You Will Die Mid-Scroll

The vibes are off

Allegory of Fire by Jan Brueghel the Elder, 1608

I. Mark Zuckerberg

You will die mid-scroll. Think about it: what’s the second most time-consuming activity you do every day after sleeping (if you’re not an insomniac)? Work? You gotta be kidding me; unless you’re a skipjack tuna fisherman tuning in on this blog from the Indian Ocean, or a newborn nurse on her seventh delivery tonight (that’s my brave, hard-working girlfriend right there), or a chess prodigy on your road to becoming a grandmaster (I’m binge-watching Magnus Carlsen’s matches), the truth is you’re working much less than you’d admit to your boss or your friends, and surely to me.

The label “fake email job” is not a dunk on people’s lack of ambition or willpower (I do work one, and it takes a substantial amount of determination to endure the perpetual hype machine), but an honest reflection of the office-centered service economy: most of us are being indirectly subsidized by the government’s need to keep us employed. We are proof that universal basic income works insofar as it’s well concealed, like a placebo but for work. So no, after sleeping, the thing you do most often isn’t work.

Let me check my phone: “Monday, September 22th. Screen time: 5 hours and 55 minutes. Substack (1h, 38min), X (1h, 27min), YouTube (49min), WhatsApp (42min), see all sites.” This is it. Almost hitting the six-hour mark, which is more than most tuna fishermen or newborn nurses will sleep tonight, the thing I do most is scroll my damn phone. And you, dear reader who got to this post scrolling yours, are the same.

As things stand today, if you don’t change—and provided you are not a regular practitioner of extreme sports other than chess—here’s my safe bet: you, like me, will die mid-scroll. But Mark Zuckerberg doesn’t care. That’s why Meta created Vibes. For those of you still in command of your impulses and thus not chronically online, Vibes is a new feed (or app) that exclusively features short AI-generated videos you can cross-post on Facebook and Instagram. Your first question, like mine, is “why?” Your second question, like mine, is “who the hell is this for?” I will try to answer both.

If you know Mark, and everybody does because he’s the poster child of the self-made young tech genius, you know he doesn’t care about anything except one thing: he wants to win. The how and why are mostly irrelevant. He failed with the metaverse, but he’s relentless, or rather, relentlessly fickle—the epitome of extreme agency toward nowhere. So he focused on AI (his company had a strong AI presence already, but Zuck’s attention was elsewhere). However, he didn’t prioritize the kind that his right-hand man, and Turing Awardee, Yann LeCun, supports (world models, JEPA architectures, and whatnot). He favored the kind that sells: generative AI. So he bought a lot of computing resources and tried to win the community over with Llama open models, but he failed again because winning at AI requires two things: money and principles, and he has yet to realize that they’re not interchangeable.

So Zuck prepared a final blow earlier this year: he put together a dream team to build AGI and then superintelligence (but the personal kind, lest he loses sight of Meta’s original mission: connection) and then the singularity and whatever other buzzword might be in vogue, he’d build it, too. He hired fellow wunderkind Alexandr Wang as part of a $14 billion deal to lead it, poached a few researchers from other labs, especially OpenAI (I think his fight against Sam Altman is twofold: OpenAI is the leader, and CEO Sam Altman is one year younger than him, which is unacceptable). It seemed, just for a fleeting moment, that Meta had a chance of victory by brute force.

But so far, Zuck and his galactic team have little to show for the investment, except vibes and Vibes. Plenty of AI people have already piled on Meta for releasing an AI video feed, as you can judge for yourself from the quote-tweets to Wang’s post—how can one choose “vibes” as the name and be so inept at reading the room at the same time?—so I don’t really have to add anything in that regard for it to be clear that not even AI people want this shit. It’s just Wang and Zuck and those poor researchers who sold their souls for a fatter paycheck. I will say, however, that if you have any doubt about whether AI art is art, Vibes is where you must go to find the answer.

But be sure about this: You won’t find Zuck there. He sells the poison; he doesn’t consume it. “In the blacksmith’s home, a wooden knife,” we say in Spain, but the smith doesn’t forget to bring one because he’s ascetic or absent-minded; he knows that a steel knife will chop your finger off. So Zuck prefers to train mixed martial arts, read books, and engage in yearly learning challenges instead of scrolling. You, on the other hand, should let vibes flow and eventually, in some forsaken dwelling, die mid-scroll.

