People are worried that AI will automate their jobs. Yet Brian Merchant has interviewed hundreds of workers disrupted by AI, and what he sees isn’t a technological revolution but a centuries-old power struggle: bosses wielding new tech as leverage to deskill and destabilize workers. 

Merchant is a tech journalist and critic whose book, Blood in the Machine, gives a sympathetic history of the original Luddites. He joins Jon Bateman on The World Unpacked to explain what 19th century English craftsmen can tell us about AI, why he doesn’t lose sleep over Chinese models, and how to have a better industrial revolution. 

Transcript

Note: this is an AI-generated transcript and may contain errors

Jon Bateman: Brian Merchant, welcome to The World Unpacked. Hello, pleased to be here. With a name like Brian Merchand, do you think it was inevitable that you would end up writing about capitalism for a living?

Brian Merchant: Yes, I do. It was my destiny. It was I'm really

Jon Bateman: You know, into this nominative determinism thesis, um, my name is Bateman. So here I am trying to bait people into engaging with the podcast. Kind of makes sense.

Brian Merchant: Right, well with a line of questioning like that, who could resist? You got him a hook, light, and sinker. You got, you know, from the from the get-go.

Jon Bateman: A very auspicious start to this conversation.

Brian Merchant: Well, so my name actually does, it's it's a little maybe more interesting than it is. I it's, it' a translation from the Armenian Chorbachian. So my great grandfather was indeed a soup seller of merchants. So when he translated the name to the to the English, that's that that's where we get the merchant.

Jon Bateman: It's one of the better names out there. Um, there's others too. You know, there was a woman named Elizabeth economy who writes about the Chinese economy. I mean, that one's like totally on the nose. So, um, you're not at that level yet. Okay. Jokes aside, um AI.

Brian Merchant: Debatable. That's debatable.

Jon Bateman: Some people think it's a joke. I think you take it pretty seriously in your writing. That's actually one of the reasons why I wanted to talk to you. There's a lot of critics out there of AI, of its role in the economy and society. I would say most people on the whole in America seem somewhat afraid and disturbed by AI if I just have to take a mood of the country. Critics seem to be divided between two camps. There's the camp that thinks the technology itself is ineffectual and useless and then there's the Camp that takes it at least seriously and says it's actually a powerful useful technology and that's sort of part of the problem actually that we'll need to be grappling with. I see you almost as maybe more toward the second camp. I don't know how you'd describe yourself.

Brian Merchant: Yeah, I, you know, I don't know how useful sort of trying to separate those strains of criticism into two separate camps or like or sort of dichotomizing them. I think most critics who say it's useless and bad are still taking it seriously, if only because everybody else is. Um, we are seeing the AI being used by managers by uh by C suite executives by Fortune 500 companies in you know corporate managerial contexts and so it's being adopted. It's being used. It is being used as an automation tool. So, I you know, in the long run, to what extent and at what scale um that will that will play out on the one hand like even Sam Altman has said that a lot of companies are doing what he calls AI washing where they're you know laying off thousands of workers and saying oh it's AI when maybe there's a number of reasons why a firm wanted to do that already throughout the history of automation going back 200 years the question hasn't really been is this technology powerful or how power is it powerful enough to replace somebody's job. It's What you know what can management do with it like can management, you know justify laying off X number of people can management you know get people to buy. Lower quality goods on mass for a lower price point and then put their competitors out of work is that a use of automation like it's not really a question of how powerful the technology is it's sort of like what does this facilitate what does a enable the second part of that is i think we just we have to ask to like. Paraphrase Jeff Goldblum in Jurassic Park. Well, does that that should it then like just because it can and because it can do it just as cheaper? Should it like should we you know? Yeah. And so I think that this maybe is a moment where whether or not AI can or cannot perform these tasks, we should be asking what we do and do not want machines to do in a human society.

Jon Bateman: Totally fair answer. I think maybe the reason why I started with this question of your view of the technology and its power is that a lot of things are downstream of that, right? Yeah, like If all we're seeing right now is AI washing and it's companies that primarily wanted to get rid of jobs for other reasons, cut costs, improve shareholder value, and AI is just a kind of useful excuse for that, but then the technology flops, probably that would kind of burn itself out like relatively quickly. Um, and then there's this kind of curious phenomenon where you mentioned, you know, Dario Amadei, Sam Altman actually have some more extreme worries that they're putting out about the impact of jobs than actually some people kind of on the left have, right? Maybe if you do expect it to kind of flame out or be more of a narrative than actual change in the material organization of the economy, if that makes sense.

