and not the technology itself, whether that is a pile of circuit boards or a pile of phonemes.

The “scene of reading” — the situation in which a text is presented to you — establishes horizons of meaning as well, and this scene will also have irreducible human relationships at its base. If some researcher tries to fool you by presenting you a LLM-made text as human-made, it doesn’t matter if you take the bait: The researcher is now the writer. Max Read argues here for “close reading” of LLM output, but it seems like the wrong tool; maybe a phenomenology of prompting, or statistical table of word frequencies, or a network analysis of Sam Altman’s tweets makes more sense. Or some sort of Franco Moretti-style “distant reading” in reverse. Maybe reader-response theory. Trying to parse LLM text closely without prioritizing the context of who made it and distributed it and why someone opts to engage with it seems pointless, like trying to understand why q comes before r.

When language circulates, the presence of another person is automatically implied; even LLM text implies human intentionality at some level. But it is maybe a mistake to think that the point of machine-generated text is to simulate the thought and feeling of another person. The point may not be to elevate our appreciation of the machine’s creativity to a human level, but to allow a human to degrade their encounters with language to something more predictably mechanical.

A critic on Bluesky pointed out that entrepreneurs like Altman “fail to understand that creative writing isn’t a slop bucket that needs refilling by any means possible. The reason creative writing is beloved is because it gives us insight into the thoughts and imaginations of fellow humans, not homogenized and plagiarized slurry.”

That’s probably true, but it may be that people are sometimes seeking an escape from “the thoughts and imaginations of fellow humans” because they are threatened or inconvenienced by having to take other people into account, and what they want is an endless stream of content that negates human creativity and frees them from having to live up to it. In other words, LLMs promise to turn language — ordinarily polyphonous and uncontrollable and irreducibly social — into something more like machine gambling, a solipsistic flow experience that provides an illusion of control by impoverishing the range of experience and reducing it to refilling the slop bucket, over and over again.

The term “vibe coding” has been around for only a week or so, but it already has a Wikipedia page. Credit for its coinage is assigned there to one of OpenAI’s co-founders, Andrej Karpathy, who defined it in an X post:

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. I ask for the dumbest things like "decrease the padding on the sidebar by half" because I'm too lazy to find it. I "Accept All" always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I'd have to really read through it for a while. Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away. It's not too bad for throwaway weekend projects, but still quite amusing. I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.

This seems like a routine piece of hype for the capabilities of LLMs, with the usual overtones of proud and eager submissiveness, boasting about being “too lazy” as though it were the best luxury imaginable. I am so excited to evacuate myself from the work I do! So amusing! I can’t wait to repudiate self-consciousness next! Just see stuff say stuff whatever I don’t know why!

But the term appears to have quickly caught on, perhaps because of some coordinated inauthentic behavior on the part of journalists like Kevin Roose, who enthuses about the idea in this New York Times piece. “You don’t have to know how to code to vibecode,” he writes. “Just having an idea, and a little patience, is usually enough.”

Such claims for “vibe coding” seem to place it in the tradition of hyping “AI” by dismissing craft, as though the process of working in a particular medium and struggling against one’s limitations within in were irrelevant to the end product. Instead, finished work can supposedly be unfolded by rote from whatever lightning-bolt idea has struck someone. This all-theory, no-practice attitude posits a sort of rain-dance approach to creation in which one tries different “random changes” without worrying about “comprehension.”

Perhaps one could make the case that the prompting becomes the medium in the sort of process Karpathy describes, and that the series of iterations turn the natural language of the prompts into a refined set of tools to be handled with artistry. But the whole tenor of Karpathy’s post seems to work against that idea, encouraging us to understanding prompting as the “dumbest thing,” as a way to forget that the underlying medium even exists. It points to a fantasy of creation without a medium — content without the hassle of form.

