00:00:29.340Today, Jazza and I are recording on the lands of the Wurundjeri Woiwurrung people.
00:00:34.240We would like to pay our respects to Elders past and present and extend this to any Aboriginal or Torres Strait Islander listeners today.
00:00:45.500Corey Doctorow is a Canadian-British author, activist, journalist and blogger.
00:00:50.800his work in the tech culture space hit new heights after the publication of his 2025 book
00:00:57.080and shittification why everything suddenly got worse and what to do about it
00:01:01.480cory has written dozens of books spanning science fiction non-fiction policy young adult graphic
00:01:08.340novels and short stories cory will be at the sydney's festival of dangerous ideas from the
00:01:13.86022nd to the 23rd of august and melbourne's now or never festival presented by the wheeler center
00:01:20.220on the 25th of August. His work in the privacy, media, copyright, AI, surveillance and tech space
00:01:27.880is vital. Thank you so much for joining us today. Oh, it's my pleasure. Thank you for having me on.
00:01:33.480Of course. So as some background, we actually started this podcast a good six years ago.
00:01:38.620We were strictly a pop culture and celebrity podcast, right? But now we find that we're
00:01:44.160speaking about big tech and AI almost every week. Have you seen these worlds merging more and more?
00:01:50.220Well, sure. I mean, you know, if you're talking about celebrities, it's hard to be a celebrity without the Internet in the same way that it's hard to do anything without the Internet.
00:01:59.120I often say this to colleagues of mine because I came up through, you know, progressive politics that say, well, why should we even try and fix the Internet?
00:02:10.340And, you know, as someone who spent his teenage years cycling around the streets of Toronto in the middle of the night in sub-zero weather, wheat pasting posters to the sides of telephone poles to get people out for demonstrations, you know, if you think you're going to run a mass movement without the internet, I want you to spend a couple of sub-zero nights out with a bucket of wheat paste.
00:02:29.880And in the same way, if you think people are going to like care about anything you do and you're not using the Internet, I'd like to see your plan for it because it's hard to imagine how it is that you are going to become someone of public regard and importance unless your conduct in some way intersects with the Internet.
00:02:49.860We also cover books on our podcast quite frequently. As a sci-fi writer, your job is to
00:02:56.120invent these futuristic stories based on the current world that we live in. We know that
00:03:02.780you can't predict the future, but as a Gen X, we're wondering if you could ever imagine the
00:03:08.220future looking the way that it looks now in the mid-2020s. Yeah, you know, I think that
00:03:13.620A science fiction writer's job is far less to imagine a new gadget than to think about who the gadget does things for and who the gadget does things to.
00:03:22.720I think science fiction is maybe the antidote to Margaret Thatcher's maxim that there is no alternative.
00:03:27.780You know, our riposte to that is there are six alternatives we can think of every day before breakfast.
00:03:33.020And there's nothing inevitable about the idea that if you have friends, Mark Zuckerberg needs to be privy to all of your conversations with them.
00:03:39.180And so I certainly, you know, if you go back to the 80s, 90s and aughts, you see two currents in the way that people imagined technology in literature and film and in art.
00:03:55.940And they are a mix of incredible hopes for how things could be, you know, the possibilities that technology might extend to us.
00:04:06.960And at the same time, absolute stark terror about what technology might do to us in terms of control and surveillance. And especially, you know, when you look back at the cyberpunks of the 1980s, you know, one of the main characteristics of those writers is that they're writing about a world in which a small number of firms have metastasized and formed cartels and taken over the world and become multinationals.
00:04:33.760Now, for the most part, they imagine those companies are going to be Japanese, which, you know, in retrospect, seems quite quaint. But at the same time, you know, there is an absolute imagining in the dawn of the Reagan years, in the dawn, you know, at the heyday of the Thatcher era, when Margaret Thatcher and Augusto Pinochet in Chile and Brian Mulroney in Canada and Helmut Kohl in Germany and so on, when they're all sort of taking over the world with neoliberal politics,
00:05:03.160that the end state of that really is like a digital world which as um tom eastman from new
00:05:10.360zealand says we have five giant websites filled with screenshots of text from the other four
00:05:14.660and and you know that that i think was absolutely absolutely not just foreseeable but foreseen
00:05:21.700um and while you know most science fiction writers are not very up on economics in my experience
00:05:28.980They really do, I think, get the vibe right when you look back on that era.
00:05:35.120And I think a lot of science fiction writers from that era are today at great pains to
00:05:39.900remind us that cyberpunk was a warning and not a suggestion.
00:05:44.300Yeah, watching The Terminator or any 80s movies nowadays, it's like, this is actually a bit
00:05:53.660What draws you to, I think, looking at these themes of surveillance, tech, etc., through the lens of fiction when you dabble in it so, you know, constantly in the nonfiction world?
00:06:06.960Yeah, so I've been a fiction writer for most of my life. I started selling stories as a teenager. And about, well, when I was in my early 30s, I went to work for a nonprofit called the Electronic Frontier Foundation, where I've worked now for 25 years.
00:06:26.980So most of my adult life, I'm turning 55 in a few days. And EFF's mission is to, you know, uphold digital rights in the, you know, on the virtual world and in virtual policy.
