← Podcast – Cory Doctorow's craphound.com
Podcast – Cory Doctorow's craphound.com
The Reverse Centaur’s Guide to Life After AI launch at Kepler’s Books with Angie Coiro
This week on my podcast, audio from Sunday’s launch in Menlo Park for The Reverse Centaur’s Guide to Life After AI at Kepler’s Books with Angie Coiro. Catch me next tonight in Toronto at Osler Records, tomorrow in NYC with Jonathan Coulton at The Strand, Thursday in Philly with David Williams and Friday in Chicago... more
Loading summary
Transcript105 lines
- [00:12]
A
Well, hello and welcome to the Cory Doctorow Podcast. Here is a little bit of cinema verite for you. I'm recording this on the Danforth in Toronto outside of Book City Books, where I've just finished the last of many bookstore signings for the day, about, I guess, 10 racing around town before continuing on with my book tour tomorrow. And I'm recording this just before going off to get some Greek food as one does on the Danforth. And it's by way of an introduction to today's podcast, which is a recording of yesterday's session. God, I slept on the red eye last night. Was it really yesterday? It was yesterday's session at Menlo park with Angie Coiro in Kepler's books. So as I mentioned on the last podcast, I will be sending you some of these updates from the road and well, I got a go get dinner, so I hope you enjoy this hour. We had an amazing time. It was a really good discussion and I hope I will have audio for you in a few days from the Strand in New York, where I will be with Jonathan Colton. All right, talk to you later.
- [01:33]
B
Thank you all so much for coming to Kepler's today. I'm Heather Burchell, the program manager here. I hope that many of you have been to some of our author events here, but if this is your very first author event at Keplers, then welcome. Now we all know that magical feeling when we come across the word for very specific situations or feelings, like schadenfreude means feeling satisfied from other people's misfortunes. Trappen witzen. This is the perfect comeback that occurs to you only after you've left the conversation under walking away. And for my fellow colleagues at Kepler's and anybody who shops here often, sundoku is the Japanese word meaning acquiring books and letting them pile up unread. Last year, Cory Doctorow coined nshitification, which refers to the gradual decline of a platform of service as it prioritizes profit over users without naming any names. Amazon is a great example of this. This phenomenon is also increasingly being applied to AI. Now, there's no shortage of books written about AI or in fact, books written by AI. But you won't find any of those books at Kepler's, obviously. But this button sized book that we're celebrating today, the Reverse Centaurs, Guides to AI how to Think About Artificial Intelligence before it Too late. Before it is too late, is the one that we all have to read. AI has colonized our lives and as Doctorow writes, AI genies now magically Appear all the time, whether we're sending a text or an email or taking a picture. It's both useful and harmful. And Doctorow has come up with a term for that, reverse centaur. And he is gonna be up here. We're so lucky that he's gonna be up here describing this concept to you. Cory Doctorow is an author, activist and journalist. He has written for just about every age group. He's written a picture book, the Little Brother series for young adults, and several sci fi novels. He was up here talking to Charlie Jane Anders just last year about his high tech thriller Red Team Blues. He's also written several nonfiction books and certification, of course, the Internet con, Checkpoint Capitalism, and talking to him in the green room. He definitely has more up his sleeve. One of the Bay Area's most skillful and elegant journalists, Angie Koiro, is going to be up here asking the questions. This is a subject that is particularly dear to her. So I think for all this is going to be a really, really lively conversation this afternoon at Kepler's. You will get a chance to ask some questions right at the very end of the conversation. So you can be thinking of those. And as you know, Cory Doctorow will be staying to sign copies of your books straight after this presentation. So without further ado, let's give a great big hand to Cory Doctorow and Angie Coiro.
- [04:44]
C
Great crowd. Wow.
- [04:49]
A
Hi, everyone.
- [04:53]
C
This is pretty awesome. I have been wanting to interview Corey forever. It's not about me, but I have been wanting to interview Corey forever. So that's really cool. Well, where Heather left off, initiatification. Draw the through line between that book and this book.
- [05:10]
A
Well, I think that if you want to connect the two theses about insidification and the core thesis of this book, which is really an exploration of the material ideological basis for the AI bubble and also set of tactical notes about how to undermine that material on an ideological basis to diffuse the bubble before it absorbs the life savings of too many people who are just trying to not starve to death in their old age. Which is the whole point of a bubble, is to separate people who are trying to just survive from all of their money and transfer it to insiders who are lying to them about market opportunities. So if there's a connection between those two projects is they're both exploring different negative effects of allowing firms to secure and maintain monopolies. And this is a kind of
- [06:05]
D
iron
- [06:06]
A
law of economics that if you let companies go long enough, they'll form a monopoly. It's not the Great forces of history. We just had this policy we adopted 45 years ago called the consumer welfare theory that said that monopolies are efficient and we should encourage their formation. That's why we have monopolies. We didn't get hit by a meteor, right? We just have monopolies because we wanted monopolies. And so, you know, in Inside of Vacation, we talk about how a firm that's too big to fail becomes too big to jail, and then it becomes too big to care. And so they can lock you in and then squeeze you. And in this book I talk about this seeming paradox of the monopoly, which is that when a firm saturates its market, right? When Google has a 90% search market share, it can't grow, right? Like, how do you grow a company with 90% market share? I mean, you can raise a billion humans to maturity and hope they become your customer. That's a product called Google Classroom. And the problem is that it takes like 15 years to pay off. And the stock market wants to see growth now and not because growth is like an ideological thing. I think progressives have done ourselves and other people a disservice by repeating this Eternal growth is the ideology of a tumor phrase. It's not an ideological matter, right? Companies that are growing are more than companies that are mature. And the reason for that is that a share of a company is a claim on its future earnings. And so if you have a company that makes a million dollars a year and it's going to make a million dollars next year, and it's going to make a million dollars the year after that, you can see why that company, a share of its future earnings would be less than a company that made a million dollars this year, but might make 1.5 million next year and 2 million the year after that, and 5 the year after that. And the corollary is that when a company stops growing, it is grossly overvalued because now it's mature. And so what we see are these corrections whenever growth tapers off in these companies where suddenly the stock market does these panicked sell offs. And so if you're an executive at one of these companies who've been compensated in shares, well, it's not because of ideology that you don't want the share to fall off a cliff. It's because that's where all your money is. And not only that, if you plan on growing by buying other companies with shares, you better hope you have a growth stock with highly liquid shares, because companies will take those instead of cash. Whereas if you want to buy a company for cash, you've got to get it from a creditor or a customer or an investor. If you want to get more shares, you just type zeros into a spreadsheet, right? It's an endogenous substance to the firm. So you have this paradox, right? When firms cap out their growth, they have to act to build a narrative that investors will believe about how they will continue to grow. They started by claiming they would absorb each other. Facebook said, we're going to become Google because we do the pivot to video, and that's going to be YouTube. Google said, we're going to become Facebook because we have Google, and that's going to absorb Facebook. The problem with that is that when you claim that you're about to become Facebook, there are a bunch of Facebook people who will say that you're not, right? And so then you have this argument in public. The upside to it is, like, everyone knows how much Facebook is worth because the balance sheet's just a matter of public record. By contrast, if you just make stuff up and say, we're going to conquer this market, that doesn't exist, sure, no one knows how much it's worth, but who's to contradict you when you say it's worth infinity? And so we've had this series of bubbles. We had cryptocurrency, web3 metaverse, and so on and so on. And in the manner of a man running across a river on the backs of alligators and not losing a leg. Every time one of these narratives started to get a little washed and stale, there was a new one. And AI is the latest. And AI is different, partly because I think there's more to it. I think from a computer science perspective, it's very interesting, but also because it's so much bigger. $61 billion set on fire to build the metaverse. 1.4 trillion so far in AI and 700 billion of it in the last year alone. So from a sector that makes billion dollars worldwide gross. So that's the connection, right? Monopolies who saturate their markets start to do very weird things.