In this 404Media article about a vibe-coded game, Emanuel Maiberg defines vibe coding as “being less methodical and detail oriented, telling the AI tool what you want, and getting it to work without worrying about the code base being messy.” We are supposed to be excited that AI will liberate the world from the tyranny of expertise, and we can all be equal in our ignorance. But why is “vibe” part of this term? What does Karpathy mean when he says “give in to the vibes”? Is it just because that word still retains a residue of virality? What is the point of describing deskilled work as a “vibe”?

When I first saw “vibe coding,” I assumed it was pejorative, meant to mark it off from “real coding,” where the coder understands how the code is put together. I was thinking of how vibe was being used back when it was described certain TikToks, as evoking a feeling that can’t be pinned down in words, or a claim that couldn’t be substantiated with data. Then it was a general word for a mood or a feeling; it seemed to mean the opposite of “having an idea,” insofar as having ideas also means being able to slot them into causal relationships. Vibe indicated an inability to analyze a certain situation that is accepted instead as a gestalt, a mystic whole.

“Vibe coding” retains the implication that you can’t explain or even understand how something works. But under the pressure of AI hype and its championing of incomprehension, there seems to be not a “vibe shift” but a “vibe” shift occurring, as the term drifts toward a different connotation. Where vibe once conveyed something that can’t be analyzed, now it conveys a purposeful indifference to analysis or explanation, as well as to the components that make up something. It is as though the preponderance of vibe talk made explanations irrelevant in all cases, and now we speak of vibes to forbid comprehension, which would be unfun.

“Giving in to the vibe,” then, means deliberately refusing to understand, as though that would be to defy AI’s supremacy. One should let AI handle the data and the details so that you can just have gratified impulses, which are ultimately just a matter of data being manipulated to your liking — prompting, waiting, tweaking, and trying again until you are satisfied or bored with what you’re doing. Being involved with the thinking process, the details, would be to go against the vibe. It’s a superfluous burden that sets you against the spirit of the times.

“Vibe” in this usage tries to lend exuberance and cachet to what amounts to being passive and negligent. It borrows from the earlier connotations of relaxing into conditions and not disrupting them by trying to pick them apart, but now applies it to the alienation that “AI” must produce if it is going to be profitable for its makers. “Vibe” evokes the the dream of a life that contains no processes, no skills, no research, no effort — just feelings that machines interpret and turn into substance. If “vibe” once could refer to a range of nameless feelings, now it refers to thoughtlessness, in every sense of the word.

When I first saw the Comedy Central show Mystery Science Theater 3000, in the early 1990s, I was most likely home on holiday break, because I didn’t have access to or interest in cable television at college. The premise of the show is that you watch a group of friends watch a bad movie and make fun of it, and I remember thinking it was slightly pathetic to have to turn to TV for that experience. I didn’t need a show that simulated the experience of having friends; I lived with friends, worked with friends, ate nearly all my meals with friends, was continually in the midst of friends, and there was no shortage of sarcastic commentary and inside jokes about whatever we encountered.

Seemingly aware of the innate pathos of its premise, Mystery Science Theater represented its group of friends not as a cool clique you could vicariously join but as a single man isolated in outer space who has been driven by loneliness and inadequate entertainment media into building himself “robot” puppets to talk to. (It prefigures the current vogue for pushing chatbot companionship.) A few years later, after I graduated and moved across the country to an alien city where I had no friends and spent nearly all of my time alone, I became a fervent viewer. I had dozens of episodes on VHS, recorded at SLP speed to fit three per tape, and most nights I would fall asleep with one of them playing. Some of the jokes still recur in my mind unprompted when I’m cued by certain phrases or names, a sad mode of self-soothing that has become permanently embedded in me.

Mystery Science Theater achieved a cult status in the 1990s, and one could read in press coverage of the show about how people would get together in their own groups every week to watch it. But this never was my experience and it never made sense to me; the show felt like a substitute for and not a facilitator of sociality. I understood it as being about alienation (I can’t find straightforward relief in what the world offers to amuse me) and coping with the ensuing isolation. It shows an attempt to wrestle entertainment media into being a kind of conversation partner and not just a sedative. I saw it as modeling not so much collective experience as a mode of quasi-Brechtian nonpassive consumption, in which you refused identification with a film’s characters and situation, and resisted getting caught up in a story but instead stood aloof, assessing with irony and occasional sympathy what a movie was trying and invariably failing to do.

