AI Content and the War for Your Attention

Does AI so ruthlessly optimize for what you will pay attention to? It totally alienates you from what you want to pay attention to.

Everyone kind of carries themselves like they are a Kardashian everywhere they go. In the future, we're not all going to be famous for 15 minutes. We're all going to be famous to 15 people.

If it's on every waking second, you can now mine that attention and extract and commodify it in a way that wasn't possible before.

But it's only going to be waking hours, Chris—16 hours. We're going to need full Neuralink to actually put ads inside your dreams. Actually, we're going to need the full 24.

Well, I'm so excited to have you on, Chris. Antonio is a longtime friend and collaborator of mine. We have a show. Antonio, why don't you briefly introduce yourself?

Yeah, I started in ad tech, I guess, in 2008, in attention capitalism, as the press would call it. I worked at a number of startups. I was an early member of the Facebook ads team, back before the IPO, when the ad system was horrible. For those old enough to remember—I suspect Chris is probably in the same bucket—Facebook ads used to be these little postage stamps on the right-hand side, and then it became this huge money machine.

I wrote a book about that time called Chaos Monkeys, which came out in 2017 and did really well. I was a media figure, writer guy, for a while. I don't know how you do it, Chris. It drove me basically crazy. And so I went back to tech.

Antonio famously says the prize for beating Matt Yglesias in the Substack rankings is that you then become Matt Yglesias, and that wasn't inspiring enough.

I'm not pretty funny. I'm not anti-Matt, but we were briefly among the top 10 or 15 Substackers on a revenue basis circa 2021, and he was definitely a comp. I guess he and Andrew Sullivan were definitely the winners of that race. I feel like I would probably have more to learn from you guys than you have from me, so I'm happy to talk, and maybe I'll ask you some questions.

I sort of think the whole thing is about to collapse. I think there's just a huge pollution problem. I think that the treatment of spam in the book is the signature pollution of attention capitalism, of the attention age. If it's lucrative to aggregate lots of attention where attention pools and collects, there will be money.

To the extent that you can cheaply automate the extraction of that attention, even if it's done in a brute-force, wasteful way, there's going to be spam, right? That's already happened everywhere: it's happened with junk mail, phone calls, text, and email. But I think the idea of AI content creation now puts that at scale for all the social media platforms.

I guess the question becomes: Is AI-generated social media content actually compelling enough that it doesn't feel like spam and just dislocates all human creators? Do people not like it, so the algorithm stops serving it up, or does it kind of overwhelm the experience in a way that starts to feel like the way our inboxes feel?

Is AI slop going to win? Yeah, is AI—that's one way of saying it—will AI slop win?

I mean, yes and no. You're seeing it in the For You tab on Twitter already, right? Maybe I would think this because I was a creator paid for my creations, to some degree, but I think a lot of it is empty and sterile. I think humans who are able to be creative and use AI tools correctly will have superpowers.

Yeah, I'm pretty obsessed right now with AI slop, and particularly these subgenres that feel like—what's funny about a lot of AI is that it doesn't actually feel that sophisticated. It just feels reverse-engineered in a way that you can trace. Not all of it, right? But some of it does. There are moments we all have with this technology where you're like, “Oh my God.”

For instance, it's clear that religious themes and babies do well, right? So there's this whole universe of AI babies singing, “Bless the Lord, O my soul.” There was one I saw yesterday with Jesus and a baby; they're both singing “Bless the Lord, O my soul,” and Jesus is holding a disembodied foot, which is just signature AI slop.

To me, the question about all this stuff is—one of the theses of the book is that there's this kind of disconnect between what we will pay attention to and, in some volitional sense, what we want to pay attention to. Does AI, from the generative perspective, and AI from the machine-learning, algorithmic perspective, ruthlessly optimize for the former—what you will pay attention to—and totally alienate you from the latter, what you want to pay attention to? If it does do that, do people stick with it, or do they reach some point where they feel too alienated from what they want to be spending their time doing?

Well, I think the challenge there is what you put, Chris. By the way, I think it's time we just plug Chris's book. I think you wrote a phenomenal book.

Thanks.