However, and I feel a certain disgust in defending Meta, they’re not alone in this, and it’s unfair to pretend so; they’re just the least PR-savvy. Google has quietly added generative AI features to YouTube and forces short videos on us all the time (both human or AI, who cares), and proudly announces that we watch them, collectively, 200 billion times a day. OpenAI, for its part, is going to put ads on ChatGPT and also give it the ability to start chats (they call this feature Pulse, as in “Hey, Impulse! ready to engage in your fave distraction?”). Just picture the invasiveness of a friend who texts you in the morning like “Yo, I’ve been thinking—have you tried the new matcha late-flavored protein powder? You have it available in shop.GPT for ONLY $99.99! Click the link below, or I will stop responding for 24 hours :).” No wonder you hate your friend.

Anyway.

II. Alexandr Wang

Who is this guy? No, really, who is this guy? The title that shows up when I search his name on Google is “youngest self-made billionaire” (Zuck is facing great competition these days). At 24 years old, no less; how did he do it? By founding Scale AI and growing it into an empire dedicated to outsourcing data labelling work to the third world through worker-facing subsidiaries. Ok, so he’s the worst person on the planet, then, a more fitting title for the guy “excited to share” Vibes.

He goes and announces it on X as if it is a good thing! As if Meta is winning! As if this is what the world wanted or needed! What the actual slop. But it makes sense that he’s diligently pushing this poison into the feeds of 60-plus-year-olds (the only demographic left on Facebook) when you listen to him:

Basically, I wanna wait to have kids until we figure out how Neuralink or other brain-computer interfaces . . start working. There are few reasons for this: in your first seven years of life your brain is more neuroplastic than at any other point in your life by an order of magnitude. I think when we get Neuralink, kids born with them are going to learn how to use them in crazy ways; it’ll be a part of their brain in a way that’ll never be true for an adult who gets a neuralink hooked into their brain.

I get it, he wants his kids to have a chip in their brains so they can scroll Vibes videos without a phone. Genius! So, when trying to answer “who is this for?” The immediate answer might be: Alexandr Wang and his family. But that’s false, for although he’d be willing to drown the world in slop, he, like Zuck, would never fall victim to his own.

Vibes’ target demographic is the people who have given up the fight, who can’t resist the Infinite Jest; they’re The Matrix’s Cypher, eating that juicy steak as their brains melt inward, serving as battery supply for machines. They’re people who have accepted that we are, at best, the germ of the final species, AI. And who is that? Why, it is clear: 60-plus men and women who, for some unfortunate reason, decided to make an account to spy on their kids on Facebook some two or three decades ago and now inhabit that dead morass of bots, like trapped by an uncontrolled fire that burns their bodies but they can’t put out because it’s mostly invisible to their untrained eyes.

They unwittingly accept we’re the caterpillars to AI’s butterflies. Humanity is, as of now, in a chrysalis state; the larvae will feed on them first. Not even boomers are naive enough to believe that AI will fill this God-shaped hole in our hearts, but, having lived for so long, they’re certain humans won’t either. They may as well pass over the torch.

III. Richard Sutton

Sutton belongs to Vibes’ target demographic, but he doesn’t hide. He is 68, but he doesn’t need Facebook to fall in love with the idea that humanity is, merely, a disposable seed for true intelligence. Reason, not vibes, got him there.

He’s a giant in the AI field, well known for having authored the landmark essay The Bitter Lesson. Less well-known but still considerably so for being the father of reinforcement learning. And much less known than he should be, for being the most accepting among his peers of humanity’s fate. He’s been talking about AI’s ascension and human replacement since before ChatGPT, so, unlike Zuck and Wang, he’s a man of principle. Here’s an example of his thinking from July 2023:

In the ascent of humanity, succession to AI is inevitable. This would be the next great step: technologically advanced humans and then, AIs would be our successors. Inevitably, eventually, they would become more important in almost all ways than ordinary humans . . . this need not be viewed as bad in any way.

Another example from last week, on the Dwarkesh Patel podcast:

I do think succession to digital intelligence or augmented humans is inevitable. I have a four-part argument. Step one is, there’s no government or organization that gives humanity a unified point of view that dominates and that can arrange… There’s no consensus about how the world should be run. Number two, we will figure out how intelligence works. The researchers will figure it out eventually. Number three, we won’t stop just with human-level intelligence. We’ll reach superintelligence. Number four, it’s inevitable over time that the most intelligent things around would gain resources and power.