Brian Merchant: Mm hmm. I would definitely not go so far as to say that it's, it's not useful. I don't think that there's a lot of controversy or debate over the fact that it can, that AI can be used to write code and can produce like some, some decent, you can do rote code production outputs still need to be checked. So like it's still like, you still have an ROI to deal with when you're like investing in, you know, AI, even as a purely sort of code production tool, like what is your, like how high, you to argue about sort of. Nailing the code base? What kind of vulnerabilities is it opening up? How much can you trust your, you know, remaining engineers to like go over the codebase, make sure there aren't any errors, or it is doing what you want it to do. So I do think that there's a value proposition that still needs to be considered there. But no, I think the big question now is to, you know, is going to be like how, how niche that the automation is going to be, how far it really can scale, how much the market is going to bear, tons of AI produced digital goods. You know, AI is also doing this weird thing that tech has traditionally not taught us to do for the last 30 or 40 years, which is to sort of accept lower standards, right? Like now, generative AI is teaching us that, well, 10% of the time it's going to wrong. You can get, you know, you can deal with that. That's okay. Like that's fine. And we still haven't been able to clear that hurdle. Maybe we'll get within four or three or two percent. But how is that, you know, is that acceptable? Is 96 percent accuracy acceptable for most tasks? Maybe, maybe for some, maybe not. Anyways, there's this big negotiation going on and how much, you know, accuracy, what we want out of automation, all of these questions that we still, you know, that are largely unanswered, whether or not and underlying it all. Like, is this ever going to be? Uh cost efficient enough like to with huge data sets and all the energy that it costs to run and train and block probably again the answer is probably there will be use cases but i think we're still all over the map about like what in terms of you know where we're actually going to land on how much you know 10 years down the line how much ai is going to be a sort of you know, your average corporate office. Like, I don't know, it could be a major key sort of foundational element, or it could something that's a little more peripheral. And I think, you know that is still, it's anywhere that scale is wide and there's a spectrum and we're gonna land somewhere on that and then we can sort of sort things out accordingly.

Jon Bateman: Yeah, so people in the AI industry, the boosters of this technology, they will paint a picture of a future in which the AI becomes so intelligent and sophisticated that it can automate highly complex tasks, can get rid of law firms, you know, Wall Street, you know, kind of professional white-collar intellectual work. I think for right now, many people's experience of AI is maybe more in this kind of 80-20 realm that you were talking about Brian, where. Um, it's goofy at times it's got characteristic mistakes, but it's fast, cheap, available 24 7. And so you could think about it as like, flooding the labor market with cheaper, in many cases, lower quality versions of human produced outputs. It actually makes me think about your book about the Luddites. I didn't think we'd go there so quickly. But you're one of the foremost modern chroniclers of this historical movement from two centuries ago in which, um, English textile workers were fighting to, um hold on to their livelihood in a way that was also a similar transition, would you say from like the highly skilled version of clothwork to maybe a more machine oriented factory driven textile industry in which there was I think lower quality in many cases, but it was much cheaper.

Brian Merchant: The biggest misconception about the Luddites is that they didn't understand technology, that they were backwards looking, that they did get it, that they saw this stuff and they said, I don't know, I think they're cavemen. And so they smash it with a hammer.

Jon Bateman: Which is how we use that term today, that's kind of what's been handed down as the lore.