I would recommend it to most of my ad people. One of the things you cite is—to use the marketing speak—acquisition and retention are different things, right? It's one thing to acquire attention. I think you cite the example of someone going into a room and firing a gun in the air. Well, you've acquired attention, but then how do you maintain it? Even with a guy with a gun in his hand, it might be a little hard to maintain it, right? And that's the challenge. If you're a sophisticated marketer, you realize it's not just about acquisition; it's about the retention side of it. I don't know. A positive side of me thinks humans aren't going to sit there and stare at the slop all day. That said, I've used TikTok and, like a methhead getting into a Fenty fold on the street, I emerge two hours later, I come to, and I'm like, "What the hell just happened?" And then I delete the app from my phone because I never want to do that again. So that's maybe the—

No, that's—well, first of all, thanks for the kind words. I appreciate it, especially because I'm not a practitioner in ad tech, and so much of the story is an ad tech story. I think it's deeper than ad tech; ad tech is sort of epiphenomenal to the story.

I've had the exact same experience. That's, again, the defining metaphor of The Sirens’ Call: Odysseus on the mast, trying to avoid the sirens with your volitional, conscious self against the kind of lulling that happens. All of us are in that position, but I think I have the same faith as you: fundamentally, there's something irreducibly human, and people are going to rebel—basically, we'll rebel—against slop.

Which brings us back to this first question, which I do think is really interesting: If the slop overruns it, what does that mean for the platforms? Is it something that they have to start actively managing in the same way that, at scale, somewhere like Twitter, they started to have to deal with a bunch of really difficult content-moderation questions that were tough calls? They're trying to create a user experience that is not going to repel people and is going to attract advertisers. I do wonder if AI-generated content becomes a set of questions like that for TikTok and for Reels and stuff like that.

We've had a slop problem before AI, right? I mentioned LinkedIn earlier, in terms of how fake that network feels. But also on Instagram, people have Finstas—fake Instagrams—for a reason. So this idea of who you present yourself as, who you actually are, and what you actually think has been a problem with social networks for a while.

It's not obvious to me that AI will make the problem worse in terms of accelerating or emphasizing either fake stuff or stuff that people don't want to be engaging with. Why are you so convinced that it will expand that gap between what we want to be engaging with and what we actually engage with?

I just think, to me, it's a question of quantity and automation, right? One of the things about online success is that it's not a batting average. It's your total number of hits. I mean, it's a batting average if you're trying to make a living, right, and you only have so much time. But MrBeast talks about this. Other people talk about it, too: always be posting and see what works.

Sometimes you can post the same things 15 times, and one time it takes off and other times it doesn't. If you're thinking about it as this quantity game, right, this is where the brute-force spam analogy comes in. If you're automating 1,000 videos every day and just throwing them at the algorithm for content-farming purposes, does that start to overwhelm everything?

You can just do it. I can't make 1,000 videos a day, but you can now, right? Antonio, to your point, you could even imagine this not as a slop-spam thing.

You can imagine it as someone who's actually pretty good, right, using that automation. To me, it's the kind of brute-force scale issue. Maybe it never comes up, but it seems to me like I'm starting to see it crop up, and I just wonder what it's going to do to the existing models for these platforms.

One thing I find interesting is the impact of social media. You cite a lot of the negative impacts of social media, and you cite one book, “The Frenzy of Renown,” that I actually started reading based on your recommendation.

It's a great book, isn't it? I'm going to be that guy myself: I think I have a line in “Chaos Monkeys” that says, “Andy Warhol was wrong. In the future, we're not all going to be famous for 15 minutes. We're all going to be famous to 15 people.” Yeah, and it's a great line.

Yeah, and the fact that everyone carries themselves like they are a Kardashian everywhere they go. As a tourist in Europe, the fact that you literally have a line of people taking selfies in front of the thing already shows that people are imagining themselves under the spotlight.

I work at a tech company. It's very young, and I think the 1 thing the Zoomers themselves cited is that you feel like you're being observed all the time. Everyone is like a celebrity who goes out in the wild, and if you yell at a waiter or do something stupid—which we've all done, by the way—you can suddenly become Insta-famous for it, and nobody wants that.

I think Antonio and I are probably old enough to remember that, before, when we were young, dumb kids, if you didn't go to jail and didn't go to the hospital, it didn't happen. That was it; it was gone. Maybe it remained as local lore, but it never got bigger than that. You weren't really worried about anything, and somehow that's totally changed the way that people think about themselves and the world.