00:06:42.580And so, you know, I've done treaty work for them and standards and lobbying and just activist rabble-rousing and street demonstrations and everything in between for the last quarter century.
00:06:54.940And it was a very natural extension of the science fiction I was writing to, you know, like the reason the themes of the science fiction matter to me was the reason I became an activist.
00:07:07.460And so, you know, I was always going to end up doing both. And there's lots of times where you can take this very abstract technical debate that people are mostly not engaging with, because for very good reasons, people mostly want to engage with things that are concrete and immediate, because those are the things that are, you know, matters of life and death, as opposed to these things that won't be matters of life and death for a couple of decades.
00:07:37.460And by turning those into stories, you can elevate the salience of them and maybe avert them becoming a matter of life and death.
00:07:46.340You know, it would have been very nice if we had felt as vividly about the climate emergency now, if we felt that way in the 1980s.
00:07:56.880And so we didn't have to wait for wildfires to be on our doorsteps before we decided it was time to start doing something about it.
00:08:04.180and a lot of our audience will know you for the terms that you coin one of your most famous is
00:08:10.040of course and shitification for people who aren't familiar with this term can you please explain it
00:08:15.240for them sure well you know like i said i've spent most of my adult life trying to get people to care
00:08:20.080about these issues and i've come up with any number of you know narratives and framing devices and
00:08:24.760similes and parables but it turned out that uh dirty little word and the accompanying minor
00:08:29.220license to vulgarity at moments when you're frustrated with the networks that you rely on
00:08:33.340but which are getting progressively worse and which you can't seem to escape, that that was
00:08:36.920the key. And so I coined this term in an essay I wrote, and it has since taken on a bit of a life
00:08:43.380of its own. It's in the Macquarie Dictionary. It's just been inducted into the OED. It's in Webster's.
00:08:48.760It's been several different kinds of words of the year, and it's been an international bestseller,
00:08:53.860the book of the same title. And in Shittification is a theory about why platforms are going bad
00:09:01.940and a description about the normal pathology of platform decay,
00:09:05.960the way that platforms typically go bad.
00:09:08.680And in that pathology, we see a three-stage process of decay.
00:09:13.320In stage one, platforms are good to their end users,
00:09:17.700And in stage two, once they know the users can't readily escape,
00:09:20.840they make things worse for those users in order to extend benefits to businesses.
00:09:25.820And then once the businesses are piled in and locked in,
00:09:29.060they make things worse for them as well.
00:09:30.660This is an area of departure from, I think, common places that are not really true.
00:09:36.740Like if you're not paying for the product, you're the product, which seems to imply that if you just gave them money, they treat you better, which is not true.
00:09:43.540You know, they will treat you as badly as they can get away with.
00:09:46.740And so we see in stage two, the advertisers, publishers, other business customers that supply the income to these platforms, they get locked in.
00:09:56.180And so in stage three, they get extracted from as well.
00:09:58.920And, you know, whether that's Amazon charging 60% junk fees to all of its platform sellers or Uber lowering the wages of drivers monotonically or advertisers on Facebook seeing their prices going through the roof, even as the quality of ad targeting goes down and the amount of ad fraud explodes, that the business customers also have all the value drained off from them.
00:10:20.980And so in stage three, you have this end state where the value left in the platform is a kind of mingy homeopathic residue that's calculated to be just enough to keep users locked to the platform and then keep businesses locked to the users.
00:10:33.540And all the value is harvested for the executives and the shareholders.
00:10:39.420Incredible. Just what everybody wants, right?
00:10:41.820Can you use this and paint a picture of what the, I guess, the kind of Google searching looks like today?
00:10:49.800Yeah, well, I mean, we've all encountered this, right?
00:10:52.380You type a search query into Google that just a few years ago would have given you a nice
00:10:56.740crisp answer to whatever you were trying to find out, and instead what you have is like
00:11:00.980a paragraph of AI slop, and then you have five ads, and then you have five scam pages
00:11:07.000that have escaped and that are themselves just garbage ads, and when you click on them
00:11:11.140and you scroll through and you get a hundred interrupters and a thing that demands that
00:11:15.120you subscribe to some email and some auto playing video and then a chum box with someone's weird
00:11:20.720tooth implants uh and then if you're lucky maybe you find out you know how many um ounces there
00:11:27.200are in a in a in a cup uh which is the thing you're trying to find out from the beginning
00:11:31.900and you know this is not an accident uh google uh you know when they uh over the last 25 years
00:11:39.140is a company that has grown to acquire a 90% search market share.
00:11:46.740But the way they got that market share wasn't by being better than everyone else.
00:11:50.240The way they got that was by buying all of their competitors.
00:11:53.340So Google wants you to think they're Willy Wonka's idea factory
00:11:56.660and that they're just coming up with product after product.
00:12:00.100And when you actually look at the history of the firm,
00:12:02.200almost everything they've done internally has been a failure
00:12:04.660and all of their successes, so their video stack,
00:12:07.520They're ad tech stack, server management, documents, collaboration, maps, directions, you know, mobile, everything that they do except that search engine they launched a millennium ago is something they bought from someone else.
00:12:23.260They're just rich uncle penny bags and they're buying other people's toys, right?