Brian Merchant: That was very purposeful that there it became a derogatory epithet, right? Like it serves the state, it serves, the the industrialists to have anybody who would oppose their, you know, their factories and mass production, to be sort of written off as foolish. Anyways, we're getting ahead of ourselves. You know, I'm one guy I run a series called AI killed my job on my newsletter where I solicit stories from from workers. It's a little bit self selecting, but I do have talked to, you know, hundreds of workers at this point from different fields. And I chronicle them at AI killed by job. And AI and there are a lot of similarities to what happened 200 years ago with the Luddites who are sort of this sort of skilled group of craft workers, usually cloth workers who had sort of worked more or less the same way for hundreds of years, they worked at home. Using machinery. We're talking about framework knitters who are who are knitting stockings, mostly weavers who are weaving with either wool or with increasingly as the industrial revolution goes on cotton. And so they're using looms and then sort of the attendant machinery that would be used to sort of finish the the the cloth or to or to cut it to prepare it for sale. Um, and so as the industrial revolution moves on, it really wasn't about the technology becoming more powerful in most cases. What's happening at the time of the Luddites is sort of more crude automation. They're sort of versions of technologies that have been around. So somebody took what the the stocking frame and made what's called a wide frame, where it I mean, you could see why they're they're mad if you're a craftsman in this field making stockings. And you pride yourself on making a nice pair of stockings, that sort of is, you knit it all in one piece. What the wide frame did was allow you to sort of quickly make two pieces of cloth and then stitch it together crudely. And so it was clearly inferior and it would fall apart a lot faster because it has a seam that falls apart. And so, but it did enable for faster production, less skilled workers could. Could make this because they didn't have to, again, create the whole garment. But again, it wasn't like this insane advance, this insane technological advance. It was the use to which it was being put because the other element that's happening now as sort of machinery is becoming of more interest to industrialists and the early entrepreneurs. This term is just beginning to be in circulation. We're talking about the early eighteen hundreds. This is also the age of sort of the ideological uh sort of foundations of of sort of capitalism itself and Adam Smith and sort of sort of the uh the idea that it's you know it's virtuous to sort of produce uh to sort of divide labor and to and to make money and to pursue these aims which was kind of a novel formation so long story short from 1800 to 1810 uh industrialists realized that they can divide labor, build factories. Pump out cloth goods faster, cheaper, and they can start to, you know, if they have access to capital, they can build these factories and they sort of squeeze the local. Technicians and craftsmen in a local town. All of the weavers and the workers who worked at home or with one or two other people or in small shops. They could work with their families, set their own hours. They had nice lives. They modified their own machinery. They were hackers in a lot of ways, but they just didn't want to submit themselves to this new economic system where they would have to sit and punch the clock and work for 10 hours a day in a windowless room in a factory. So they were fighting that as much as they were fighting. More than they were fighting the machinery, right? And so.

Jon Bateman: So there was some technological evolution happening, but in your story, the real revolution was social organization, that there was the realization that people could be packed into factories and there was a philosophical underpinning behind this. And I think government policy as well, tons of resonance with the AI era that we're entering into. You tell the story in the book about a young man who's kind of coming of age, just having finished the second seven-year apprenticeship, preparing for his life in textile work when he now is entering this totally changed textile industry. All I could think about was the college graduates who are graduating into an AI-driven workplace to some extent and are booing their graduation speakers. It's almost a one-to-one correlation.

Brian Merchant: Yeah, it's really similar. It is, you know, I and when I was read when I first encountered the story of john booth, that's how I could I he wasn't a college student, but I couldn't help read him that way. Even then he was like a saddle makers apprentice who is is a sort of a key part of the Luddite lore, who yeah, does things the right way goes through the apprentice system. And I should note that the the laws it wasn't just again that the technology was changing. I think there was a lot of opportunism as there is now where there were laws on the books that said, if you want to join the cloth trade, you have to apprentice for seven years. Yeah, you have to produce garments that are of this quality. With this, you know, they had rules about like the stitching and they had Rules about, you know, They had all these laws that were on the books that, you know, like our tech companies today, the early industrial Let's just. Ignored them, right? They said like, well, this, you know, the times are changing. Like, this is the technology is new. They didn't say this, that's not a quote verbatim, but that was the attitude. And that's how they made their case to Parliament at the time, just saying like, you're like, no, it's just truly just kind of a move fast and break things ethos. And it just gives you an idea of how long, you know, this has been around and how technology often gives those who see an opportunity or a chance to sort of break down social norms and to override them for you know, often for personal gain, right? Like they say, like, a critic would say, Uber's major innovation wasn't like just tapping GPS to unite a someone who's looking for a ride to a taxi driver. It was, you know arbitrage, it was getting around the taxi code of that city and saying, oh, we're not a taxi. We don't have to, you We don't have to. Know, take all these extra steps and you don't have to pay for a medallion. You don't have to do all of this. We're just kind of, you know, it's technology, so it's new. So, you know we don't to think about sort of local labor laws in the same way that a traditional business would. And then usually it takes years and years, but, you know governments catch up and say, oh wait, like maybe you should be an employee, or maybe if you're working 40 hours a week and you're doing the same thing, Maybe you should, you know, maybe that, maybe. Maybe it was less about the technology and what the executives and the company did with the idea of the technology.