Yeah, and I think one of the things I write about in the book a lot is the effects of attention and social attention on you and what it does, and that democratization of fame as a genuinely new phenomenon that's different from past eras. I think you're right: It is behavior-modifying.

For me, fame is always totally relative. Different people are famous in different places, but people recognize me on the street. I do have that awareness that, for instance, if I screamed at a car that cut me off or almost hit me as I was rounding the corner, people might know that was me and associate it with me.

But there's also something profoundly warping about it psychologically. That point you just made is a really interesting one: There is some kind of prosocial effect of not wanting to be a complete maniac in public because that might go viral or something. That's an interesting prosocial aspect of it.

I want to get back to the group-chat thing, because I think toward the end of your book you become a little bit of a doomer, Chris, but you start trying to suggest solutions to the problem. I think the group-chat one is an interesting one, but I want to focus on one thing you said, which is: Is it prosocial?

One thing I noticed right after I wrote “Chaos Monkeys”: I lived—I kid you not—on a small island in the Northwest, in what's called the San Juans, which are these beautiful islands. Yeah, I've been there. It's beautiful. Yeah, it's a cool place, a very small town.

The reality is everyone's kind of famous in that neighborhood, and everyone knows who everyone is and who their family is. If you did something totally crazy and off the deep end, you'd still catch flak for it; you wouldn't get away with it.

A lot of our focus on privacy—in fact, this is historically true, and I know this is a bit of a rabbit hole—comes from the history of privacy. Louis Brandeis basically created it in 1890 with this disquisition called “The Right to Privacy,” which he more or less invented. By the way, the word “privacy” doesn't appear once in the Constitution. The reason why is that, at the time, it wasn't used in the sense that we mean it: the right to live as a stranger among strangers. That right was more or less invented to cope with urban living, where you were losing the ability to actually know everybody.

Long windup to getting to the group-chat phenomenon: Not everybody can join the group chat. There's an admin. If someone gets completely out of control, they get booted, and then everyone kind of knows everyone. You don't really want to piss people off because there's typically some theme—at least in my group chats, often professional. I'm in a group chat with Erik and a lot of people who have worked together or invested in those companies. You can't be that much of a dick, to be blunt, because you'll take a hit for it. So it's kind of self-regulating.

If I were to propose a solution to our current bind, it's not less technology and a Luddite jihad against this thing. I think it's using technology in some way to recreate—

I totally agree with that. I do think the scale ends up being the problem. The question to me is: Where is the revenue in that model? This is the interesting question. It's something I'm really obsessed with, and it's the germ of something that I explored in the book and am writing about now in relation to AI.

We tend to conflate, particularly in this era, useful tech and lucrative tech, and those are not the same. There are technologies that are incredibly useful but not particularly lucrative, like penicillin. Antibiotics are not something off which anyone is making enormous fortunes. Even solar power, which is arguably the most useful technology of our age, is a perfectly profitable enterprise, but no one is making Rockefeller fortunes off solar right now.

Then there are technologies that are incredibly lucrative but not particularly socially useful. The FanDuel and DraftKings tech is really good tech, and it works really well, but it's not particularly useful, especially compared to penicillin.

One of the things that happens in a digital space so dominated by commercial options is that you don't get as much useful tech that might not have great revenue or a business model. The question of the sustaining business model of the group chat is an interesting one.

Obviously, WhatsApp, which Meta owns, is an incredibly valuable company—an incredibly valuable division of Meta. Then there's Signal, which is fascinating because it's a nonprofit and, I think, a really interesting model in that respect.

To your point, I totally agree about reclaiming community, reclaiming Dunbar's number or IRL interactions. What I say in the book is that normal relationships are born of bilateral exchanges of social attention, as opposed to unilateral ones. That's the way we're all conditioned to work. The question is: Can you do that at scale? Is that a particularly lucrative technology to reclaim? To me, that's the question: Is there some fundamental tension between what the revenue-model incentives are and what the tech is best for us as people?

One take on that, Chris: Just to give you the view from the crypto trenches, because I work at a crypto company, group chats are actually huge in the crypto space. They're usually on Telegram, which is an app that most normal people don't use—or most Americans don't use, certainly.