- [10:11]
C
But you mentioned cryptocurrency, and you mention it in the book as well. What we learned from the very beginning of the book, and I have to commend you how simple you make it. For those of us who don't follow stocks and investments, there's this through line where this is built of people, both what's happening with AI and what happened with cryptocurrency, where you have to believe something that doesn't necessarily exist. And when there was an old English comedy shtick where there was an apartment house that only stayed up as long as the people who lived there believed in it.
- [10:45]
D
Sure.
- [10:46]
C
And as soon as it started out, it would go like this.
- [10:48]
A
Right.
- [10:49]
C
And then they remember, oh, yes, it exists. And it just reminded me of that.
- [10:53]
A
It's very Tinkerbell coded. Yeah. Yes. So, you know, what I would say is that the idea that someday we're all going to go to work in the metaverse and we're not going to have Disneyland anymore because you'll just go there in the metaverse and the sex trade will collapse because it'll just be in the metaverse, and so on and so on and so on. That. That requires an enormous amount of belief, as does the belief that this thing that is from, again, from a computer science perspective, very interesting, will take everyone's job. That it will not only continue to grow in terms of refining the capabilities it has, but that at a certain time, quantity will take on equality all its own. And by. By becoming better at guessing words, it will become actually intelligent and maybe then become God. And this. This requires a leap of faith, Right. That I think is very. It's become very doctrinal. Like, we have a hard time getting away from it. But when you step back one step and you say, wait, the word guessing program waking up, becoming intelligent. I mean, to use an example from Stanford, where they bred horses, this is like saying if you breed horses to run faster and faster, surely one of them will give birth to a locomotive. Right? Like the idea that just guessing words better is the same thing as consciousness does require this leap that I think is not present, and it has become so saturated in our culture. So I'm flying on the red eye tonight to Toronto and doing a conference talk and a panel on Tuesday and then a bookstore event. And I was on the call with my panelists, and it's about. The panel's about digital sovereignty and the risks of digital sovereignty. And all the co panelists were talking about how Canada needs its own sovereign AI. I'm like, wait a second. What happens to Canada if we switch off the AI? Like, which firms fail? Right. Which government functions can no longer proceed? They're like, well, I guess none of them. And I'm like, well, what happens if Trump orders Microsoft to turn off M365 for every ministry of every province in the federal government? Oh, well, Canada's bricked. Right? And I'm like, so why. Why is your Entire focus of digital sovereignty, a thing that no one uses for anything important to the exclusion of the thing that is actually very important. And by the way, say we did pull off. And like all the best Americans, I'm Canadian, which is why I'm saying we
- [13:23]
D
here
- [13:25]
A
say we did pull off a national AI. What we would have won in winning this was the right to lose money faster than anyone's ever lost money on anything ever before. Again, this feels like a very strange way of describing a project of like nation building. Right. We've managed to neglect the thing that is at our existential risk in order to lose money. Does feel like a weird pitch for sovereignty.
- [13:52]
C
One of the things that's interesting to watch is the number of times AI has evidenced itself as being not quite what it's billed as. And you mentioned in particular Elon Musk presenting his how far he's gone with AI in robotics. If anything should have killed off the whole AI industry, that should have done it.
- [14:10]
D
Yeah.
- [14:11]
A
So, I mean, you will remember that Musk has now twice done demos of his robots. And the first one literally turned out to be a person in a robot suit. They were like, how did they get a robot to dance so well? Just a person in a robot suit. The second one was a remote control robot. Right. And again, it was like. It's an autonomous robot. Bartender. Not really autonomous.
- [14:31]
D
Right.
- [14:33]
A
And again, it's very smoke and mirrorsy. Like it's almost too on the nose. This keeps happening. I was just talking with someone about this mailing list I've been on forever and ever. That's of Indian geeks. It started during the PC boom in India. And so it's all these old technologists and they have all these wry jokes about AI. So one of them is that AI stands for Absent Indians because so often AI is just a low waged worker in the Pacific Rim pretending to be a robot. The other One is that ChatGPT stands for Gujarati people typing, which is very funny, but you see it all the time. I mean, you know, Jeff Bezos had these stores, these just walk out stores that were grocery stores where you walked in and you took anything you wanted and you put in your basket and you walked out again. It turned out that those were not AI cameras that were assessing what you put in your shopping basket. It was three workers per customer in India watching you over cctv, trying to come to a quorum about what item you put in your basket. Right. And this is just like trying to. This is like, there's. It's one Thing to fake it till you make it. It's another thing to just pretend that a ro. That a person is a robot. And there is something again, a little on the nose about a program, or rather a product whose main source of investor capital and whose main sales pitch is this will eliminate people and you won't have to navigate the difficulty of people. You can eliminate all the moral consideration that you would have to extend to employees or to lovers or to doctors or to whomever, right? And you can replace it with software that you can just snipe at cab drivers. Whatever you can replace with software, you can be as rude as you'd like to and extend no moral consideration to. And then it turns out that it's not really a robot driving your cab. It turns out it's people pretending to be robots remotely driving your cab. Which is what happened when a cruise had to leave San Francisco after one of their cars hit someone and dragged them 20ft. And there was an inquest and we got to look at how their software worked. And it turned out the way that their software worked is they'd replaced one low wage cab driver with one and a half high waged engineers, which again has had to turn like $10 million of real estate into $25 cash.
- [16:51]
C
When it was revealing too that in the case of the woman who got hit and dragged, one of the boosters of autonomous cars said, well, it's the pedestrian's job to get out of the way.
- [17:02]
A
Yeah. And this is not new. I mean, if you imagine when cars first appeared, roads were not for cars, right? Roads were for people and cyclists and horses and carts and everything shared the road in a kind of glorious higgledy piggledy. And then the car came along and it was just like monumentally unsuited to sharing the road with other vehicles. Doubtless notice whether you are driving a car or walking near cars, we understand that cars just don't play well with others. And so they invented the term pedestrian, right? And also the term jaywalker to describe this kind of eternal verity that people shouldn't be in the streets, that the streets are for cars. And they basically like retconned the idea that the streets were always there waiting for the cars to come along. And now that the cars are here, the pedestrians ought to get out the way because, you know, you've enjoyed your free ride while the cars were being built. But now the cars are here, it's time to clear out. And there's a very famous AI booster who talks about the pogo stick problem. And he says, well, we can probably build an av, an autonomous vehicle that can deal with pedestrians who are well behaved. But if you ride a pogo stick into traffic, it's unreasonable to think of an AV that's capable of managing it. And, you know, it sounds superficially reasonable, but the rhetorical move here is to say, once again, like, when you get killed by the robot, it's actually your fault for not being considered enough of the robot's limitations. Right. You were just, you know, the hypothetical example of person logo stick expands to whatever it was that person was doing that we can use to legitimate the use of the road by robots to the exclusion of people.
- [18:49]
C
It's interesting then when I know a lot of people here heard about when activists in San Francisco were putting traffic cones on cars. And at first it seemed, well, how juvenile. But once you realize how angering this all is, you kind of get it.
- [19:02]
A
Yeah. And, you know, I think that the traffic cone thing is actually quite useful in as much as it shatters the illusion that the car can really solve the problem of driving and navigating the roads, that, you know, it's very easy to make things that work well if you don't care how they fail.
- [19:20]
D
Right.
- [19:20]
A
You can build cars that can deal with all of the anticipatable solutions and pilot themselves to traffic. But just like as with so many things, the last 10% is the hard part. Right. Dealing with the unanticipatable, dealing with the contingency, dealing with the things that, you know, a human being might struggle with but could fail more gracefully with, that's much harder. And when we're talking about two ton robots that travel at 90 miles an hour, that last 10% really starts to matter.