00:12:29.760The competition laws of the world broadly prohibit that kind of anti-competitive merger, but we've had a 45-year period of tolerance for monopolistic conduct. And so we let them buy their way to dominance. And along the way, they bought all the shelf space.
00:12:44.900So no matter what carrier you use, no matter what hardware vendor you use, what operating system you use, no matter whether you're an Android or iOS, no matter what browser you're using, every search box you find is wired directly into Google's servers.
00:12:58.660And they're bribing Apple more than $20 billion a year not to enter the search market.
00:13:03.240And so this is how they end up with this 90% search market share.
00:13:06.000Now, here's the great paradox of a monopoly is that when you hit 90 percent, you stop growing and you stop growing because, you know, there is no one else to search.
00:13:17.940You know, we you could raise a billion humans to maturity in the hopes that they become your customers.
00:13:24.120And Google has a product called Google Classroom.
00:13:26.540But the problem is that that's going to take 15 years to show results.
00:13:31.540So Google lost three antitrust cases in 2024 and 2025 in the United States. Antitrust is what the Americans call competition law. And in one of those cases, what's called the search case, the U.S. Department of Justice released a bunch of internal emails that they compelled Google to disgorge.
00:13:52.900And in those memos, we see an incredible struggle that took place in 2019 and 2020 as Google's search market share was truly at its peak, as high as it could go.
00:14:03.800And a great tech critic named Ed Zitron reported these out.
00:14:07.280And what they show is that around this time, as the Google search revenue growth tapered off, the guy who was in charge of Google search revenue, an ex-McKinsey guy called Prabhagar Raghavan, had this idea that if you made search worse, you could increase search revenue because you'd have to search two or three times to get the answer that you used to be able to get in one shot.
00:14:31.320and you know there's lots of easy ways google has to make search worse right they can turn off the
00:14:36.080spell checking or what's called query stemming where like you search for trousers and they run
00:14:40.560a parallel search for pants and then they merge the results or you know they do context awareness
00:14:47.220so you know if there's if there's a big scandal going on somewhere in the world if someone's just
00:14:51.860thrown a sandwich at someone else and you search for the right kind of sandwich then they show you
00:14:57.020those results at the top you just turn that off and now you're gonna have to search two or three
00:15:01.040times to get your answer. And, you know, the person who is most horrified by this is another
00:15:07.660Google executive, a guy called Ben Gomes, who's on the technical side of search. And he's like an
00:15:12.560OG Google tech guy. He built out their server infrastructure from a few computers under a desk
00:15:18.000to a global system of data centers. And he's just palpably horrified. You know, he hasn't missed his
00:15:23.960mother's funeral and his kids' little league games to, you know, in-shittify this search engine that
00:15:29.140he is so proud of and that he and everyone he knows uses all day long. And he loses the argument
00:15:34.860and he loses the argument because why wouldn't you do this, right? It's not like people are
00:15:40.480going to find a search engine somewhere else. It doesn't matter where you're searching. It's
00:15:44.100always going to be a Google search box. You know, he's, he's just completely routed in this.
00:15:50.880And the thing is Prabhakar Raghavan is definitely not the first person who's ever worked for Google
00:15:55.920who said, you know, if you had to search more than once, we'd get more money because we'd show
00:15:59.800you ads more than once. But he's the first person who's won that argument. And the reason he's won
00:16:04.420that argument is that the policymakers who, unlike Google executives, actually work for us,
00:16:09.820you know, our competition regulators, our parliamentarians, our enforcers, they created
00:16:15.940an enxietogenic policy environment where the worst ideas of the worst people made the most money.
00:16:20.940and once you do that then you get this outcome you know the reason google and shitifies is the0.98
00:16:27.200same reason your dog licks its balls because no one stops them and because they can and i think0.96
00:16:33.160that's yeah why there's the why we keep going back to them like we are all complaining about0.99
00:16:39.120oh how crap google is but it's still the first search engine you open up i'm literally using
00:16:43.640google chrome right now we are really tied to these platforms and especially instagram and0.98
00:16:49.680TikTok for us as well. Your latest book is called The Reverse Centaur's Guide to Life After AI.