He is, of course, not alone in this. Dwarkesh agrees with him, also Sam Altman, OpenAI CEO, and Demis Hassabis, Google DeepMind CEO (although he’s shier about it), and probably also AI godfathers Geoffrey Hinton and Yoshua Bengio (or they wouldn’t fear that “if anyone builds it, everyone dies”). (Clearly, not everyone is as accepting of humanity’s fate as Sutton.) Here’s what Jürgen Schmidhuber (62 years old, also Vibes’ target demographic), a pioneer of deep learning (although he insists he’s the pioneer), said the following in a TEDx talk in 2012 (prehistoric by AI standards):

. . . don’t think of us vs them; us, the humans, vs these future super robots. Think of yourself and of humanity in general as a small stepping stone—not the last one—on the path of the universe towards more and more unfathomable complexity. Be content with that little role in the grand scheme of things.

So: Don’t view it “as bad in any way” that we “ordinary humans” will be less important than AI, for the succession is “inevitable.” Instead, “be content with that little role in the grand scheme of things.” Our greatest masterpiece, according to Sutton, Schmidhuber, and the Universe (she’s probably tired of our petty quarrels), is having given birth to AI. Be grateful, for we had a good run.

In Sutton’s view, and rather surprisingly, it’s large language models (LLMs)—Silicon Valley’s favorite brainchild—that are not having the best run (his arguments on the Dwarkesh podcast were thoroughly dismissed by the LLM community that once hailed him). But how do we square Sutton’s comments with the fact that LLMs are the foundational element that Zuck needed to make Vibes? No, wait, actually, it’s the other way around: Vibes is the ultimate app you can create with LLMs. They are peak modern tech; LLMs are merely a stepping stone. Vibes is the combination of an LLM that uses tokens to predict more tokens with a sorting algorithm that uses behavioral cues to predict what piece of media will result in the highest engagement. If AI is our masterpiece, Vibes is AI’s masterpiece. Who could have guessed it!

But Sutton, against the consensus, considers LLMs a dead end, not sufficiently “bitter-lesson-pilled,” as Dwarkesh put it. They don’t have goals or intelligence, Sutton says, but merely imitate what humans do. To the extent that they can learn world models, they do so like shadow-watchers in Plato’s Cave. Rather than our ascended superiors, they are the local minima we can’t escape. They can’t even experiment in the physical world; the most advanced LLM-powered robots can barely walk, much less run.

IV. roon

roon—pronounced /ruːn/, not /rʌn/—is the lowercase anon name of an anime-profile picture Twitstar member of technical staff at OpenAI. In other words, he builds AI.

He is, like Zuck, a visionary. Like Wang, a young, smart lad. Like Sutton, a rare gem among his peers: what kind of STEM software engineer guy is interested in the humanities? Fortunately, unlike Zuck, Wang, and Sutton—and those unwitting elders at Facebook—he’s measured in his post-AI form of discourse, in the Venkatesh Rao sense, which means that he accepts and embraces a world in which AI has surpassed humans across domains, including—or at least I hope he’s come to terms with including—the arts and the humanities, but he’s not ready to pass them the torch or, conversely, douse it into a toxic swamp, vibrant as it may prove to be.

This is what roon said about Vibes, which led me to reinterpret his views on AI altogether (I disagree with half of it, but let’s focus on the parts I agree with):

. . . feeds that contain human-generated things, machine-augmented content, and later autonomous machine generated content will launder some of the ‘cult value’ of humanity’s production into the latter objects too. famous artists may make things where they have little input and most of the generation is mechanical, but the original reason for their fame would have been that they made things that were alive and good on their own merits.

He’s pro-human! He doesn’t like Vibes for the same reason the other AI insiders dunked on Wang—stop the slop—but also because he seems to believe that art things are good insofar as they carry humanity’s imprint to some degree. (He’s started a weird crusade against “moral panic” around short-form video but I think he’s merely arguing that short form is equally bad, whatever the source.) He’s trying to build LLMs that write well, but he doesn’t seem to want them to surpass human prose, merely complement it, just like digital art complements photographs complement oil-on-canvas painting. To roon, co-existence is the paradigm, not succession or replacement.