- [19:50]
C
Got to get back to some definitions. You talk about the Centaur and the Reverse Centaur as being the difference between a useful tool and a technological torment.
- [19:59]
D
Yeah.
- [20:00]
A
So obviously the idea of studying how workers and automation connect to one another, it's not new.
- [20:06]
D
Right.
- [20:07]
A
It's, you know, the foundation of science fiction is really asking these questions. The first science fiction novel, Frankenstein, written by Mary Shelley, was an allegory in support of the Luddites. Mary Shelley was a Luddite supporter. As was common of the day, the Luddites were really, really popular. It was only afterwards that the history got rewritten by the people who defeated them. The Luddites were skilled technical workers. Right. To work on a textile machine, you needed a seven year apprenticeship. It was like getting a master's degree in mechanical engineering from mit. They weren't angry about the machines, they were angry about the social arrangements of the machines. Shelley, her novel, was about Luddism and about this idea that when workers are not allowed to have co determination and automation, that you have these horrible hubristic machines that are unleashed on the world to the great detriment of everyone. Her husband's maiden speech in the Lords, when he was made a lord, was a speech in support of the Luddites. It's really hard to overstate how popular the Luddites were in their day. So, you know, from the Luddites on, there has been a broad understanding that when workers are in charge of automation, they tend to adopt automation in order to improve quality. But when capital is in charge of automation, it tends to be to improve throughput. And that's not because of the sinister character of the boss. It's because of the economic relationship of a worker to a tool versus a boss to an asset. When you acquire an asset and then it starts to depreciate off your books, you want to maximize your return on that asset, which means you want to maximize the utility you get from it, which means you want it used as much as possible. And broadly, machines work faster and longer than people. And so as soon as there's a machine on the premises, as the person who bought the machine and then employs people to use it, you want those people to use the machine as much as possible, to produce as much as possible, to recoup as much as possible and make as much profit as possible. You're sweating the asset. And so this is where we come to centaurs and reverse centaurs. A centaur is someone who chooses to use a machine to make their work better. They partner with a machine in order to do things that they can't do. And they set the pace and they direct the machine. So a centaur is a human head on a machine's body. You have something tireless, you have something that's fast, you have something that can do physical feats that the human can't do. But the judgment is left up to the person. You ride a bicycle, you're a centaur. You even look like one. Use a spell checker, you're a centaur. You use an IDE to program or a debugger. My first programming environment, a basic interpreter on an Apple II plus, did not have a debugger. All of these things are ways of becoming a person in charge of a machine. And the machine helps you do your work. A reverse centaur is the reverse. A reverse centaur is when the machine directs the activity of the human. And the human is conscripted to be a peripheral for the machine. Think of an Amazon driver watched by all of those cameras. Your route is set by AI that's determining where you have to go. And your pay scale varies by up to 500%. The minimum pay scale per parcel is 10 cents. The maximum is 50 cents. And that variation is entirely based on your compliance with the machine's orders.
- [23:20]
C
Right.
- [23:21]
A
So that's a very reverse Centaur arrangement. It is characteristic of automation that is driven by capital, that that's the arrangement we come to. And it's why the most automated warehouses in America, which are Amazon warehouses, have the highest rate of injury in America. Because you're not just being used by the machine, you're being used up by the machine. And again, it's not new, right? The Luddites were attuned to this. You know, the reason that the entrepreneurs who bought the steam mills and the stocking frames were interested in them were not just because they were easy to use. It was specifically because they were so easy a child could use them. And the children they were thinking of were Napoleonic war orphans, of which there was an enormous surplus in London who could be effectively kidnapped to Manchester to 10 years of indenture on the machines. That's why they had to be so easy a child could use them. Many of those children died or were maimed. One of the survivors was a guy called Robert Blinko, who wrote a memoir that was such a bestseller that it inspired Charles Dickens to write Oliver Twist, which is Luddite fanfic. And, you know, you see this as a recurrent motif, right? Charlie Chaplin caught in the gears in Modern Times. Lucy and Ethel trying to get the chocolates in the chocolate box at the end of the assembly line. Why is the assembly line running at the edge of human endurance and reflexes? Because you work the machine to the rate at which the slowest component can go, and the human is the slowest component. So you work it to the maximum rate that the human can work at. And that's how systems are often designed. When you couple multiple components in a system together. And so this is centaurs and reverse centaurs. And it resolves another seeming paradox, which is how is it that you have people who are skilled workers who are historically reliable narrators of their own experience with. With their tools, and some of them are using a tool, and they describe it as the most powerful increase in their capability as a skilled worker that they've ever seen. And others are using that tool and they're describing it as a source of immiseration. And not just that, but it's such a drastic reduction in quality that you hear programmers say, get your money out of the banks. Because I've been, you know, they've been rewriting their software and I've been. I've had a front seat for it. And oh, my God, they're just going to get. They're going to get wiped out by hackers. Right. Or don't ever fly in a plane again or whatever. So how is it you have these two groups of workers who have this seeming contradiction? And the answer is, it's not the tool, it's how they're being directed to use it. The first group are centaurs. The second group are reverse centaurs. The first group is figuring out how to use the tool to do their job better. The second group has seen 9/10 of their colleagues fired. And they've been put in charge of marking the AI's homework and taking the blame when the AI gets it wrong. And it's. It's not what the gadget does, it's who it does it for and who it does it to.
- [26:07]
C
Parenthetically, as I was reading. Hey, we just lost my audio.
- [26:11]
A
I think that's why you got a stick mic there.
- [26:13]
C
I think you're so lucky. Okay, let's get this on. I'm back.
- [26:19]
D
You're back.
- [26:20]
C
Testified. Okay, thank you. But parenthetically, I was wondering as I was reading all of these samples, and now you bring up the Napoleonic orphan, the triangle, shirtwaist fire. Thank you, thank you. I appreciate that. I just found myself wondering how much of the situation we're in now would be the same if we had that acute knowledge of labor and the history of labor that we used to have.
- [26:47]
A
Yeah, I do think that, like, civilizationally, we have so little object permanence that we'd lose a game of peekaboo.
- [26:53]
D
Right?
- [26:56]
A
Because, like, these are not, you know, like, AI is doing new things. It is certainly like, and I know I've said this like three or four times now, I'm going to say it again from a computer science perspective. AI. Super interesting. But the issues that AI raise, although they have their own special character that is in relation to the contingent characteristics of AI and what it does, they also have so many commonalities with existing problems that we've had, and yet we always come afresh to this as though nothing has ever happened before. And moreover, when we experience problems with these systems that Echo problems we've had before. So bubbles, as I said before, they're really bad and really destructive and they steal money from ordinary people and they provoke economic crashes that are met with austerity that drives people into the arms of fascists. They're really bad. And we've had this string of bubbles, and every time a new bubble comes along, we're like, surely this can't be a bubble, right? Like, why would they spend this much money on it if it wasn't real? Right? Like the thesis that the pilot shit is big enough, it must have a pony underneath it somewhere. You know, the fact that we just can't remember this from day to day and year to year, and also that we can't remember the public officials who presided over the policy choices that made a fertile ground for the bubble and that failed to visit consequences on the people who created the last bubble and stole so much money from so many ordinary people. You know, that is like, as a policy person, I should say I work for the Electronic Frontier foundation. It's my 25th year with them as an activist, I have. Thank you. I hope if you don't support them, you will. And if you do support them, I thank you. Efficiency. But in 25 years, I've had many opportunities to see policies be made where we said, this has the foreseeable outcome of bringing about this harm, and then the harm arises. And people are just like, why are you blaming us for this? How could we have foreseen this? Right? This is not. And maybe it's not even connected to my policies, right? Like, the idea that we should tolerate monopolies. It's not like no one in the moment was like, you know, monopolies are bad, right? You know, like, we had a bunch of problems with monopolies. Like the first anti monopoly law, 1890, Sherman act, was passed in the face of the Granger movement. Farmers who got pretty Luigi Mangione about railroad executives, allegedly, right? So, like, we knew what happens with monopolies. And then we got all these monopolies because we decided to tolerate monopolies. And you go to these economists who told us that monopolies were efficient and that we should both tolerate and encourage them. And they're like, why are you blaming me? Right? It's as though we didn't used to have a rat problem because we were using rat poison. And these guys were like, why do you need the rat poison? Now rats are eating our faces off. And when we call them on it, they're like, why are you blaming me? For the rat problem, this is like the time of the rat. Right? The iron laws of economics have made rats more fecund than at any time in human history. And if the rats did buy all the rat poison factories and shut them down, well, if you're not using rat poison, that's just economically rational. Stop blaming me for this. I have to go polish my fake Nobel Prize now.