00:16:57.460Can you please tell us what a reverse centaur is? Sure. Well, you know, this isn't the first time
00:17:03.540we've had a big conflict between labor and automation. And so there's a pretty good
00:17:07.320literature about how labor and automation play out. And one of its truisms is that when workers
00:17:13.740drive automation, they typically do so in order to improve quality. Like you might adopt a tool
00:17:18.440like Riverside, the one we're using now, because if you use Zoom, then, you know, you're at the
00:17:24.520mercy of network latency and you're not getting recordings at both ends. And so you end up with
00:17:28.620all kinds of substandard audio at the other side. But when bosses drive automation, they do so
00:17:35.940because they want to improve throughput, right? If you were working for someone else and they said,
00:17:41.000hey, we're using Riverside, it wouldn't just be about improving audio quality. Be like, now you're
00:17:45.100not going to be doing all those retakes and all that editing we'd like you to do twice as many
00:17:48.920podcasts right and so uh this is uh not because bosses are necessarily just mustache twirling
00:17:56.540villains it's because when your relationship to a tool is that it's an asset that's depreciating
00:18:01.900or an expense item on your budget you want to maximize the return from it right and so you
00:18:07.400want to use it as much as possible before it wears out or you know while you're you're paying for
00:18:13.200that subscription every month and this is where centaurs and reverse centaurs come in in automation
00:18:19.400theory a centaur is someone who is assisted by a machine you know in greek mythology you have the
00:18:25.300human head on the horse's body the horse is tireless it's dexterous it's fast uh but and
00:18:32.700it's strong but it it's not in possession of judgment the judgment is coming from the human
00:18:38.620head that's directing its operations and so when you are married to a machine you're the head on
00:18:43.900top of the machine's body and it might be stronger than you and be able to work longer than you and
00:18:48.980be able to lift things that you can't lift but you're the one telling it where to go and what
00:18:53.900to do you know a human on a bicycle is a centaur you even look like a centaur when you're riding
00:18:58.440a bicycle but also a human you know using a spell checker is a centaur and so that centaurship is
00:19:05.400quite nice and it really corresponds to labor-driven automation adoption a reverse centaur is the
00:19:11.040reverse right it's it's the horse's head on the human body it's the machine directing the
00:19:15.340operations of the human and of course the machine is faster than you and can work longer than you
00:19:20.640and can lift things that are heavier than you can lift and so it wants you to work at the absolute
00:19:25.240limit of your capability because it has capabilities that exceed yours you are the weakest link in the
00:19:31.220chain. You're the bottleneck in the system. And so the machine is pushing you and pushing you
00:19:36.800and pushing you. And so, you know, Amazon warehouses are the most automated warehouses
00:19:42.320the world has ever seen. And Amazon warehouses have the highest rate of injury of any warehouse
00:19:47.680in America. And that's not a coincidence. It's a consequence. You're not going to spend eight
00:19:52.160figures on warehouse automation only to leave the plant sitting idle. You're going to work the
00:19:57.180people who are the peripherals for the warehouse machines at the limit of their capabilities. And
00:20:02.880when you work at your limit, you eventually make a mistake. And when you make a mistake in a
00:20:07.740warehouse, you sometimes end up impaled on a forklift. Yeah, wow. And I think adding to that
00:20:13.160whole issue is how extra surveilled these delivery drivers are, you know, can't take
00:20:19.500pee breaks, can't listen to music while they're driving to meet their KPIs. It's very dystopian
00:20:25.680stuff. We also want to talk more broadly about AI. A lot of our audience, including ourselves,
00:20:32.420are quite anti-AI, but we're at this weird standpoint where I feel very conflicted. Do we
00:20:39.080risk being left behind, especially as women in this space? Is there a line to draw?0.54
00:20:44.700So I think being anti-AI per se is weird. I think that opposing statistical inference is kind of a
00:20:51.840dead end right but if you're in charge of how you use ai you might for example uh use whisper
00:20:59.560to produce transcripts of your podcast so that when you're doing edits you can pull up a clip
00:21:04.500you know if you want to find the clip later and insert it in that's fine right it's like that's
00:21:10.780just that's just a plug-in for your audio editing software you know i my first word processor i'm
00:21:16.420old enough that, you know, I had a Apple two plus when I was like nine years old in 1980.
00:21:21.540And my first word processor, uh, I got by going to the news agent and buying a magazine with a
00:21:27.600word processor program printed in it and typing it into my computer. And yeah, that's how we used
00:21:32.860to get software. Um, and in the years since, um, I have had many new plugins added to my word
00:21:40.480processor and some of them are good and some of them are bad. And some of them other people use
00:21:44.640in ways that I think is stupid and makes their writing worse. But I've never had a word processor0.98
00:21:49.060plugin to which my response was, hey, this is so good. Let's bet the world economy on it. And while
00:21:53.940we're at it, we should put all the writers in a wood chipper. And so if you think of AI as just
00:22:00.220a normal technology, as a tool, and if it's being driven by you as a worker, you might find it
00:22:05.380useful. You might find, for example, that you need to pick up a line and one of you is away
00:22:14.620from the mic and so you deep fake a line so that you can pick it up because you made a mistake and
00:22:20.900you know you've realized you've libeled someone and that living in Australia where you have the0.90
00:22:25.060world's second stupidest libel laws after the UK that you might end up losing everything you have0.99
00:22:30.300and so you you say no Rupert Murdoch isn't an immortal uh you know uh blood-sucking vampire0.99
00:22:35.660he's an alleged immortal blood-sucking vampire and and thereby escape penury and you just you
00:22:41.220know you know jasmine deepfakes maggie's voice saying that who cares who cares right like that's
00:22:47.800fine right that's you driving adoption but the idea that you have to like i don't know uh deepfake