There, you either pay to enter the group chat, or some apps have literally tokenized access to the group chat. You have to buy a certain number of the tokens, and then the owner of the group chat—the person who's leaking the alpha, because this is all very speculative—the token represents his value.

There's a company called friend.tech that was short-lived. I was part of it for a while, and I had a channel. You could buy the AGM token and join the group as part of it.

It feels like—but I think the lesson there is that if you overfinancialize things, you lose the social side of it, which crypto does a lot, right? It makes everything—AI is everything computer, and crypto is everything money. Literally everything.

But, yeah, I agree with you. I mean, one place I'd maybe quibble with you a little bit—or I guess I don't know; I'm confused myself. I think the world we're heading to, the digital version of things, is kind of like the mainstream, premium-economy version of the world. And I think our friend Marc Andreessen—you might know him, Erik; his is the A in the sign behind you—famously said that some people are reality privileged. Some people are wealthy enough or live in a certain context in which they don't need the VR headset, because their life is good enough and they don't need it. But everyone else gets the VR headset.

So how do you strike that balance between the digital versus the real versions of things? I think what happens is—and again, I think it's the subtext to your entire book, Chris—who we interact with and who we agree with is completely uncoupled from our actual physical place in the world. I think that tension between the physical and the—I wouldn't say spiritual, necessarily, but the intellectual—is at the heart of all of this.

I think the nation-state is kind of cracking up in a way, just to get doomer for a hot second. I think the internet's definitely behind part of it, because you don't have that sense of collective consensus that the media and TV, for example, used to create during the Walter Cronkite era. They used to have some level of synchrony or synergy that would happen with Cronkite saying, "And that's the way it is." Somehow now I scroll my feed versus Bluesky, and it's like, "No, I don't know what the way it is. Actually, the way it is."

What's really interesting, too, is to think about it from the opposite problem. So much of what happens in the 1920s and 1930s, particularly when you have these huge mass-broadcast mediums that can capture attention at scale, is that you have this opposite problem, right? The problem of mass media is that this is an insanely powerful tool: You can blast out messages at scale to all these people and get them all thinking the same way, or moving the needle in some way. That's a problem that Lippmann wrestles with, and all sorts of different folks wrestle with.

Now we have kind of the opposite problem, which is that there is no getting anything like that. One way to think about a public is paying attention to the same thing together. I've even come to feel weirdly old-man nostalgic for the Super Bowl, or when everyone's talking about Love Island, because it feels like, "Oh, here's a thing that everyone is paying attention to together." I'm not watching Love Island, but I know that everyone is, and I find some weird comfort in that.

I think it's so funny, because 20 or 30 years ago, particularly if you were Gen X and thought of yourself as alternative in whatever way, mass culture and the middlebrow felt stultifying. I've come to kind of miss that, because I think it is the thing we're missing. Football is really the only thing left that everyone pays attention to together.

The elections.

Yeah, there's something lost in that. I don't know how to get it back. I'm also fully aware that it might just be that I'm 46 and succumbing to the nostalgia of middle age.

What was the last sort of literary event that every American paid attention to? My vote is for Jonathan Franzen's The Corrections when it came out. I was literally going to say Franzen and The Corrections, and then the whole thing. Yeah.

Right. Everyone had to read that damn book, and it's a great book, by the way. I like Franzen as a writer. Everyone had to read The Corrections, and since then it's never been the case that everyone's read the same book at all. With the possible exception of David Foster Wallace and Infinite Jest. But it's one of those books everyone claims to read, but no one's actually read. I think people actually did read The Corrections.

It's weird, because if you talk to Gen Xers, and particularly people like Marc who came from maybe not core mainstream America, what they always say about the internet is, "Oh, I finally found my people." They were interested in some weird little niche obsession, and in their little Midwestern town there was nobody doing it.

Although what's also interesting to me, and what I'm always curious about, is the tech.

I feel like I don't have a great sense, actually, of how sophisticated the algorithm is, and maybe you guys have a better sense of this. I think it's very funny that we refer to "the algorithm" as a definite article. It's sort of totemic, almost like a deity: "The algorithm served me this," or "the algorithm..." It's always just unclear to me how sophisticated the algorithm actually is.