- [30:14]
D
Right.
- [30:16]
A
Nobel Prize in Economics. Not a real Nobel Prize. And so, you know, it's very frustrating. It can be a little demoralizing to exist in this world of the eternal now, where we are as babies who do not know what happened before and therefore can't predict what's going to happen next. But I do think that there's a real tonic in writing books that connect the events of the past to the present, even the recent past to the present.
- [30:44]
C
Well, you also have to look at the role of wishful thinking. You talk about the eternal goal of the employer not wanting to have to deal with nasty human beings being, you know, employees get in the way. And then you have the people who've invested in one bubble after another, and they can see through all of the fallacies of what's happening with AI. You mentioned an MIT study where it showed that 95% of the applications saved no money at all, didn't increase profits at all, and yet their wishful thinking has them continuing to invest. Yeah.
- [31:14]
A
You know, Sartre said, hell is other people. He wasn't wrong. But it's not because other people are bad. It's because they're good. You need other people to do anything that exceeds the things you can do on your own. And yet other people stubbornly refuse to acknowledge that everything you want is right and they should just do what you tell them to do. As an Angelino, I'm keenly aware that other people do not acknowledge that when I need to get on the five, which is what we call it in the southern part of California, when I need to get on the 5, everyone else should leave so that I can go where I'm going. And, you know, the fantasy of a world without people, it's a very powerful one. And I think if you're in the ultra rich, if you're a billionaire, you kind of live surrounded by the human equivalent of AI, right? You are surrounded by people who never tell you you're wrong, who, when you come up with an idea, just go and do it. If it comes out horribly wrong, they make up excuses why it's not your fault. If you insist that they keep doing it they come up with the most incredible workarounds. They pour peroxide into the reflecting pool to prove you right. They arrest people who go and look at for being saboteurs so that you can preserve this ego health. And I think that it's a very solipsistic way of living. And I think that on the one hand, that primes you to believe that AI can replace people, right? And I think if you run an enterprise, especially a large firm, and you know that if you don't show up for work, everything is fine, but if the people who work for you don't show up for work, the business shuts down, that you're not really in the driver's seat, you're probably in the backseat with a Fisher Price steering wheel. And with AI, maybe you can fantasize about wiring the toy steering wheel directly into the drivetrain. So you have an amazing idea and no one tells you it doesn't scale. No one tells you you're going to jail for making it. No one tells you about all the people you're going to harm by deploying it. You don't have to convince people that they should stop doing what they're working on because it's like holding the business up and come work on your idea. You can just vibe, code it and poop it out and then you get someone around, someone else to come and fix up the edges and keep it running and sort of nurse it along. There's a great line in Get Shorty where the two criminals are talking about writing a screenplay and they're like, it's not that hard. You just write it down, whatever you think should go in there and then you hire someone else to put the commas in, right? And so this fantasy, right, of not having to have these ego shattering confrontations with people who know how to do things, it's very seductive. And I think that as a billionaire, on the one hand, like, you know how seductive that is. And so if someone says to you, I'm going to send out a salesforce and all of the sales calls are going to go, how would you like to have a workplace without workers, right, that they're going to, you know, they're going to sell a lot of stuff and also you kind of want it to be true, right? Now, I don't think this is unique to the ultra wealthy and I don't think it's unique to capital. I think one of the reasons we see AI partners, particularly AI boyfriends, because speaking as a man married to a woman, I think it's objectively harder to have a male partner than a female one. The reason we see it as hell is other people, right? Romance is hard. Relationships are hard, right? The screenplays without screenwriters, movies without actors. When I worked in a bookstore, our joke, it was terrible, but I still laugh was retail. Be great if it wasn't for the customers. You know, I've just finished shelving these books. Stop messing them up. You know, that, that, that there is something very seductive about the fantasy without people in it. And you know, you see it like echoed in some of the kind of more outlandish parts of the way the ideologies have swept through these circles, like this long termism effect of altruism that says you should earn to give, which is to say you should inflict as much harm as you need to on living people today to earn as much money as possible. But then you should give that money to benefit people, but not people alive today. You should give that money to benefit 10 to the 53 artificial humans who will be on the planet of Venus in 10,000 years, right? You could not get more solipsistic than this, right? This belief that it doesn't matter who you harm today, provided virtual people in the future have a better life, right? It's a very kind of ketamine ish belief. You know this the. I have chronic pain. I once had a ketamine infusion. There comes a point in the bottom of the K hole where it feels like you've just imagined all of your life and you've just like you're, you're the, you're the king underneath the tree and Alice through the Looking Glass, imagining the world. And if you wake up, the world will stop existing, right? It's that we are living through a moment in which other. The inconvenience of living with other people has a group of people who propose to solve that problem and not solve that problem the way other people have with like relationship advice or unions or organizational advice or politics, but just by making everyone else obsolete.
- [36:11]
C
So much of what you brought up here and so much of reading the book kept reminding me of a guest we had here earlier, which is Theo Baker, who wrote how to Rule the World. And one of his assertions is that there's a rise of at least amorality, if not immorality in a Stanford pipeline. And the students are essentially learning that people are there for what you can wring out of them. And there's a direct line out of Stanford into Silicon Valley. I just want to put three things Together, if you can get a feel on that. We have everything we learned in your book. We have everything from learn to how to rule the world. And then we have this long conversation online and at Reddit, attributing the growing lack of warmth in San Francisco and on the Peninsula to the growth of technology, amorality and immorality. Can you put those three together?