00:22:53.860your voices and then use ai to write your scripts and then blah blah blah blah i mean ai has nothing
00:22:59.260to say right the reason ai art is so washed is that when we first saw it we'd never really
00:23:06.040encountered like paintings that didn't have painters and you know books that didn't have
00:23:10.200writers and so we imputed or you might say hallucinated writers and painters that were
00:23:16.340directing the painting as opposed to just statistical inference that was producing the
00:23:20.380painting or producing the words and so you know we put the meaning in it but it has no meaning
00:23:26.860and it can't have meaning uh you know when if you're an atheist like me you might think that
00:23:33.920a sunset is interesting or beautiful but you don't think it's trying to tell you anything
00:23:37.620but someone who takes a picture of a sunset and you know texts it to you or someone who paints a
00:23:43.640painting of the sunset part of the aesthetic experience of looking at that image is trying
00:23:48.360to understand what the person who made that image was feeling and wanted you to feel and so we can
00:23:55.280get that from AI so long as we don't realize that AI has nothing to say but as soon as we realize it
00:24:01.260has nothing to say it just all turns into something that's very washed it might be technically
00:24:06.060interesting or aesthetically striking but it's not beautiful and it's not important right it can't be
00:24:13.560you know if there was like a a spot somewhere in the world you know my kid is going to a university
00:24:19.400in california called santa cruz and uh there's this tourist trap nearby that's very famous called
00:24:25.380the mystery spot where water seems to flow uphill it's just an optical illusion right but um if there
00:24:31.320was like a mystery spot where you went there and you threw a pile of leaves in the air and they
00:24:35.000fell down and made a sentence we would flock to that mystery spot for quite some time and look
00:24:40.320and marvel at the sentences but eventually we'd realize they're not very good sentences and that
00:24:45.500the important thing about them is not their literary merit but the strange technical underpinnings of
00:24:50.460this phenomenon and you know like i think that ai is technically interesting i just don't think that
00:24:57.620it's it's art i don't think it can do your job i think you know most people's bosses are you know
00:25:03.700credulous weirdos who are infinitely horny to replace workers with automation. And so that
00:25:09.160the fact that the AI can't do your job doesn't mean that it won't try that your boss won't fire
00:25:14.680you and replace you with AI that can't do your job, because they're an easy mark for an AI
00:25:19.040salesman who's like, you know, let's fire Maggie and split her wages between me and my boss, Sam
00:25:22.940Altman. Right. But but, you know, that that that doesn't mean the AI can do your job. And we should0.91
00:25:29.540be, you know, kind of careful materialists about this. We shouldn't like help the AI salesman raise
00:25:35.520money by going out there and stipulating to facts, not in evidence, like the idea that the AI can do
00:25:43.260your job. Totally. And I think that's what we, how we feel like being anti-AI is like anti-AI
00:25:50.780of what it's doing to the culture and like the AI slop it's producing and the way people are
00:25:56.220treating it like it's this new god when we work in media here in australia and we've seen people
00:26:02.540so many of our journalists and writer friends be made redundant so it feels very real being told
00:26:09.680you know you should use ai to output more and more and more like we are becoming the
00:26:13.720robots essentially so i think that's why we are so critical about it but you wrote a great article
00:26:20.020for the guardian titled ai companies will fail we can salvage something from the wreckage and in
00:26:26.120you pretty definitively say that AI is a bubble. Why do you think AI is a bubble?
00:26:31.360Well, there's two parts to that. So the first is, why do we have tech bubbles? And the second one
00:26:36.540is, why is AI one of them? So tech bubbles are part of that paradox of market saturation, that
00:26:43.320while you're growing towards monopoly status, the capital markets love you, right? A share in a firm,
00:26:51.380I think people don't really understand this. The fundamentals of a share in a firm is that
00:26:56.780it's a claim on the future income in that firm. And a growing firm has more future income than
00:27:01.740a mature firm because tomorrow it's going to make more than it made today, right? So if you have
00:27:06.640two firms, one of which made a million dollars this year in profit, the other one made a million
00:27:10.720dollars in profit, but the second one is going to make $2 million next year, then a share in that
00:27:16.500firm is worth more, right? Now, the corollary of this is that the instant you become mature,
00:27:21.820say, because you've got a 90% search market share, your firm is grossly overvalued.
00:27:27.780And what we see are these flash sell-offs in tech firms when their growth slows down. And,
00:27:33.180you know, I think leftists kind of have done the world a disservice by repeating this truism.
00:27:38.400Endless growth is the ideology of a tumor, right? As though the reason these companies want to keep
00:27:44.440growing or the people who run them want their firms to keep growing is purely ideological it's
00:27:48.100not it's material right if your net worth is in shares in the company you work for or founded
00:27:53.560and a failure to continue to grow results in an 80% reduction in the value of those shares
00:27:59.300the reason you want to keep the growth going is not ideological it's because you wouldn't you
00:28:04.440don't want to be 80% poorer and who amongst us wants to be 80% poorer and so there needs to be
00:28:11.300a story about how you're going to keep growing, either when you saturated your market. And, you
00:28:15.180know, you can ground that story in real things. You can say, oh, we're doing the pivot to video,
00:28:20.180we're Facebook now, but tomorrow we're going to be Facebook plus YouTube, and we're going to
00:28:24.100absorb all of YouTube's revenues. Or you can say, oh, we're launching Google Plus, so we're Google
00:28:29.020today, but tomorrow we're going to be Google plus Facebook, and we're going to absorb all of
00:28:32.080Facebook's revenue. The problem with that is that, like, when you claim you're going to become
00:28:35.660Facebook, Facebook has other ideas, and we'll go out and tell people why that's not true.