Jon Bateman: I want to come back to the Luddite story and kind of get your take on what should have been done back then and kind therefore what we should do now. But just to make sure that we're hitting on AI in the current moment, you're pretty unique in the amount of time that you have spent talking to and interviewing and publishing stories from the field, from workers who are in what economists often call kind of AI exposed jobs or AI exposed industries. Translators, tech workers, educators, others. What's happening out there, right? I mean, are you seeing people lose their jobs? Are you seeing the nature of work changing? It seems to me that part of what you've been chronicling is maybe power dynamics within the workplace changing, but what just what have been your takeaways from all of these conversations?

Brian Merchant: Yeah. Again, it's a pretty broad spectrum. You know, creative work is not the largest sort of, you know, it doesn't constitute sort of like the largest by volume number of workers in the American economy, but creative work, is uniquely sort of as the economists might put it, we would rather. You know, use mid journey or, you know, chat GPT or whatever, whatever to produce text or images instead of paying a paying paying a per a human. And so it has had a fairly immediate effect on a number of fields like illustrators translators and and copywriters who they've, you I've again sort of talked to those who've been impacted. Most negatively, but there have been a lot of them. And studies do seem to reflect that. I did a piece for Wired a couple of years ago when the AI boom was even just getting off the ground. And it was pretty immediate that calls came down and some of the more aggressive companies that were like, well, we don't need half of our concept artists anymore. We don't as many. Sort of people sketching out you know designs for those extras that people like to buy so that it did kind of swing like a wrecking ball through through through the field there. I've you know, illustrators, a lot of them, you know their public facing work is often with magazines or book covers or whatever, but that's often not enough to do you know to pay rent. So many, many, if not most also supplement their comes. By doing work for ad agencies or marketing companies or companies that need like sort of internal art for for presentations and things like that internal magazines. Yeah, well, if it's internal now, don't really and you know, you could publish AI generated stuff and it maybe looks 60% is good or 70% is as good and you just kind of need something there and you're not going to catch flat for it because it's not public facing. So that's where all of the illustrators and sort of copywriters report seeing the biggest declines in their work. So yeah, it's staggered. It's all over the place. So it's

Jon Bateman: Yeah, and this goes back to the our discussion earlier of the kind of 80 20 aspect of AI that like it may be less than what a skilled professional can do with great effort, but it's cheap and fast. And then I think maybe what we're discovering is a lot of the economy actually is has that 80 20 aspects that maybe people would prefer the cheap or free 80 20 version. And then we'd still have kind of the premium sort of service, but there would be less of it. And then just other things that I hear from you is, you know, gig workers or freelancers, independent contractors, people who already have kind of a marginal attachment to their employer, their most vulnerable. One of the things I thought was really interesting from some of your tales is people whose job whose bosses wanted to use them to use AI. They themselves really didn't get it or felt like it wasn't a clear value add. But there was just the sort of like top down directive. Um, that maybe really didn't have a clear business logic from the perspective of the person doing the work day to day. Yeah. But from this larger kind of corporate strategy perspective, there was a sort of a ham fisted belief that like, we need more AI.

Brian Merchant: I mean, I think that was pretty prevalent, and I think that we're seeing that fever break maybe like right now, a lot of sort of the road, the you know, the reporting on sort of recent, you know? I think corporate earnings and stuff like that. A lot of it has been like, well, our AI spend has gotten way out of hand. Because for the last two or three years, yeah, I think there was huge FOMO among, you know, CFOs and managers who are like AI is the thing. You got to learn how to do it. You got to do you guys, especially in the tech work, there's a lot of frustration with having that top down. Um, sort of imposition of AI into the workplace. Key part of that before we move on past it is that it like really erodes the satisfaction that people get from doing their jobs. If they have to, you know, if they, if, they're just sort of dealing with AI content, if there, if your job goes from writing code that you feel proud of creating something to, You know, supervising, you know the output of a, of gen AI, a lot of, I've, I hear from a lot of folks that, you other. Levels of interest in their work and their satisfaction goes down accordingly and that's again not just coding but you know you name it a lot of translation work has sort of been seated to sort of checking AI output as well and of course what then does that give management opportunity to do, but reduced wage is just saying like, well, know, it doesn't require as skilled of a worker to sort of edit the output. We don't really need a skilled translator to do this. We can just have somebody we can just so so yeah so so it's and in that way it's kind of classic de-skilling right so you have AI being used as justification to sort um you know erode wages we see work satisfaction declining we see some job and task replacement outright. Um, and then you see sort of. The looming over it all of kind of is the use of AI by management as leverage, right? It exists kind of it's omnipresent, right. I'm management can always say, you know, like, oh, well, you know, we work either use it as a as a means of trying to get productivity to speed up or as a mean of sort of chilling maybe efforts to Um, organize or ask for raises even, or just sort of, it's, you know, like we have the AI waiting at the gate. So, you know, don't, you don't get any big ideas.