Mystery Science Theater illustrated in a backhanded way how hard it was to make entertainment and, accordingly, how hard it is to entertain ourselves, to be entertained, how much work we ordinarily have to do to allow entertainment to work on us, covering up for its mistakes and implausibilities, embracing its stereotypes, and fleshing out its formulas. The show frequently zeroed in on the incompetent attempts made by marginal production companies to mimic slick entertainment products, and it elevated these failures into a kind of found folk art, as though a kind of genius could be spotted when you learned to look at things with the right kind of cynicism. It basically depicted camp sensibility in action, which felt like a consolation prize for a more general cultural exclusion.

But even as Mystery Science Theater discouraged vicariousness with respect to the movies it mocked, it generated it at a different level, with viewers being invited to identify with an idealized audience and see themselves as part of a community of elites who could see through culture industry pabulum while besting it at its own aims. This never seemed like a potentially real community to me but a fantasy one: You could imagine yourself not as a character in the film but one of silhouettes in the theater seats at the bottom, one of the show’s true stars. Being in the audience was presented as preferable to being onscreen; the show dramatized the audience’s presumed power, its claim on the last word about any entertainment product and its ability to draw on deeper pools of knowledge and feeling, and ultimately its transcendence of the need to explain or defend itself.

At the same time, the audience onscreen did the watching and enjoying for you, saving you the trouble of having to pay close attention yourself. Through the show, viewers were able to consume “watching TV” as an activity without having to perform it; it made viewing more passive than it already was, more passive than passive,
”hyperpassive,” as Baudrillard might say. You watch in the same way as before but can believe that now your watching somehow expresses a criticality toward the idea of merely watching things.

Relatively early in MST3K’s history, Comedy Central made this behind-the-scenes promo documentary about the show, which, if I remember right, first ran during one of the “Turkey Day Marathons” when the channel would air something like 15 of the two-hour show’s episodes in a row. Ostensibly the documentary was meant to expand MST3K beyond its cult status, but at the same time, it seemed intent on proving the opposite point, that the audience for a show like this could never scale, that chasing popularity was antithetical to its existence.

One line from Joel Hodgson, the show’s creator and primary star, sums this up: “We never say who’s going to get this; we always say the right people will get this.” This line was in the commercial for the documentary, which ran as MST3K aired and thus ended up on my VHS tapes and lodged in my brain for life, perhaps to my detriment. It expresses the quintessential Gen X anti-populism, the suspicion of pandering to audiences or trying to be “relatable” in a general, nonspecific way. It also captures the idea of an aristocracy of taste, built on being able to “get” certain kinds of ironic humor and camp sensibility. The show’s detractors naturally regarded it as elitist and irony-poisoned, if not fully nihilistic, because it problematized people’s supposed naive enjoyment of media.

Looking back, it’s clear that this was a core contradiction of the 1990s, traceable not only in MST3K but in other highly reflexive entertainment products of the era, like The Larry Sanders Show, or Nirvana and the ensuing flowering of alt-rock radio, or grunge fashion and the ethos of “my image is I have no image”: It was a moment in consumerism when being a conforming nonconformist seemed especially viable, when mainstreamish media products could try to sell themselves as being too complicated for the mainstream, when being “anti-commercial” was understood as having a lot of commercial potential. It seemed like a central cultural dilemma, how to avoid becoming popular while also becoming popular. (The internet would turn it inside-out, making popularity the only remaining measure of reality.)

Whether or not I would have admitted it to myself then, I really wanted to consume authenticity through these kinds of semi-popular things that seemed to embody “not selling out,” and then later I felt like I wanted to repudiate that whole period of my life and the empty comfort of having tastes that I needed to believe that nobody else shared. I came to realize how that whole orientation depended on a stable cultural backdrop, a set of banal but integral shared values, that I was perhaps too quick to take for granted.