00:28:39.800and so you can claim instead that you're going to conquer a market that you've just pulled out of0.89
00:28:44.220your ass right we're going to conquer the metaverse we're going to conquer uh cryptocurrency or daos0.63
00:28:49.920or nfts or you know web 3 and while those things don't exist you know the advantage of a thing that0.97
00:28:56.460doesn't exist is that no one knows how much it's worth and since you just pulled that out of your0.94
00:29:00.620ass you're the world's leading authority on how much it's worth and so you can claim it's worth0.89
00:29:04.640empty trillion dollars and provided you have a new story by the time that old story is proved not to
00:29:09.760be real, you know, you're like running across the river on the backs of alligators without losing a
00:29:14.620leg, right? Then you can just keep the growth story going. And so long as the growth story is
00:29:19.900going, you can grow by other means, right? Like if your shares are highly liquid because you have
00:29:25.460a high price to earnings ratio because you're growing, then you can buy another company with
00:29:30.860your shares. Buying other companies with shares is great because shares are an endogenous substance
00:29:36.600produced within the four walls of the firms you make them by typing zeros into a spreadsheet
00:29:40.480right boring old money is something you have to get from like a bank or a customer or an investor
00:29:46.720and if you try to make the money within the four walls of the firm someone from the treasury shows
00:29:51.260up and takes you away in handcuffs right and so you know that growth stock is like it's worth
00:29:56.640a lot because it can also beget more growth it can become a kind of self-looking ice cream cone
00:30:02.460But when you stop growing, well, the reverse happens. You enter a nosedive. All the employees who are compensated in shares leave. You can't use them to restart the business. All the other companies you buy to restart your growth want money now and you don't have any, right? So it's a self-fulfilling prophecy.
00:30:29.500And some of that size can be accounted for by the fact that there is a real computer science thing happening at the middle of AI in a way that exceeds the computer science value of, say, the incremental improvements in VR that were supposed to give us the metaverse or the somewhat interesting trustless ledger maintenance.
00:30:50.840that was supposed to give us blockchain futures or whatever.
00:30:55.520You know, AI is real in the sense that like 10 years ago,
00:30:58.440some researchers took existing machine learning techniques,
00:31:17.440and you get more and more major variants.
00:31:20.000Now, we've reached the end of that, like they're getting severely diminishing returns to scale. But like, that is a real economic thing. When you get positive returns to scale, that is a real investment opportunity. But it's not this investment opportunity. When I wrote the book, AI total capital expenditure had reached $700 billion US. That was just over a year ago.
00:31:40.780When the book hit shelves, the number being bandied about was $1.4 trillion, so that's doubling.
00:31:48.440But Gartner now says by the end of this year, it's going to be $2.5 trillion.
00:31:52.280So $2.5 trillion in CapEx and OpEx, just the total burn rate of all these companies.
00:31:59.880Now, how much are they making? $50 billion a year, right?
00:32:03.500And $50 billion sounds like a lot until we're talking about $2.5 trillion.
00:32:08.200And although the assets themselves, they're not durable, right?
00:32:14.960So they're having to turn over the data centers and the GPUs like every couple of years.
00:32:21.780So it's not just that they have to replace the GPUs.
00:32:24.260They actually generally have to replace the data centers too.
00:32:26.860Because the way that NVIDIA is actually getting more performance out of each generation of
00:32:31.540GPUs, which they need to do if they're going to convince people to buy new GPUs every year,
00:32:35.360is they're throwing away the design constraint
00:33:04.160The seven companies that are 35% of the S&P 500, six of which are losing money to the seventh, which is Nvidia. So when they run out of money, so does Nvidia. They're passing around the same $100 billion IOU really fast and pretending that it's in everyone's bank account at once, which, you know, my other life, I write science fiction novels, including a bestselling series about a forensic accountant who unwinds tech scams.
00:33:28.160And to call that an accounting trick is to do violence to the noble accounting trick.0.99
00:33:56.740should we stop talking in this way? Is this how we start to kind of turn the conversation around?
00:34:02.860Is that possible? Yeah, yeah. I think that's the trick here. I mean, look, when the bubble goes
00:34:07.620away, we're still going to have lots of hardware. We're going to have lots of skilled technicians
00:34:11.480who don't have to answer to their stupid bosses anymore because they're out of a job. And we're1.00
00:34:15.620going to have open source models that will have lots of room for optimization and improvement.
00:34:19.960And so whatever use you get out of AI is going to be around for you. You're going to be able to do
00:34:24.800What we're not going to have are these companies that are selling $100 bills for a dollar a piece, losing thousands of dollars on every customer. And, you know, this inevitabilism, I think, it's a very cheap rhetorical trick, as are kind of the invidious comparisons that are made between AI and other technologies that ultimately panned out after losing money, like, you know, Amazon or the web.
00:34:51.420You know, the web did lose money when it started, but the web had very good unit economics, which is a finance term for how firms perform at the margin, right? When they add a new customer, when the customer makes a new use, when you get a new generation of the technology, if you have good unit economics, then all of those are positive factors. And that was true of the web.
00:35:12.380Every web user made the web more profitable.
00:35:15.320Every time that web user returned to the web, the web got more profitable again.
00:35:19.040Every generation of the web was more profitable.
00:35:22.780AI loses money on every customer and loses more money on every customer's every use.
00:35:27.880And every generation of AI loses more money than the previous generation.
00:35:31.800And so the idea that it's inevitable is absurd.