Sometimes it just seems very dumb to me. I look at one thing, and then it serves me 15 of that, and it's like, "That is—I probably could have coded that." That's not that impressive. It's not that sophisticated. Sometimes it does find things where I'm like, "Oh, wow. There's something interesting happening here." I write about it in the book, like taking a gummy and realizing that it's just showing me pictures of sandwiches. After 30 minutes of just looking at it, it's like, "Oh, it knows that I'm high." That's a kind of magic.

I do wonder—one of the things I think is interesting is that there will be these kinds of things that go viral or have this kind of algorithmic resonance, that everyone is then talking about or, particularly in Gen Z, referencing. The way that people talk—this sort of brain-rot, Gen Z way of talking—is really interesting to me, because that does seem recognizably mass, right? In the same way that in the 1960s it was "far out" and "groovy" and all these things, there is this really distinct argot of Gen Z. But they're getting it all not from mass culture, but from basically algorithmic social media. To me, that's super fascinating, because there's something enduring beneath the technological or institutional layers of how culture is being mediated, and how attention is being captured and bought and sold.

Yeah, I forget what it was. I recently visited family, and all the kids had the exact same argot as in the U.S. Versus if you had visited Ireland 20 years ago, you wouldn't have been able to understand them, because they spoke a dialect that would have been unintelligible to American ears.

And you see that, actually, in the U.S. I mean, this is the thing that I'm really obsessed with, too: the lack of geographic distinction between cultures. You're starting to see this show up in politics, right? It used to be the case that rural Mississippi and rural Minnesota voted very, very differently. That's because the people who occupied those places had different ethnic backgrounds—Minnesota was, like, Scandinavian socialists—but also different religious traditions: Southern Baptist versus Lutheran. There were all these defining geographic features of different places that would show up in politics.

Increasingly, there is this kind of homogenization happening, where rural folks of certain demographics more and more vote like each other, and also listen to the same music and have the same kind of omniculture. That's also true of big-city, college-educated, upper-middle-class folks. You could go to the cool neighborhood in Columbus, and there's going to be some restaurant called Stem and Stern that serves whatever farm-to-table gastro-pub food. It's interesting that these omnicultures supplant localism at this moment when we don't have massness as the central cultural feature.

The place is called Pine and Spruce in Seattle. But it's exactly what you're saying, and it's exactly right. It's all called Stem to Stern or Pick to Snout, or whatever. It's always this kind of folksy thing, and you can already imagine the distressed wood—the artificially distressed wood—the artificially distressed people, the exposed iron, and the bare filament bulbs.

I think the person who first called this, by the way, was Dan Savage, who was the sex columnist or whatever for the Seattle local rag. In the aftermath of the Bush election, as a total Gen Xer or millennial, he said, "It's not blue states and red states. It's a blue urban archipelago in a sea of red." The culture was actually going in 2 directions, and you see it everywhere.

Again, to quote our friend, the analogy he cites is this: If you remember, back when we had Windows machines, you would do something called defragmenting the disk. What that means is Windows is such an inefficient operating system that a single application stores its memory throughout the hard drive, and then it slows it down because you're reading from many places.

So you'd have to do this thing called defragmenting, which would basically write the memory back to the same place, so the app would start running fast again. What he feels is happening—and, of course, he believes in this thing called the network state—is that we're basically defragmenting reality, right? What we need to solve this is to take all the people who think like each other and get them back in the same state and, in some sense, defrag the disk. I think the realities of that are a little challenging, but it's an interesting thought experiment to imagine that if you were to do that, in some magical way, then you would end up with a wholly red and a wholly blue country or something.

Yeah. I mean, this question of the individuation of the current attention marketplace and how it relates to what we share and what we don't share—there are just a bunch of really interesting questions in there. Some of which, again, like the point you were making before about “I found my people on the internet”: for a trans kid who's in a place where there's no one else like them, to find on the internet people who are going through this is the most incredible thing, right? And that's true for a million different ways of being that you might be as a teenager, as a young adult, or as an old adult, right?