He’s failed to make LLMs write well so far, though (probably for the very reasons Sutton enumerates). The best he can offer is the occasional joke about an impossible quest or an unfruitful defense (he mentioned the line “I am nothing if not a democracy of ghosts,” which is fantastic, but turns out, as Erik Hoel revealed, that AI “lifted it from [Vladimir] Nabokov”). I personally don’t think an LLM will ever (and with “ever” I mean “short-term” because it makes no sense to make predictions about the end of the world) write, for instance, like blogger Scott Alexander (I don’t think he’d agree, though; he’d likely take the optimistic stance if I take the pessimistic one). To do what he does, you need to observe humans in their real-world habitat for a long time and preferably be one yourself. Unfortunately, AI is disembodied and unfamiliar.

Among humans, Scott’s talent is rare. Even among observant humans, it’s rare. But so is world chess champion Magnus Carlsen and Garry Kasparov’s talent, and the last time I checked, AI was ultrahuman at chess. So you never know. Anyway, roon said something else on that tweet that got me thinking (and inspired this essay):

. . . the art that we actually like is rarely the most ‘technically adept’ or even the ‘most tasteful’ (whatever that means) but rather whatever builds cult value in the Walter Benjamin sense. Techne is an aspect of cult value formation, but not all of it.

What moves people, he says, isn’t technical polish or even beauty but the sense of belonging an artwork carries; the story that connects creator to creation to world. Techne (craft, skill, substrate) contributes to that but it’s not enough; what matters is the cultural charge, the feeling that a piece means something beyond itself, that it belongs to a special circle, moment, or reverence, just like a kid’s “crude crayon drawing of their family is priceless to her parents,” to use roon’s intuitive example.

He cited philosopher Walter Benjamin, so let’s move on.

V. Walter Benjamin

He writes:

. . . photography freed the hand of the most important artistic functions which henceforth devolved only upon the eye looking into a lens.

Throughout history, we’ve steadily abstracted away the human from the artistic process due to the increasingly complex technological involvement. Vibes is the ultimate form of this abstraction: the human is forever gone. In that sense, I imagine Walter Benjamin would’ve found Vibes fascinating. And even more because he would have recognized that, unlike photography, it displays a striking dual nature.

Cult value, as roon argues, is absolutely lacking in automatically generated videos; however, it is huge in the sense that Vibes only exists as an extension of LLMs, and these are, above all, a technology of cult. An LLM is an element of ritual par excellence in Silicon Valley and AI circles; an object of veneration with a powerful aura. Except for Sutton and other heretics, AI people believe they’re our best chance to engineer a mechanical God. They check all the boxes: AI labs act like modern prophets—that’s why they’re pouring billions into building temples for LLMs—and keep them hidden from the masses in the Cloud—their Heaven—like relics hidden from pagan eyes.

At the same time, LLMs enable a sort of artistic reproduction—stochastic in resemblance to an undetermined original version, and infinite in quantity—that makes LLM-based visual content the quintessential example of exhibition value. The existence of Vibes, an endless chain of AI short videos created by LLMs, is contingent on their being exhibited to a larger degree than any other art form, even more than photography and cinema, which keep a glimmer of cult-like devotion. AI videos are not reproductions of any particular human-made object, which breaks any connection with tradition. Vibes is context-less content; without exhibition, it has no existence.

These two facets—LLMs as supreme cult value and AI videos as supreme exhibition value—are inextricable: Vibes has come to exist only because LLMs are ritualistic elements (or there wouldn’t be enough compute resources to launch it in the first place) and because infinite random reproducibility of visual art is now possible.

Walter Benjamin realized that the artistic function of a work of art can be “recognized as incidental” as other functions emerge, and that mechanical presence accompanied by human abstraction in nothing takes away from the value of a work of art, but he didn’t entertain the possibility that all these functions could collapse into an ultimate art form enabled by the most advanced technology ideated by humans, without humans playing any role whatsoever (prompt-making is also automated):

. . . the work of art in prehistoric times . . . by the absolute emphasis on its cult value . . . was, first and foremost, an instrument of magic. Only later did it come to be recognized as a work of art. In the same way today, by the absolute emphasis on its exhibition value the work of art becomes a creation with entirely new functions, among which the one we are conscious of, the artistic function, later may be recognized as incidental.