- [37:01]
A
I mean, so there's. There were the. Paul Krugman had a long conversation this week on his substack with labor economists, and he was talking about how labor economics is the. Is the most rigorous area of economics because it has all these natural experiments where you have things like one state raises its minimum wage and the next state over doesn't. And at the border they have really similar economies. And you can say, like empirically, what happens when you raise the minimum wage? Does it depress employment? And so on. And one of the empirical findings is that businesses, when a business is run by someone with a business degree. So you have this other natural experiment where oftentimes a family firm is taken over by someone with a business degree when they, the descendants of the founder, no longer want to run the business. So you can just see what happens when a firm is taken over by someone with a business degree. Wages fall, but productivity doesn't go up. So it's just a pure transfer. Executive compensation goes up, employee compensation goes down, but productivity remains flat. So there is a sense in which the role of the MBA is to teach you how to lower wages, right? Not improved productivity, but to lower wages. I say that as someone who used to work at an MBA program, right? Like. And there's lots of things they teach you in MBAs that I think are useful, like modeling and so on. But I think that there is at least if the purpose of a system is what it does so without interrogating the morals of the people who teach there, which maybe we could do, but you don't even have to get that far, right? You can just say if what happens when you go get a business degree is you go out into the field and where you arrive, wages for workers go down and compensation for executives goes up, but productivity remains flat, then we can say that the ideological purpose of the mba, irrespective of the ideological basis for the mba, the ideological purpose for the MBA is to teach you to transfer wealth from workers to executives. And so, you know, I think we know when we look back through history, we know that we went through this period where economists decided that all of economics could be treated as a kind of physics Problem that you could solve everything with models. This is kind of the neoclassical way of thinking about economics, that everything can be solved with models and that therefore anything that you couldn't model well, that didn't quantize well, like power, should be discarded. And that even if you knew power was important, you could safely leave it out predicted, provided the model predicted things. So this is what Mill, Milton Friedman said that all models are wrong. Some models are useful. If the model predicts things well, it doesn't matter if you know your assumptions are wrong. And this leads you into this position where you habitually decide that power doesn't matter and outcomes that the model, if it predicts things well, sometimes, is always predicting things well. This leads you to these very torturous chains of logic. So one of the bedrocks of this economic theory is something called revealed preference. So the reveal preference is the idea that if someone says they want one thing but do something else, that what, what they do matters than what they say. And you know, there's a sense in which that sounds about right. But where it leads you is you observe someone in the world who sells their kidney to make rent and you say, look at that person with a revealed preference for one kidney, right? If you have like the specific kind of neurological injury that you get from doing one of these economics courses that makes you incapable of perceiving or reasoning about power, you go into the world and you assume that anytime you can bully someone into doing something that coercion is the same as consent, that they cash out to the same thing that the signature at the bottom. And if you wanted the most perfect epitomization of this, just look at the end user license agreement. By being dumb enough to use this product, you agree that we're allowed to come over your house and punch your grandmother and wear your underwear and make long distance calls and eat all the food in the fridge. I mean, you know, click okay to continue. Right. And by the way, you've just agreed to binding arbitration. So even if we violate whatever you think you're getting from this, you can't even go in front of a judge to sue us. You have to go in front of a lawyer we paid to decide whether we're wrong, right? I mean, like that, that, that is like the, that, that is like a foundation of coercion built into so much of what we do online. So I, I can't imagine that if you work in a firm that this is the way it conducts business, that it wouldn't rub off on you and that it Wouldn't at least create a habit of thought in as much as if there was a dispute within a firm, like a factional dispute, where one group of people want to do the right thing, the other group of people want to do the wrong thing. But the people who want to do the wrong thing always win that you would at least start to be able to model their, their thought process. Because the only way you could get something done is to understand what the, what the devil on the left shoulder of the firm is whispering, even if you're one of the angels on the right shoulder. And I know I've spoken for a long time answering this, but I'll give one quick example from Inshidification and this is based on reporting at Zitron did. When Google had the search antitrust case, the DOJ published a lot of their email that was relevant to it. And one of the things we saw is that when Google saturated its market in 2019, 2020, their search revenue growth stopped because they had 90% search market share. And two different factions had a dispute about what to do about this. And the faction that one was a faction led by the revenue guy. He was an ex McKinsey guy called Prabhakar Raghavan. And his plan was to make search worse because if you have to search more than once, you get to see ads more than once. So he's like, we'll just turn off query stemming where if you search for trousers, we also do a search for pants in the background and merge the results. We'll turn off context awareness. So if you search for algae, when the reflecting pool is full of algae, we put those links at the top. We turn off spell checking. So it just takes two or three queries on to answer your questions. The other group was led by a technologist named Ben Gomes who, old school, you know, Googler, like, you know, overseen their server build out from like a couple of computers under a desk to the, you know, worldwide data centers. Who's really horrified by this but. But can't win the argument. You just see him like trying all these different arguments, but money talks and bullshit walks. Google had bought all of the places where you could see a search box, right? They're paying Apple $20 billion a year not to enter the search. They were buying the default search on every browser, every manufacturer, every carrier, every search box is wired into, into Google. So it's like you go into the grocery store and all the shelf space belongs to Procter and Gamble. Why shouldn't they make the products worse? What do you where are you going to find another product? Right? So that's how he wins the argument. So you can see how even if you want to do the right thing inside the firm, that if the policy environment is in shitagenic, right, if it's designed to reward the worst ideas of the worst people with the most money, doesn't matter how much you want to make things work, because the gravity is dragging the firm in shit words. And you can pull it and pull it and pull it, but gravity is remorseless and is always going to pull back and it will never stop.
- [44:08]
C
And you could always say that, well, with our government, we can't count on them to rescue us from all this. You mentioned the fact, for example, you're not allowed to reverse engineer certain things so that you can get on top of the way they use the equations. You can't say, well, Expedia is showing you this cost and showing somebody else another cost. Well, you look at where we are now and when you look at, you know, the inauguration and you look at scotus, you could make the argument that we're in the worst possible place for the government to help us with this at all.
- [44:39]
A
We are. But on the way here, I turn in the third draft of my next book, which is called the Post American Internet, and it's about what the rest of the world can do now that Trump has convinced everyone that they no longer have to care about what America wants from their technology policy. So global technology policy has really been determined by the US Trade Representative who said you must have certain rules and not have other rules. And the carrot and the stick was tariffs and free trade, or rather the other way around to free trade and tariffs that either you do what America says or you will have tariffs on your exports. But if you do what America says, you can export freely to America. And so every country in the world has enacted what's called an anti circumvention law. And these are laws that make it illegal to reverse engineer and modify things. So if your printer comes, and it's designed not to accept third party ink, ink modifying your printer so that you can put anyone's ink in it, in this country, it's a felony. It's punishable by five year prison sentence and a $500,000 fine for first offense under section 121 of the Digital Millennium Copyright Act. But it's also illegal in Europe under Article 6 of the Copyright Directive and in Canada under the Copyright Modernization act of 2012. In Mexico, it came in through USMCA in 2020. Every country in the world has a version of this law. But here, what it means is that the s and P500 and the NASDAQ do really well because everyone's paying $10,000 a gallon for printer ink, right? Everywhere else, it means that you can't go into business. You can't implement the Bezos plan. Your margin is my opportunity. You can't go into business jailbreaking printers and selling ink for pennies on the gallon. And so the basis for that, that for allowing the people in your country to have their pockets picked and their data stolen by American tech giants, was free trade with the US So happy Liberation Day, everyone, right? The economic case for or for keeping this law on the books has collapsed. And then to make the case even more thoroughly, Comrade Trump has gone around the world and convinced people that they can't trust American technology either. So the chief prosecutor, the International Criminal Court, lost his Microsoft accounts when he swore out an arrest warrant for Benjamin Netanyahu. Now, Microsoft says that this was a coincidence and that saying that this was really bad and he should be punished has nothing to do with it. It's one of those he said clip, he said things. I know who I believe, because it happened again when High Court justice in Brazil sentenced Giobro Senoro to prison. Trump's ally, who had been the dictator of Brazil, who staged a coup after he lost an election. It keeps happening. So, you know, if Trump wants to take Denmark, he doesn't need to roll tanks, right? He can just brick every ministry in Denmark, right? And like Hormuze, the world's supply of, you know, Ozempic and Lego and deliciously strong black Licorice, right? That is a giant geopolitical risk, right? So now you have people like me, digital rights hippies who spent 25 years saying, it's bad for you to lose your privacy, it's bad for you lose your technological autonomy. We got so far, no further. And then you have a bunch of entrepreneurs and investors and technologists who'd like to actually make a dent in the universe and not just do a fake startup so it can be aqua hired by a company that mostly innovates by doing stock buybacks and chatbots, but who actually want to make a thing that people like, right? Who now have been chased out of this country one step ahead of a guy in, like, a gator in Oakley's and, like, a plate carrier over a Totenkof tattoo, who are now back in their country doing reverse Brain Drain and who would like very much to turn America's trillions into their billions, right? But you also have national security hawks who are like, wait, they can turn off all the John Deere tractors? That's what they did when and Putin's thugs stole all those tractors from Ukraine and took them to Chechnya. They could do that to us too. And the courts aren't going to protect John Deere if they don't want to do it. So, you know, this may sound like a bad deal for you, but I will put it to you that Americans are the beta testers for every terrible idea that American tech companies have. And that if we can buy reasonably priced pharmaceuticals from Canada through the U.S. postal Service, we're going to be able to buy the software that lets us jailbreak our printers, run different app stores on our phones, get our cars fixed by any mechanic we choose, turn off the surveillance in our cars, all of that stuff from wherever, the eu or maybe, like, maybe it's Ireland. Like, maybe Ireland's. Like, maybe the wrong regulatory arbitrage to do is tax cheating. Maybe the right regulatory arbitrage to do is like the tools of technological liberation. Or maybe it's Mexico. There are a lot of Mexican technologists who've just gone back to Mexico and who have an axe to grind, right? And frankly, they deserve it more than the rest of the world does. So maybe it's Mexico, maybe it's Nigeria, maybe it's Ghana, maybe it's India. India will get the double whammy of like, now we can also get all the Chinese firmware out of our solar stuff because anything that can receive an over the air software update can receive an over the air software update that permanently makes it unusable, right? So this is a great moment, right? Like, comrade Trump has done more to convince people to stop using American technology in one year than I did in 25 years, right? Like, he's also convinced more people to give up on fossil fuels than I did in all the summers I rang doorbells for Greenpeace. Like, he's doing praxis. He's an accelerationist. It's very weird.