But there is a level of individuation built into the technology and into the market technology of it now, which is how it all drives toward ever more individuation. But that individuation fails in really interesting ways. Can I ask you a question, Chris, because I'm really curious about this? To me, one of the big unsolved questions that I never quite wrestled to the ground—I talk a little bit about this notion of subprime attention—is that there's sort of 2 different ways of looking at ad tech, to oversimplify.

One is that it has all this incredibly personalized data that no advertiser has ever had access to. It's not just talking to a generic man, head of household in the ’50s. It's talking to you, Chris Hayes: I know where you live and what you do and how much you make and what your medications are, right? That allows it a level of sophistication and optimization that's never been possible before. And then the other side of this is, for all of that being true, it's amazing how bad, schlocky, and ineffective most internet advertising is. I think there's some truth to both of them. I wrestle with them a bit in the book, but never feel like I've resolved the question. I'm curious how you think about that.

Yeah. No, I mean, it's a good point. It's funny: this is going to take us way back to the last time I was on your show, probably because it's what I was on for—the whole Cambridge Analytica scandal. Exactly. We don't need to go down the rabbit hole. It's a whole story. Everyone's probably heard as much as they want to hear about it.

There's no way to do that, right? Any media buyer will tell you that a lot of ads are still crap. It's true. It's still kind of a statistical phenomenon. Just to put numbers on it, the CTR, the click-through rate—how often a user engages with something like Instagram—at least 97% of people who looked at it were just completely indifferent to the thing, right, and didn't engage.

Yeah. I mean, look, you can really do a technical deep dive there. Yes, that data is out there, but a lot of it is super fragmented, and it often exists in varying silos. Then there's drift, and to be blunt, in advertising, they'll sell you data that's not nearly as accurate as you think it is. Companies like Facebook that actually do have a first-party relationship with the user often have better data for that reason.

If nothing else, they have a stable notion of identity: it's the same Chris Hayes or the same Antonio that shows up on the app every time. So they kind of know who you are. If you just show up on a website, it's based on a cookie, and that's so transient these days that it's basically meaningless.

So there's just a lot of practical problems. And, again, it's funny: I actually did a story for Wired about it because it pissed me off so much—the whole conspiracy theory that Zuck is listening to you through your phone or whatever. I don't want to go into a total detour. But here's my question: this strikes me a lot. When you're on TikTok, TikTok has gotten pretty ad-heavy these days, and I'm always struck that every ad feels like something from the Home Shopping Network or an infomercial I would see at 5:30 a.m. before cartoons came on on a Saturday morning in my youth.

I just keep thinking to myself, why isn't Progressive Auto Insurance or GM serving me ads here if this ad tech works? Why are the people that spend the most amount of money on advertising—and for whom this is particularly true, brands like car brands and insurance brands, where there's unbelievable continuity, right? People just keep the same thing. That's why traditionally the younger demographics in advertising were so valuable: you want to sell someone their first Toyota because they're just going to buy Toyotas after that. So you're really purchasing a lifetime of purchases.

If that's the case and young people are on TikTok, if this is such effective advertising and Detroit spends a lot, why are they not there? Why do they never get an ad there? I don't know. Maybe they're just dinosaurs and they're wrong, or maybe it's not as good. But I just don't know the answer to it.

Yeah. I don't know. I mean, it's totally true. There's TikTok. There's an app called Whatnot. It's basically people stumping for a product and then buying from it. I think it works for DTC brands, right? Very niche brands that people associate with and buy very easily.

You're absolutely correct that that's the whole point of brand advertising, right? Again, to use more ad-techie language: high LTV, high lifetime value. If you become a Ford buyer for the continuity of your life—you buy a Mustang and you buy an SUV when you have a family—that's literally lifetime value, you know? I'm not sure, to be honest. I don't know what the answer is.

Yeah. No, no. I don't know if it's a thing to answer. It's just so striking to me that that's the case, and particularly how much that still dominates these older media—that still, when you watch a big sports game on broadcast, if you're watching Sunday football, right, there's a scale there that's very hard to replicate with TikTok.

To your point about how scalable the TikTok stuff is, if you make some nifty little gizmo that you can drop-ship from China and you have low overhead, you can sell it to a small number of people on TikTok. That's not the way you're going to sell a Ford F-150. So there's a sort of interesting scale question there.