I was thinking about this because while I was in Ireland last week I happened to see a show on a British network that put the basic premise of MST3K to what seemed to me a different end from self-isolating irony. This was Gogglebox, a reality show that has apparently been on for a decade or so and consists of “ordinary people,” noncelebrities who appear to have been selected to be demographically representative, watching and responding to other popular TV shows. Whereas MST3K presents its surrogate audience as shadows, on Gogglebox, we see the designated viewers from the perspective of their television, sitting in their living rooms on couches and easy chairs, eating their snacks and sipping their beverages, making wise cracks to each other as they watch clips of shows we’ve presumably also seen. There is no presumption of “indie cred” to it, no implication that making fun of TV makes you some kind of outlier. Instead it gestures toward the idea that watching TV is something that normally brings households together in a spirit of light-hearted skepticism.

I don’t mean to overly romanticize I show that I saw only three episodes of, but nonetheless, Gogglebox came across to me as easy to watch and nonchalantly charming, and I was surprised someone hadn’t tried a similar show in the U.S. But then it occurred to me that the show was utterly at odds with the U.S.’s media environment. Gogglebox seemed like an advertisement for the prosocial effects of media consumption: It posits a national audience, a fantasy mainstream to replace the monoculture that no longer exists, and shows that “we” all watch, defining that “we” as a melting pot including people from different races, classes, family structures, and regions of the country. No one is ever watching alone.

Gogglebox represents that normative audience not as a slavering fan culture overinvested in the stakes of entertainment products and willing to fight and humiliate people over them, but as people who automatically tend to put the absurdities of other reality TV shows, news digests, Netflix dramas, etc., into a proper perspective. Though the show models a kind of talking back to the screen, it also suggests that watching TV is a low-stakes distraction, not something to build an identity around.

Though Gogglebox seems a reaction to social media and reads in part as an attempt to incorporate the “second screen” back into the first screen, it also rejects the imperatives of the social media feed to direct the most attention to the most unscrupulous attention-seeking figures and to generate outrage, conflict, and emotional intensity that can keep people scrolling indefinitely. In that way it feels like pure nostalgia, what we might think “social media” was if the platforms were never launched. Like MST3K, it makes a passive spectacle of “audience participation,” but this doesn’t contradict or betray its purpose; instead curtailing and interrupting the viewer’s need to participate in media production at all times seems to be the point. The commentators onscreen are showing viewers at home how low the stakes are supposed to be.

According to Gogglebox’s Wikipedia page, the show’s producers “did not want to feature people who wanted to be on television” and sought potential cast members in unconventional ways:

One of the methods used to find participants is termed “street casting,” whereby the show's team looked in everyday public places; Leon and June were found in a bridge club, and Stephen and Chris were found in a hair salon. In later series, members of the show’s production team visited random houses and held up a card that contained something, such as a picture of the British Prime Minister or a Daily Mail headline, and noted how quick the person responded and any funny, interesting, or insightful comments they had.

I don’t watch a lot of social media video, but I have spent some time in bridge clubs, and Gogglebox seemed to owe much more to that vibe than to TikTok. It took me a while to put my finger on it, and perhaps some of it derived from my being a foreigner, but it was unusual to see people onscreen who weren’t begging for my attention. But it also didn’t feel like I was spying on them. It conveyed instead the impression of very different people who shared in common the permission to be themselves publicly from within the safety of their own living rooms, and that this permission didn’t produced the id-driven megalomaniacs of the anonymous internet’s content cesspools but fundamental civility. Gogglebox makes of TV shows a kind of common target that people can vent their skepticism on, so they don’t have to regard one another as entertainment to be similarly mocked or seek to become entertainers themselves to remain visible and seem worthy of respect.