00:35:33.940you know that there may be some kind of technology down the down the pipe that is going to make some
00:35:41.300money with this but it's not going to look anything like these companies and these companies
00:35:45.360I don't think they have a path to survival after having committed all of this capital
00:35:51.360you know we are really running up against sums that are proper business threatening sums even
00:35:58.980for very large very profitable businesses you know Facebook's a good example here
00:36:02.780Meta pissed away $61 billion on the metaverse. And that sounds like a lot. But over the last
00:36:08.840three years, they've spent $150 billion. They spent another $150 billion this year on AI.
00:36:14.320And they're talking about more next year. And they've had to do thousands of redundancies
00:36:18.480because they've run out of money. And now they're saying, well, we can't keep the lights on because
00:36:24.260we don't have the programmers anymore because we thought the AI could do their jobs. And it turns
00:36:28.420out it can't. But we don't have the money to hire back the programmers because we spent it all on
00:36:32.460ai 100 do you think like where do you think this ends or how do you and this is such a big question
00:36:39.540i'm just so interested by your answer but how do you think this bubble will burst or when yeah
00:36:45.880could you predict when is it going to be like the defining yeah so i just i think the 2030s would be
00:36:52.100would be uh a very good long innings i'd be very surprised if we were still talking talking about
00:36:57.300us in the 2030s. But, you know, I earlier mentioned Stein's law of finance that anything
00:37:03.040that can't go on forever eventually stops. There's an equally important law, or rather aphorism,
00:37:08.420John Kenneth Galbraith, who said the market can remain irrational longer than you can remain
00:37:12.120solvent. So I don't know when the bubble is going to burst. You know, I think whenever the bubbles
00:37:18.040do burst, we look back and we say, well, in hindsight, it was obvious that it was about to
00:37:23.280go. And certainly, you know, the AI finance bubble is there. It has a lot of very brittle
00:37:28.640contingencies, right? Like, so think about how much money, uh, on data centers is coming from
00:37:34.740Gulf sovereign wealth funds and individuals in the Gulf whose income is tied to the oil sector
00:37:39.760there. And who's like, uh, you know, pipelines and refineries and LNG terminals keep getting
00:37:46.280bomb by Iran. And maybe one of them pulls out of a key deal. And these deals are all structured so0.59
00:37:54.020that if one bit of them fails, then there's a cascade of failures through a bunch of other
00:37:58.760deals because they're cross-collateralized. And they have very, very tight tolerances. It's a
00:38:04.060machine with a lot of tight couplings and without a lot of slack in the system. And so anything
00:38:09.840could could knock it over uh and and that's why i think you see these panic sell-offs from time
00:38:15.900to time whenever there's like a little bobble i think that there's a difference between the
00:38:21.880investors who are in all in on ai because they think ai is about to become god uh and do all
00:38:29.400the jobs and make you know infinite money hack and then the investors who are like i'm just in
00:38:34.480because those weirdos will buy AI at any price, and so I'm just going to buy AI and sell it to
00:38:40.980them. And when those people bail, because those people are sleeping with one eye open and one
00:38:46.420fist poised over the sell button, and when those people bail, it's going to be a bloodbath for
00:38:53.400everyone, because they're the ones who are making up the real money in this, the bulk.
00:38:59.840Now, everyone is at this phase trying to unload on us, right? This is the point in the bubble where you try to find exit liquidity in the form of everyday suckers, because that's what bubbles are, right?
00:39:13.140Even productive bubbles, even railroad bubbles or the web bubble or whatever, they're very evil.
00:39:20.420The point of a bubble is you have people who are just everyday savers who are putting away money that they can barely afford to not spend today because they're worried that when they get old, they'll be homeless and that they'll starve.
00:39:33.940and you take those people who are desperate and scared and who no longer have defined benefit
00:39:40.040pensions and who have to have these market pension accounts where they know they're the sucker at the
00:39:45.240poker table they know that they don't understand the market as well as the real market players
00:39:49.920and that they're in danger of losing everything and you say here's a sure bet and you get them
00:39:54.480to pile in and they they take the dodgy stock off your hands and then they're left holding the bag
00:40:01.020when it all pops. And we're seeing a real effort to bring in those suckers now. Elon Musk's IPO is
00:40:10.300the kind of exhibit A here, where we've all been told that if you want to be financially literate
00:40:17.640and you want to save money for the future, you should just buy stock indexes, these index funds
00:40:23.220where you want a little of all the shares, top shares in an important stock exchange like the
00:40:28.400NASDAQ or the S&P or the FTSE here in London. And he got these exemptions that allowed his
00:40:34.760company to debut in the indexes because normally, even if a company is in the top 100 of the
00:40:41.200S&P or the NASDAQ, it doesn't get included until it's been listed for a few consecutive
00:40:49.160quarters of growth because it might just be that there's a lot of exuberance around the
00:40:53.660initial initial flotation. And so you don't want to stick a bunch of pensioners or would-be
00:40:59.500pensioners with this garbage stock. And he got an exemption, right? Allowing his shares to be
00:41:04.700listed immediately on the exchange's indexes, which means that every index fund manager is
00:41:12.220required to buy those shares, right? They can't say, well, I think this one's dodgy, right? If
00:41:17.900you run a NASDAQ 100 fund, you are legally obliged to have as many shares in that fund
00:41:27.220as are representative of their proportion in the top 100. So if SpaceX is 5% of the 100,
00:41:33.660then 5% of your shares in your fund have to be SpaceX at any price, at any price.