But I also just wonder what the next—basically, we have seen these movements from text-based to video, and if there's something after that, what basically the next big thing is. What is the next ByteDance that does to TikTok what TikTok sort of did to the incumbents, Facebook and stuff? I mean, my vote would be for AI, to frame it as a bigger thing, right?

We're going from a textual culture. I think you cite somewhere in the book that the thought that we're going to be reading a 6,000-word thought piece in The New Yorker—I doubt most young people can even do that anymore. I can barely do it, actually, even though I used to do it regularly. I've forced myself to do it, which is weird.

It doesn't mean we're going back to an oral phase in which we're going to be reciting Homer in hexameter or something, right? Because the oral phase, which Neil Postman writes about a lot, was very mnemonic. It was very memory-intensive. People had these recognizable tropes and forms, but they also remembered a lot. They knew they could recite poetry; they could recite biblical verses. There was an enormous amount stored in the brain.

Yeah, and I mean, I think we were at the tail end of that. I still had to memorize Shakespeare in high school, which I don't think anyone has to anymore. But I think the oral-textual vibe wasn't even about the format; it's the form of thought, right?

And when I talk to the AI, it's like I'm having—I don't think they're human, by the way. I think everyone who thinks they're human is crazy. But I talk to it as if it were human because that's actually the best way to interface with it. You engage in this sort of Socratic dialogue with it, trying to get an answer. Sometimes the AI bot gets crazy and hallucinates. I mean, so do humans, and you kind of have to put it back on the right course.

But I think that's the thing, right? Talking to the human, and when you manage to map that to e-commerce or travel, when I can say, “I want to go”—which I do: I go to France for a week in August to see my daughter and spend a week in Brittany. So book me an Airbnb, book me a flight, and make it not cost more than $3,000, if that's even possible, and just give me a button that says, “Go,” and do it.

That's, I think—I would literally bet large portions of my net worth on it, but that's the future.

But that's going to be the gateway. That's how their AI is going to upstream everything about the consumer experience.

I totally agree with that. My bigger question is what it does to what you do on your phone. I agree that there are these very obvious, incredibly useful use cases if it gets good enough—particularly the sort of agentic idea that you're describing. There's lots of stuff.

And the idea of it essentially becoming, in the way the browser was, the sort of interface being the thing that you're talking to. Right. But that's to do all the useful stuff, right? My question is: What does it do when you get that screen-time notification on your phone? What you're doing a lot of the time is not booking flights and doing stuff. Some of it is, but what it does to that part of it is the big question.

I don't have the answer. I don't think anyone has the answer. If you had the answer, you could probably make a fortune with the right bet. But I just don't know what it's going to do there. The place we started was my wondering if it produces a kind of brute-force level of pollution that then actually becomes hard to manage for the platforms; 2, it gets so good that it kind of creates this surreal new genre of pure attentional drug, like the slot machine that people just watch and that doesn't have any real meaning because it's just this complete, neural-network-produced hive-mind creation; or people just opt for humans and it doesn't really change that much. Anyway, those seem like the 3 big options.

There's a paper from Yahoo Research from around 2001 or 2002 in which they use the multi-armed bandit problem, which precedes advertising, obviously, but then they adapt it because if you look at a page, there are several bandits there, with different ads. If you click on one, there's a chance of it leading to a conversion or not, and a lot of the math behind this one—I forgot—is that—

As a side footnote, look at the Apple Vision Pro, for example. It's way too bulky, way too expensive, and it's not quite good enough yet.

No, and I think the thing about the Apple Vision Pro and VR, to me, is—and I write at some length in The Sirens' Call about this—that at a certain point, if you're thinking about this commodity, you come up against these hard limits. At a certain point, it's like, okay, we have people looking at their phones as much as they possibly can. If we keep them awake a little longer, we can get more out of it. We start going to children. We sold the phone to everyone in the world. How do you keep expanding the frontier of the commodity to mine?

That's why VR is so important, because if you have it on literally every second, you've just unlocked this enormous new landscape. You've discovered these new deep-sea reserves that no one had before, and now, literally every second, if it's on every waking second, you can mine that attention and extract and commodify it in a way that wasn't possible before. Apple did this once before with the smartphone as the crucial piece of hardware to massively expand that supply. That's, I think, the commercial logic of the device, whatever it does after that.