- [50:20]
C
So we are gonna turn this over to the audience for questions. I have, have two more, not one more question, a two part question for you. I'm going to the last couple chapters of the book, putting together a couple sentences. AI companies are the most data hungry, privacy disrespecting firms in tech history, which is a tough, highly competitive field. They've never found a data source they will not scruple to ingest. Back to back with Another quote, if we are good AI critics, if we carefully identify the pathological aspects of AI and relentlessly target the financial basis of the AI bubble, we can make sure all the terrible things billionaires want to do with AI never happen. And we can consign all the terrible things that are currently being done with AI to history's ash heap.
- [51:10]
A
So look, some of you probably have been living here since I first moved here in the late 90s. And so you'll remember in the early 2000s when suddenly you could buy Aeron chairs for like a nickel and you could buy servers and office space and there was all kinds of amazing stuff. Now the bubble was very bad. The dot com bubble was really bad. It was like all the other bubbles, a way to transfer money from people who just didn't want to starve to death in their retirement to industry insiders who lied to them about how much pet food you could sell online and other lies of the day, right? But it left behind a very productive residue, right? They're like millions of humanities graduates who've been available to drop out of university and become Perlin, Python and HTML jocks. Built web2 once they no longer had to pretend to build dumb products for their bosses, right? Because they could hire each other. Their rents were cut in half, servers were too cheap to meter. Right? So there's a bunch of stuff that I think we will do with AI long into the future because again, super interesting piece of computer science. Once the GPUs are being sold at a nickel on the dollar by bankrupt companies, people will figure out cool things to do with them. They won't sweat the assets so hard that they burn out over two years or three years or five years. There'll be lots of unemployed people who work in the field who can hire each other or can go to work for people with more interesting ideas. And there'll be these open source models that have barely been touched, barely been optimized. And whenever anyone tries to optimize them, they get so much performance out of them. Remember Deepseek got 6 million bucks for from a Chinese hedge fund. They took a meta model and they tweaked it to see if they can make it run on the older chips that you're allowed to buy in China. And it was so performant that Nvidia lost $600 billion in market cap in 24 hours. It's the single largest decapitalization of any firm in the history of markets. Even inflation adjust makes South Seas bubble look like nothing. And so there will Be these models, we can get more performance out of workers who know how to use them and the hardware to run them on. And so everything you like about AI, if you like AI, I think you will still have. It's just the bullshit that will go away. And the sooner we get rid of it, the better. And so to do that, we have to remember that the pitch for AI, the way you go from $50 billion a year in gross revenue to somehow convincing people that you can make up $1.4 trillion trillion dollars in spending 700 billion in the last 12 months, is you convince people that it will do the jobs of skilled workers with high wages, right? Like robot taxis just don't matter in terms of the balance sheet of AI companies. You take all the cab drivers income and you wipe it off the expense side of an AI company, right? And you split the wage with the AI company. So you give Uber half of it and take the other half and you give it to the company that makes the algorithm. It doesn't come close to touching the sides of the, you know, vast hole in their balance sheet. Same with illustrators, right? Commercial illustrators are the most immiserated precaritized in the American workforce. Except for cab drivers and truck drivers. The other two groups of people, they say they're going to bankrupt, right? It is horrible that they're like gloating about the idea that they can take these poor people who just love doing their job so much, they'll do them under these miserable conditions and say we're going to bankrupt all of them with slop and we're going to use their own work to do it. But if they succeed, if you transfer the wages of every illustrator in America working today to Sam Altman, it does not pay for the free kombucha he gives to the workers who oversee the next run to train an image model, right? It's, it's not, it's Same with election disinformation. Not a highly waged activity. It's gross. It's terrible. It threatens our democracy. Ban election disinformation. A thing we should do and we should figure out how to do it. But when we ban election disinformation, we do nothing to the expected return on capital for AI investment. It's a peripheral use that soaks up some excess capacity at the edges. Same with deep fake porn. Disgusting, awful, terrible. We should do something about it. Not economically important to the bubble, right? It's not what investors are not going like. Okay, so what's the ROI on deep fake porn? Right? The ROI they're interested in is like radiologists, lawyers, teachers, doctors, skilled engineers, workers. Right. People who have high wages. Right. It's about eroding their wages. So we need to focus on when we talk about the impact of AI and its shortcomings. We need to talk about what it can't do in those fields and why the story about how you can fire 9/10 of the radiologists, take the remaining radiologists, put them in charge of marking the chatbots homework, and then make them into the accountability sinks who take the blame when someone dies because the AI missed something. Why that's not adequate. Why that's not a thing we'll permit. Regulatory, socially, medically. Right. That's the thing that you do to diffuse the bubble. And if you want radiologists to have AI tools, I think that's good. A little personal talk here. Four days ago I talked with a radiologist who told me I'm now cancer free, which is very good if the pitch is right. Now, radiologists review 100x rays a day and we're going to give them AI and twice a day it's going to tap them on the shoulder and it's going to say, go look at those two again. And so we have to hire another radiologist for the Kaiser system in Southern California to take up the slack as we reduce their throughput by 2%. If that's the pitch, I'm all for it.
- [56:36]
D
Right?
- [56:37]
A
That's the pitch that you get if we diffuse the bubble. If we don't diffuse the bubble, the pitch that we get is we all die of cancer and the radiologists get the blame and the hospital cuts its costs.
- [56:48]
C
Let's grab this though, before we move to the audience. I got to grab this. I'm hearing you say we have to do this, you have to do this. Who's we? I mean, for the individuals here, what do they do?
- [56:59]
A
So I have bad news for you, which is that shopping isn't politics. I would love, hell, as other people, I would love it if the way to solve all of your problems was to just make the personal political and just like think very hard about what you buy and only shop at the good stores and don't shop at the bad stores. You're not going to shop your way out of a monopoly. It's like recycling your way out of a wildfire. A boycott isn't shopping hard. Right? The Montgomery Bus boycott was not a bunch of people deciding to walk to work. Right? It was an organized movement and the organization was a feature, not a Bug the hours they spent organizing carpools, leaning on employers who were threatening workers who were showing up late because they weren't taking the buses. All of that formed the institutional backbone of the civil rights movement. That's what a boycott is, right? A boycott isn't just you making personality consumption choices. By all means, shop at bookstores instead of Amazon, because you love the bookstores and you want them to stay in business, right? But don't think that Jeff Bezos has less political influence because you didn't buy your books at Amazon, you bought them at Kepler's. Buy your books at Kepler's because Kepler's is amazing and deserves to exist. But fight Amazon by unionizing his workers and breaking up his company.