Seeking attention for profit is perhaps incompatible with tolerance. The paradigm for U.S. television is the local news broadcast, full of sensationalized crime reports, fear mongering, petty grievances and suspicions, and weather. Social media platforms, in a reflection of their own prerogatives, tend to disseminate the idea that tolerance itself is intolerable and boring; they produce people whose personality revolves around being compulsively watchable. But society depends on people being considerate, largely predictable, and ignorable when necessary. Gogglebox seems intent on showing that those kinds of familiar, considerate people appear across the spectrum of social identities and they aren’t without a sense of humor. But for all I know, the show has already been cancelled.

Open AI and Anthropic have recently announced or launched their versions of “AI agents”: pieces of software, often represented as assistants or servants, that use computers on your behalf to presumably free you for more important tasks. Often they are described as booking flights or restaurant reservations, as if these were profoundly onerous burdens from which everyone seeks relief.

One illustration of this is the deeply preposterous and widely despised Salesforce commercial in which Matthew McConaughey’s AI agent — because, of course, anyone important should be expected to be using them now — books him a table at a bistro that is unwilling to seat him inside during a rainstorm or show him a menu so he can order something he wants. Instead he sits outside, alone and wet with a sopping napkin in his lap, gesturing futilely at the plate of shrimp in front of him that the restaurant has apparently forced him to have.

“The basics of this premise — a restaurant, a table, food, even a booking app — are extremely relatable to almost everyone,” Alan Kluegel writes in a piece for Defector. “What makes this commercial an avant-garde experience is that at no point are the people on-screen relating to these perfectly ordinary things in a way that any human ever has.” It’s a weird window onto a world where consumerism has ceased to function. What kind of restaurant would refuse to seat a customer (a highly recognizable celebrity one, no less) inside an empty dining room during a downpour and then try to force them to eat food they didn’t order? What kind of customer would just passively submit to that treatment? Why doesn’t anyone exhibit a bare minimum amount of resourcefulness? McConaughey appears to have become alienated from his ability to do anything, and agency itself has become an occulted mystery.

It’s also curious that the ad shows us no relations with machines at all; perhaps showing someone consulting with a computer to help them manage their life appeared too pathetic. In a December piece for Wired, Kate Crawford described AI agents as “manipulation engines, marketed as seamless convenience,” but the Salesforce ad boldly shows instead how inconvenient they will be when they go wrong. Is this because the effort that AI agents can save consumers is too inconsequential and anticlimactic to portray positively? Crawford anticipates that AI agents will seem like helpful friends that we can talk to (or bark orders at), that will “support and charm us so that we fold them into every part of our lives, giving them deep access to our thoughts and actions.” But in the ad, McConaughey sits worthlessly alone, and who can say if the tears on his cheeks are from laughter?

The situation presented in the ad is so irreconcilable with the world we customarily experience that Kluegel concludes it must be an attempt to show us the future:

This ad is not a sales pitch; it is a vision of the dystopia to come. The world depicted in this commercial is one where AI has come to dominate our lives. AI will be what intermediates you and other human beings, it will direct you where to go and what to do, it will give you what it decides to give you, its decisions are binding on you and others, its judgment is irreversible, and you will have to sit there and take it.

That is, the Salesforce commercial is an ad for the control society. It isn’t really trying to persuade the audience so much as warn them that this will be implemented and it won’t be stopped, despite the obvious ways in which it deliberately fails people. Basic civilities and accommodations will be automated out of existence, so that they can be sold back to atomized individuals as perks or bonuses — a sort of Spirit or Ryanair approach applied to the entirety of socioeconomic life, necessarily administered by machines because most of us can’t be demoralized enough to consistently and reliably treat fellow humans beings with that kind of contempt. (This ambition animates the current campaign in the U.S. to fire civil servants and replace them with “AI,” as detailed here by Erik Salvaggio.)