But of course, how could he have predicted this? Benjamin was more prescient than he could’ve imagined (the essay I’m quoting was written in 1935), but his gaze was inevitably limited by the impossibility of foreseeing AI, LLMs, short-form videos, and Vibes in the early-mid 20th century.

The artistic function of AI videos is not even incidental, but negligible. No piece of art is worth billions in datacenter investment in Zuck’s eyes. Keeping 60-plus-year-olds doing the same thing for hours on end is not art—and no easy feat—and yet worth the cost. Until now, only commercial TV had managed. But even TV shows have some grounding in reality. The watching part mattered insofar as your feeling of leisure required sequences of events and plot points and plot twists, and characters.

Vibes is not like that. Vibes is a superior form of that, qualitatively distinctive. Vibes has both cult value and exhibition value, yes, but also something else. “Watching” itself doesn’t really matter. That’s why we can’t stop at either “cult” or “exhibition”. New functions emerged that go beyond that, as Benjamin wrote: a distractible quality, an addictive power. To capture that into one overarching concept, we can say that the fundamental function of Vibes is as a tool with zombifying value.

Not even Walter Benjamin could’ve seen that one coming. Sam Kriss did.

VI. Sam Kriss

Sam Kriss, master essayist, dwells on it in “In My Zombie Era.” We are zombies; we no longer consume algorithmically-sorted content, but the sorting algorithm itself.

The videos themselves are only interesting insofar as they belong in the middle of an infinite chain. Your brain doesn’t yearn for you to consume content but for you to scroll past content. What you’re watching at any given moment in an algorithmic feed like Vibes is the least interesting thing ever. Think about it: When a short video begins, your brain immediately begs you to scroll. It wants the high from randomized rewards—will the next one be good or not?—but it doesn’t want to know the answer, just a reason to ask the question.

I compared the zombifying power of short-form AI video platforms with the human equivalent (e.g., TikTok), and this was my conclusion:

Unlike human-made TikTok videos that scream, “hey, care about me, I’m also important,” AI slop doesn’t try to convince you; it’s just there, sufficiently satisfied with its role as space-occupier in the infinite chain. It’s pure placeholder essence, pure liminality . . . . AI slop’s ridiculously huge success entails the subconscious acceptance that, in a world of digital abundance and perpetual change, things only matter insofar as they embody prior-ness [that is, existing as precedence to something else]. No pretense of self whatsoever. No pretense of meaning. No pretense that watching matters, only scrolling. AI slop videos are the perfect appetizer for a generation of zombies that, having been conditioned by the sorting algorithm, have ended up like Pavlov’s dog, salivating at the ring of the bell.

How did we get here? Kriss explains that nerds like Zuck or Wang “are people who love things because they exist.” Vibes fits his definition perfectly: It exists, and that’s pretty much it. But perhaps the most relevant question is: how do we get out of here?

You need to know that to the extent that the zombifying value of an artwork machine supersedes the other values, you are the object: if cult value dominates, the artwork is the object and you the subject who worships it. If exhibition value dominates, the subject-object relationship is mostly unimportant; the prestige of the Louvre doesn’t change with the influx of visitors. But when the zombifying value dominates, then you are the object being quietly subjected to torture by the perpetual slop machine.

I’ve long suspected that Kriss’s command of the English language and the playfulness with which he switches style from essay to essay is not so much a by-product of his evasive attachment or a targeted tactic to dissuade an unwanted audience, as it is an intentional attempt not to be swallowed up himself by the toxic swamp of fashionable currents, where everyone else happily wallows. “To be sane in an insane world, you yourself need to be insane,” said many thinkers across the ages.

So Kriss is rocking the only suitable armor to oppose the dehumanizing forces around us, for which Vibes serves as a stand-in. He only logs in if he has his safety rope tied around the waist and fastened, at the other end and with a triple knot, to a beam in his offline hermit hut in the mountains, for he knows that when the moment comes—after too many hours spent in the mire—he won’t have the strength or the will to get out on his own. That’s when he’ll pull the rope.

That safety rope is his irreproducible character, his blending of fiction and non-fiction without a warning label, his resisting the zombification, his embodying the Antivibes, his not passing the torch, his “glimpsing at the eternal things.” In other words, it is his form, not his content, that keeps him alive. He won’t be caught abstracting himself from his creations. For he has seen the destiny of the world, and he’ll do what it takes to resist it. That’s why he, unlike you and me, will not die mid-scroll.