- [58:10]
D
Right?
- [58:11]
A
And so how do we get there? You have to be part of a polity. And this is back to hell is other people. You're gonna have to work with other people. It sucks. It's hard. I mean, it's great because other people are amazing, right? And solidarity will get you through times of bad politics better than politics will get you through times of no solidarity. And so knowing your neighbors, working with them, building bonds of trust and solidarity, this is gonna carry you through the hard times we have ahead of us no matter what we do. Because. Because there are big crises on our horizon. Authoritarianism, fascism, genocide, the environmental crisis. Right? The hits are just going to keep on coming. And so you need to have these bonds with your community. So get involved with the community. I mentioned EFF before@eff.org we have lots of ways that you can get involved with community groups in the Bay Area. Especially as you might imagine, EFF having been located here for about 30 years, we have deep roots here. Also, get involved in local politics. Like, you know, Moms for Liberty showed us that even the stupidest people in the world can get a shit ton done at the local level. Imagine what we can do with these unregarded local offices, right? It turns out that local offices are what determines whether or not there's flock cameras that help get your neighbors rounded up and sent to a slave labor camp in El Salvador. Right? Or shot in the face, right? That's a local matter. Local politics determines whether there's a data center in your backyard. So get involved in local politics. I work for a 501C3. I'm not going to tell you which party to get involved with, but I will say that local politics is much easier to get involved with than state or national politics. And then finally, get involved in workplace politics. Join a union and if there is no union in your workplace, start a union. And if you work in tech, start that union by talking to techsolidarity.org and techworkerscoalition.org, that's techsolidarity.org, techworkerscoalition.org and eff.org so that's your, that's how you get, that's how you solve this. Right. As part of a polity, buy the things that make you happy, shop in the stores that you love. Don't mistake that for politics.
- [60:18]
C
Excellent audience time.
- [60:23]
A
So we do things a little differently. You might want to have two or three small plates. No. So here's how I do these Q and A's. I'm going to call on people. I'm going to point to you, you're going to stand up and ask the question. I'm going to repeat it back for everyone. And I will say, as you think about your questions, I will say two things about questions. They're not more of a comment than a question and they have one part and not two. Okay, so if you'll put your hands up, let's start in the very back there. The gentleman with a gray sleeve. Yes, you, who just turned your head when I asked it. When I pointed to you. Yes, you. Yes, yes. You're looking around. Stand up. There we go.
- [61:00]
D
Yeah. From an AI perspective, we talk a lot about the worker. From a children's perspective, you know, we look at schools and how they taught the factory workers.
- [61:11]
A
Right.
- [61:12]
D
How do you see AI influencing children?
- [61:15]
A
Yeah. How does AI influence children in their future? It's a really good, good question. I am the child of two lifelong educators who then became pedagogists. So an elementary school teacher and a high school teacher who now oversee Ed D candidates at a university. And so I know exactly how much I don't know about education, which is a lot. But I will say that broadly, like all the obnoxious uses of AI are when people use AI to do things that they don't care if it's done well. And boy, does that describe a lot of the standardized curriculum.
- [61:45]
C
Right.
- [61:45]
A
Both from the teacher's perspective and the student's perspective, like the point of a five paragraph AP essay or a DBQ or whatever is to standardize work for assessment by different assessors so that if you hand an essay to five different assessors, you'll get the same grade out. Which means that it's not a good essay. It has to vaporize every qualitative element that might make it quality. So I understand why kids are doing it in terms of, like, an AI strategy. I understand. Am baffled to hear schools say we want to figure out how to teach kids how to use AI. To me, that's like teaching kids how to, like, vape. There is a giant. Not. Not because it's bad for your health, although maybe it is, but because there's a giant industry thinking every hour that God sends how to convince kids to use AI. Right. It's not like there is a shortage of people, like, messages in the world aimed at kids about using AI. Right. Like, it's.
- [62:33]
C
There's.
- [62:33]
A
There's a massive surplus of it. If you have an idea for curriculum that uses AI and it's cool, and you.
- [62:38]
D
You.
- [62:39]
A
You've thought of it and it works in the classroom, by all means. Right. But don't like the idea that we need this, especially right now.
- [62:45]
D
Right. Like.
- [62:46]
A
Like schools buying AI licenses right now, buying AI equipment, buying AI expertise. And right now, it's like buying Aeron chairs in March of 2020.
- [62:54]
D
Right.
- [62:55]
A
Like, we are at the absolute top of the market. Why would you buy these things? Now? We can foresee a crash on our horizon. You know, Stein's law of finances. Anything that can't go on forever eventually stops. We know that this bubble is going to burst. So just wait and buy it for 10 cents on the dollar.
- [63:12]
D
Right.
- [63:12]
A
This is. This is. It's. It's crazy to be putting all your energy into it now at the top of the market. Other questions? Yeah, in the back there with the flat cap.
- [63:25]
D
Yep. So it's sort of following on what you just said. If we know the bubble's gonna crash,
- [63:31]
A
one, why should we put our energies
- [63:34]
D
into that instead of what you convinced
- [63:37]
A
me of earlier on? Yeah. So why try and hasten the end of the bubble rather than just focusing on insidification and so on? Well, the problem is that bubbles are very destructive. Right. And the longer they run, the more destructive they become. So when I wrote the book, we were in the middle of a $700 billion bubble. Now it's a $1.4 trillion bubble. And you know what's worse than a $1.4 trillion bubble? A $2.8 trillion bubble.
- [64:00]
D
Bubble.
- [64:01]
A
Right. That we are going to destroy the net worth of lots and lots of ordinary savers as a result of allowing this bubble to continue. And when that happens, we will be at a crossroads. If we're smart, we won't do austerity, we'll do stimulus. But if we do what we've done for the last 25 years. We'll do austerity. And when we do austerity, it drives people into the arms of fascism. It is politically destabilizing to have a bubble and have it burst. It makes people not trust their institutions, makes them not trust each other. And it, it wrecks your politics generationally to have these bubbles. It is corrosive to the very idea of having a common interest when bubbles burst. So, yeah, I do think that this is a very urgent project to burst the bubble sooner rather than later. Yes, ma'.
- [64:49]
B
Am.
- [64:51]
D
Actually, so as a former content moderator, one of the big checks, the
- [64:57]
C
we
- [64:57]
D
should do profit versus we should do our things that you want really, really resonated. I know that you said it can't honestly be effective.
- [65:05]
C
Is there anything that can?
- [65:06]
D
If someone is on the side of. Right, things can be done right.