Well, it's only going to be waking hours, Chris—16 hours. We're going to need the full Neuralink to actually run ads inside your dreams. Actually, we're going to need the full 24.

That's the point, right? It's like the frontier—there's a certain rapaciousness here.

No. Well, of course there is, and I can see that the ad-tech industry is completely rapacious. Toward the end of Chaos Monkeys, I think in the afterword that I wrote in the second edition or whatever it was, I mentioned that Facebook is running out of people, right?

Exactly. I made the joke that they'd start breeding new humans just to become Facebook users to sell them ads or something, which obviously is ridiculous, but I think it's a similar concept. There's actually only so many attention hours in the day, right?

And you do come up against boundaries. I'm definitely not a degrowth leftist at all. I would call myself a pro-growth left liberal.

Abundance, I think, is what we call them these days. Abundance, Chris. Abundance.

Yes, distinct from that in some ways. But I guess my point would be that there are these fundamental tensions between the need for endless growth, the boundaries you hit up against, and what it does to people.

You could create a company that just provided a useful service and printed money for a long time. In some ways, Google had that for a long time. They had a genuinely useful service. It genuinely transformed how we got information. It was insanely profitable. It was a money-printing machine, and it wasn't enough, right? You have to grow it past that.

In some ways, with the use of AI, the entire open web that was the basis for the first iteration of Google is collapsing underneath it. What comes after, when you talk about this sort of interface we have with the AI chatbot? I find that, for my work, the only way that I can use AI is with web search on and telling it to cite everything, because I can't risk a hallucination.

I need to check everything against: Did a person report this? Did a human being go and call this person? Is that name right? I do wonder about the sort of hyper-surreality that we're all entering into once the growth-revenue model collapses the open web. What's left? Who's going to maintain the stuff that's actually being used to create all that useful information if the whole thing collapses?

Yeah. I think I cite Edward Abbey and Chaos Monkeys' epigraph in there somewhere: “Growth for the sake of growth is the ideology of the cancer cell.”

Right, exactly. Yes, exactly.

Kind of true. It's funny—I cite Abundance, which is Ezra and Derek's book, and it basically tells the left, “Okay, don't be dumb about capitalism and growth. It could actually be a good thing,” right? I think there's another book by Alex Karp of Palantir called The Technological Republic that's basically telling the right, “Look, the market isn't actually the solution to everything, and it doesn't necessarily tell us what is good—the summum bonum, what the ultimate good ends of social life are. The market isn't necessarily there to answer that question.”

Peter Thiel famously asked the question, “We were promised flying cars and we got 140 characters.” I'll say this on that last point: It's interesting to me to compare the amount of capex and media attention to AI versus the insane solar-power revolution happening right now. It's a little like the flying-car, 140-character thing.

We've been burning carbon for energy before the Industrial Revolution. We've been burning trees for wood stoves, and then we got the Industrial Revolution and started burning fossil fuels specifically. But a world where we get to a point—which is totally possible now, I think, from an engineering standpoint—where the marginal cost of energy basically drops to zero, where we get the solar technology good enough that we kind of reclaim the abundance, to use a word, of the sun—that is such a wildly revolutionary thing to happen.

It doesn't get the attention, partly because it's not that lucrative, I think, in the end, but also because it's not as intentionally salient in the way that AI is. You have this weird thing happening, which is that all the capex and all the attention are on the future technology. The epoch-defining technology is going to be AI, and I don't know if I agree with that or not. I understand why some people think that's true, and I understand some of the skepticism.

But it's interesting to contrast that with the fact that right now we're in the midst of an epochal transformation in how we get energy, and it's just not that sexy.

You didn't mention nuclear, Chris. Nuclear might have to be part of the portfolio. But I agree with you that having marginally free intelligence and marginally free energy is going to be nuts.

Yeah. That's right. Those 2 things combined might be pretty close, and obviously they're related, because the energy consumption of compute is enormous.

Yeah.

And you're going to need the buildout.

But if you got to that point, it's like, Jesus, what does that even look like?

Chris, I know you've got to run. Thank you so much for coming on the podcast. The book is The Sirens' Call: How Attention Became the World's Most Endangered Resource. Thanks so much for coming on the show.

I really enjoyed that. Thanks so much.