00:41:41.040And one of the things that Musk did was restrict the number of shares that they floated from the
00:41:45.020company so that insiders could sell their shares. And the insiders were also exempted from the
00:41:51.480normal lockout that prevents insiders from dumping their shares the day the flotation happens.
00:41:56.560And so we just were left holding the bag, right? Everyone who just taken the advice that you've
00:42:01.340been given since high school, right? That you should be putting away a certain share of your
00:42:05.940wages every week and that you should take that and buy index funds. We just were left holding
00:42:12.640the bag for the Emiratis, for the tech billionaires, for Musk himself, as they unloaded this dogshit0.96
00:42:20.660stock that today is trading significantly below the IPO price.0.64
00:42:25.940It took less than a week for all those gains to disappear and for the stock to go negative.
00:42:31.040It feels so hopeless, so much of the conversation.
00:42:36.360And as we head towards wrapping up our conversation, I think a big question that we have a lot
00:42:41.960at the back of our minds is how do we continue to move forward when big tech is so entrenched
00:42:46.920in so many of our lives? And especially we're having these conversations predominantly on
00:42:51.060social media, predominantly on platforms like Spotify. Is it hypocritical to critique while
00:42:56.640we still actively engage with them? Yeah. So I think that this 40, 50 years of neoliberalism
00:43:05.560has really broken our brains about how change happens. We've been told that change is the
00:43:10.740result of personal consumption choices that the way you get rid of a monopoly is to shop really
00:43:14.940carefully and that that's all a boycott is a boycott is you agonizing over whether you should
00:43:19.860have a twitter account or what to buy at the grocery shop uh that's not a boycott right you
00:43:25.200know bds or or the south african anti-apartheid boycott or the montgomery bus boycott where
00:43:32.420people didn't ride the bus for 380 consecutive days that wasn't careful shopping right that
00:43:37.300wasn't 380 days of people waking up and going, well, I guess I'm walking to work, right? It was
00:43:41.640a movement. It, it, it, I know that like going to meetings sucks and it's full of tedious people0.95
00:43:47.720who quote Robert's rules of orders and, and, and no one can agree on who's going to do the washing0.81
00:43:51.660up afterwards. But, you know, actual like change comes from organizing carpools when you're going
00:43:58.060to boycott the bus and having crisply articulated goals and going to the legislatures and bringing
00:44:05.140people into an ideological project you know what happened after they broke the montgomery bus
00:44:09.660segregation system is they had a civil rights movement made up of the people who came together
00:44:16.680as comrades by by breaking the bus system right that's the that's the advantage of it so you're
00:44:22.660not going to shop your way out of this stuff and like worrying about your consumption choices
00:44:26.800is a mugs game because the fact that you can't change things by shopping carefully
00:44:32.700is going to make you very demoralized and it's going to make you angry at your friends
00:44:37.600because you're going to go like, I have been agonizing in the grocery aisle for the last 10
00:44:42.600years and we still have, you know, cruel conditions in our hen farms. And it must be because you're
00:44:50.400not shopping carefully enough. And so now I'm angry at you because you're not a good shopper
00:44:54.880like me, right? So the way that you actually make changes, you join a polity, right? You do
00:45:01.360politics. And so local politics suddenly becoming very salient, especially with the data center
00:45:07.860fights. Let me tell you, your town fathers are not signing non-disclosure agreements and skipping
00:45:14.420the environmental review because they know you're going to be pleasantly surprised and they don't
00:45:17.880want to ruin it. The only reason to do that shit is if they know you're going to hate it.0.99
00:45:22.580Local politics in the age of rising fascism turns out to be the difference between whether you have1.00
00:45:27.500flock cameras are not, which also turns out to be the difference between whether your neighbors are
00:45:31.460going to be kidnapped and sent to a gulag. And, you know, if you think that can happen in Australia,
00:45:35.840I've got a little offshore island full of people who are seeking asylum, who have something else
00:45:40.940to tell you about the Australian attitude towards authoritarianism, not to mention that weird
00:45:50.040Queensland governor whose wife baked the famous pumpkin pies who banned all gatherings of more
00:45:54.820than five people, right? I mean, there is authoritarianism lurking in the political
00:46:00.960capacity of every polity. So you've got to get involved. Local politics is easier than you think
00:46:07.700and can make a big difference. You can also get involved with unionization. That's very important.
00:46:13.240If you want labor-driven automation, then you need labor power. Tech workers especially need
00:46:18.260to be thinking about this because like the reason you got the free kombucha and the massages and a
00:46:22.620surgeon who'd freeze your eggs so that you could work through your fertile years wasn't because
00:46:26.840your boss liked you it was because they were scared of you quitting and now that they've had
00:46:30.720mass layoffs they're just not scared of that anymore and we know how they treat the workers
00:46:34.480they're not scared of they're the ones pissing in bottles or working in factories in iphone city
00:46:39.300with suicide nets around them and then finally you get involved in the international struggle
00:46:44.100and i as i say work for the electronic frontier foundation eff.org we have intervened many times
00:46:50.160and in many ways in Australian tech policy fights,