So the ad’s pitch of an “inscrutable solution to a non-existent problem,” as Kluegel puts it, works as an alibi for its actual aim: presenting a world in which compulsory booking apps and other modulating meshes of dividuation and permissioning are a fait accompli and even the very rich and famous default to and proudly flaunt their learned helplessness. This is a post-consumerist world, in that we aren’t expected to shop our way toward having a recognizable personality. Instead identity is expressed by what the automated decision-making systems allow you or compel you to do in front of everyone else. It is the counterpart of the post-labor vision of the world where workers have no agency or bargaining power, and persist as superfluous appendages to a system that no longer needs them. That vision will never be achieved, but capital will never cease to aspire toward it: a fully automated system in which no needs are engendered or fulfilled and nothing at all happens except capital automatically and irresistibly reproducing and expanding itself. (An all-powerful AI god is a version of this fantasy too, capital uninhibited by human inhibition or even the limits of human greed.)

This is a slight departure from earlier AI ads. A Bluesky post from Kevin Kruse (which Kluegel cites) points out that “every commercial for AI is basically, ‘Look, you’re a fucking moron, but we’ll help you fake it!’” The pitch for AI agents must take this further, because the goal is not to dupe others but to fool yourself. You need to be convinced that it’s a luxury to have machines that you don’t understand, operated by companies that profit from your ignorance, make decisions for you about what you can and can’t do. If you’re not doing something moronic and humiliating because “AI” made you, do you even have a right to be here? Are you even trying to keep pace with our brave new world? Do you really think people like you, with your souls of brass and iron, will still be authorized to make decisions for yourselves anyway? The AI assistants the Salesforce ad presumes will be ubiquitous are better understood as automated minders, incontestable guardians who will tell you what to do, along with the necessary lies to keep you complacent about it.

But these may not be the ordinary sort of lies to which centuries of ads and propaganda have habituated us. Consumerism is premised on the idea that nothing is more gratifying than exercising choice in a robust marketplace of options, and then having the opportunity to display those choices and what they signify about you. The point is to impress other people with the choices you can show that you’ve made. But technology has been pushing people toward surrendering agency and self-expression in the name of a more solipsistic comfort. Ads for technology are meant to help with this transition away from conspicuous consumption toward conspicuous compliance.

“The emergence of personal AI agents represents a form of cognitive control that moves beyond blunt instruments of cookie tracking and behavioral advertising toward a more subtle form of power: the manipulation of perspective itself,” Crawford argues. An agent positioned to make decisions for us “infiltrates the core of our subjectivity, bending our internal landscape without us realizing it, all while maintaining the illusion of choice and freedom.” The more one cedes the capability to want things, to choose things, to evaluate things to forms of automation, the more one will become incapable of taking any sort of agency, even as that experience will be framed as liberation from the burden of having to have personal desires. Better to let a machine choose goals for you that it can also meet for you than to risk even a single moment of disappointment or nonfulfillment.

“Convenience is the site of our deepest alienation,” Crawford writes — a slogan fit for a Jenny Holzer–style projection on the side of a skyscraper — and this alienation appears as a diminished capacity to will anything. Convenience is another name for lost resolve; it’s a squeamish aversion to agency that frames surrender as satisfaction. The Salesforce commercial proceeds as though everyone’s already given up on agency and can only dither over which agent can produce for them the least miserable, least effortful life.

Kluegel wonders if the ad doesn’t deconstruct itself when it shows McConaughey being rescued from the AI world by Woody Harrelson, who invites him to come join him at a restaurant across the street. “Harrelson reaches out to him with love, beckoning him back to humanity. In doing so, Harrelson gives McConaughey that which had been taken from him, that what he could not see he needed: a choice.”

That seems too optimistic a reading. I think the ending mainly contributes to how the ad tries to pummel us with irrationality and maintain the suspension of parsable cause and effect. Audiences are supposed to register the “happy” conclusion and not worry about the means that support it, just like AI agents are supposed to save you from having to understand how any of life’s systems fit together. Choices aren’t part of the program. Tech companies are so confident of imposing AI on the world that they don’t believe that they have to bother with making a coherent case for its usefulness. The refusal of coherence is itself the central message. There are no more explanations, no more messages to send.