- [65:15]
A
So how, if you do work within a firm, how do you make it? Can you make it better? How do you make it better? What's the path to it? Well, there was a long time when, when workers could, right? Tech workers used to, used to be in very high demand before the half million layoffs, right? They're extraordinarily valuable to these firms, right? When a Silicon Valley giant adds an engineer, its bottom line goes up on average by about a million dollars a year. That's why you got the free massages, right? It wasn't like, because they liked you. It was because there were 10 bosses at the factory gates who give you a job, you know, by afternoon, by the afternoon, if you walked off the job during the morning, scrum. And so you know, if you cared about, about your users, right? If you, I call it Tron pills, right? You, you, you, I, I fight for the user, right? If you cared about your, your users because like you, like me, had this voyage into tech where you, you thought you knew what your life could be and the people around you told you what your life could be. But then you found networks and those networks expanded. Who you could be, who you could know, the things that you knew. The way you described what you knew, the world became better. Maybe you became materially better off, you got a better job, but also you became just like much more fulfilled and you want everyone to have that. This is a common phenomenon among tech workers, right? Ask a Linux user about the computer and then try and get them to stop, right? There were lots of people who felt that way. And when their boss said, now it's time to initiatify the thing, you missed your mother's Funeral to build. You could tell your boss to fuck off because there were 2, 10 bosses at the factory gates who wanted to give you a job. 500,000 layoffs later, there's 10 other workers at the factory gates will take your job. And so by failing to consolidate the power of scarcity with the power of solidarity, by mistaking yourself not for a worker, but for a temporarily embarrassed entrepreneur, right, we squandered this moment. So it's hard, right? It's much harder to build worker power when you don't already have it. It's much easier to consolidate worker power than to build it. But we've done it before, right? All worker power started with workers who were powerless. Creating power through solidarity, through bonds of solidarity. It was done memory, right? This was not like the lost art of a fallen civilization. We're not not having to like figure out how to embalm a pharaoh or build a pyramid without power tools, right? We're just doing the things we've done before. But again, in eastern Canada, there's a joke whose punchline is, if you wanted to get there, I wouldn't start. Start from here, right? This is not where we. This is not ideal. But the good news is that every other worker is on your side. America has more workers who want to be in unions than at any time in living memory generations. America has fewer unionized workers than at any time in generations. The American public supports unions more than it has in generations. Support for unions is an all time high. And the unions have the largest cash reserves they've ever had. But their bosses won't spend it on recruiting drives because they think that that money is for their own members, not for members of a union that doesn't exist yet. So we need to occupy our unions too, which, you know, we started to do. This is what Sean Fain did when he took over the uaw. And we see it everywhere. There's a new wave of union militancy sweeping through the unions and also through sweeping through unorganized workers. Where do you get your power in the, in the, in the shop? Well, maybe it's through your eloquence, right? Maybe you can be so eloquent that you convince people not to do the right, not to do the wrong thing. But I just don't think that's durable. I think that you actually need to have policies. Like when you look at librarians who fought back book bans, they've done it a little bit through arguing at town hall meetings and so on. What they've actually done is They've done it by. By arguing that they have policies about how to do book removals and making those policies stick. And they got those policies by having organization and having sort of a wider superstructure. It's not about the individual merit of people. It's about the system you build collectively and the professionalism and the professional ethics that you hew to. This is what we teach in library school. I'm a visiting professor of library science at the University of North Carolina. This is what we teach in library school.
- [69:20]
C
Right?
- [69:20]
A
Is these professional ethics. And so maybe you can pull it off. You should try.
- [69:26]
B
Right.
- [69:26]
A
But like the way that you make it work durably and not just win one victory here and lose five over there is by building a superstructure or by building an institution. And you have to do it with other people and other people. Hell, as other people. But, you know, they're also the best hope we have.
- [69:42]
C
Do you ever sleep?
- [69:47]
A
Sometimes. Maybe over there?
- [69:51]
B
Yeah, only very quick. We're just. We're right over time, so. Very quick question.
- [69:57]
A
Oh, well, if we're over time, can I do this? Can I take three really quick questions and answer them in five minutes? And I will get out my notebook and my pen. I've got my timer, so hang on. By all means. This is a party trick. I do. So go ahead. Okay, question
- [70:19]
D
code is an awesome speed. I have written it to my million bytes of python code in two weeks for a thousand dollars. After the GPUs become cheap, after LLMs become democratized, how do we let everyone do behind this?
- [70:33]
A
Yeah. What do we do to democratize access to coding assistance so everyone can write code and skills? Yeah. That's one, two, and then three. Go ahead. Yeah, you.
- [70:43]
D
How do we get the people who are anti AI of all costs on solidarity of this middle ground?
- [70:51]
A
Yeah. What do we do about people who are anti AI? And then the next one.
- [70:53]
D
Imagine I'm teaching at a European business school. People want to get a master's in social business. Can you name two companies they should start when they go home to their country instead of what they would start if they were here.
- [71:06]
A
What country should you start if you're in Europe? Okay. Yeah, yeah. So how do we democratize the skills? Well, you're getting at something very important, which is that the people who use this in a way that is productive for other people and not just for themselves have to have the skills to assess the stuff that comes out. And it's hard to develop those skills if you don't have to grapple with the kinds of problems that you get when you don't have the assistance. Right. You need some of it. And I don't know how you get there, but I do think that Claude code rhymes with some of the software I was really excited about. In the 90s or the aughts, really? Well, no, the 90s and the aughts like HyperCard and Visual Basic. And I was interested in these not because they would produce production code for other people, but because if you wanted to do something for you, you could just directly do it. You know, there's this motto of the disability movement. Nothing about us without us. Do you know what's even better than being consulted on your requirements? Just fixing, solving your problems, right. You having the. Being able to produce the solutions that don't have to go through this interpretive layer where there's always going to be a loss of fidelity as to what you need. And then the question is, what do we do about anti AI people? I mean, I'm having this conversation with a lot of friends of mine who identify themselves as anti AI. I think it's not as superficial as saying, oh, you're against the use of the technology, not the technology. I think you need to bring to them a much a richer analysis about why the contingent implementation of the technology as part of a finance bubble drives it towards this. You have to show that you understand that you're not just being pat or facile when you say, oh, it's, you know, it's a dual use technology. Guns don't kill people, People kill people, whatever, right? You have to actually show that you really mean it. You're not just trying to, you're not just trying to hand wave away their objections. And then the last question. What businesses should. What two firms should Europeans start if they go to a European social enterprise business school? Look, I don't think it even has to be a social enterprise. I think that if you're constrained by the gdpr, the privacy rules there, and the labor rules and consumer protection rules, that you should just go into business. Like making dongles that hang in the checkout aisle of the Carrefour and the little that jailbreak your iPhone and install an app store that doesn't take 30 cents out of the dollar, right? Like it's Apple's single largest line of business. It's $100 billion a year in pure profit. It is a 30% tax on the digital economy. If you did it like overnight, it'd be as though every newspaper whose subscribers pay via app got 25% more subscribers. It's like it would be transformative. And the thing that would constrain you is not ethics, although ideally ethics would play a part in it. The thing that would constrain you is that if you got caught stealing people's data or money or screwing them over on their working conditions, that they would fine you so much money that you'd lose more than you could possibly stand to gain. And then we would enforce antitrust laws so you couldn't buy all of your competitors or do predatory pricing, and then you couldn't become too big to fail and they're thus too big to jail. And I think that's actually, that's the answer. I don't think it has to be a social enterprise. I think that the reason that greed is good plays is because it works. And I think if greed didn't work work. It's not like we invented greedy people in the era of Margaret Thatcher. We just made it so that greedy people could make more money than good people in the era of Margaret Thatcher. And then they took over the world. Okay, so I'm an Internet user, which means I get to try all these cool gadgets. I got this. It's a device for making books non returnable. I'm giving demos of it at the time table over there. And I do have a child in the uk, so I. I am paying her tuition. So thank you.
- [74:49]
B
Okay folks, that's a wrap.
- [74:55]
A
That was the Cory Doctorow podcast. Licensed Creative Commons Attribution Non commercial share alike 4.0. Or as Woody Guthrie put it in another context, this song is copyrighted in the US under seal of copyright 1540 for a period of 28 years. And anyone caught singing it without our permission will be a mighty good friend of our because we don't give a dern. Publish it, write it, sing it, swing to it, yodel it. We wrote it and that's all we wanted to do. Many thanks to John Taylor Williams of Ryneck Studio. That's W R Y N E C K for engineering and mastering. John Taylor Williams is a Broadway broadcast technology specialist, an audio engineer, and a musician. In his spare time, he likes to carve useful objects out of wood, antler and steel.