Jon Bateman: We've been talking about this as a kind of internal battle within companies, as sort of workers versus bosses, workers versus owners or capitalists. Could you also reframe this as kind of workers vs. Consumers? That part of what's happening is consumers out there see value in AI produced outputs. Translation would be an interesting example of this, right? I mean, I've worked with professional interpreters before. They're very skilled, impressive people. They often understand cultural subs text. They're also expensive, hard to schedule. They actually kind of tire quickly. So sometimes you have to have two to swap out the others. Doubles the amount of time that you have to spend with someone because, you know, so there's a lot of complications with it. And I'll just say if someone eventually creates like a really, really impressive live translation device, which people are kind of they're marketing right now and I haven't tried them. But like if that existed to me as a consumer, that would be very exciting. So there is like there's this other side of it, too, which is the consumer. How do we balance that against what a worker might be experiencing?

Brian Merchant: Yeah, I mean, that's a million dollar question. I mean I'm in France right now and I use Google Translate to sort of enhance my understanding of what's going on. I was in Japan not too long ago and they have those like ordering systems where it's like a kind of a vending machine where you like you could order your ramen or your sushi through like a but it's all and I you know you can use computer vision to sort of translate in real time you know what those are so you could order the ramen that you want and that's great

Jon Bateman: It's a great example, because Japan is somewhat of an interesting, it's a difficult place to navigate for an English speaker who doesn't speak Japanese. Yeah, there's not a lot of English spoken, you know, or written in the country. And so yeah, that's an interesting example.

Brian Merchant: Yeah, so here's some genuinely useful, you know, to where you could even consider that as like a new a new market. Like there was never really a market for you know individual translators or interpreters for the lay traveler who's just going for a week's vacation on Japan. That's you could you could kind of look at that as a novel market. And in cases where AI is opening up a novel Market like that, that I think that that's great. I think it's when it becomes intention. With a with an existing market or a marketplace that you start to have the the the worker strife and it's I mean you do that that is sort of like the age old question because ultimately right which if you run a company that you know is selling translation services and now you have this new tool that's going to reduce your costs but at the expense of your workers Um, you know, there's a pretty clear incentive to do that. And so how do you keep one with, without the other? Um, and again, like these are questions that we often have to navigate in, like sort of not purely economic terms, because like, like, as you mentioned, like there's great importance in a, in a good translator or, or an interpreter that can like retain the context and the human element and like you know, before too long, they're sort of correct some of the drift from a the inevitable sort of, you know mistakes and AI generated output that will come along. So like there has to be there has to be some some ad mixture of both. And you know I don't I you know i don't have I don t have the answer there. I do think what I will say is that I do think the market is rather distorted right now. And the way that tech products in general, especially software has been developed and and sort of deployed over the last 15 years, at least, is, is one that, you know, it's, can you even call like this like a true like free market capitalism when you have venture capital larding, you're backing them up, like they these Companies don't have to make. Profit anytime soon. Uber famously didn't have to turn a profit for 10 years because it had such a war chest of venture capital and what it uses that capital for in many cases, especially in Uber's case was to penetrate markets, squeeze out competition. Once they squeeze them out, start raising prices and then on its terms again, we mentioned earlier what a lot of the sort of the industrialists were trying to do was to sort of. Um you know disrupt uh uh a sort of structure of of labor norms and and standards and you know laws in some cases on the books and that's what Uber was trying to do too and they had billions and billions of dollars to do it. AI similar right like they can they can and and so when you're looking at these questions of like what can it automate and that's why we don't know the answer yet is because these companies aren't operating as traditional businesses. They have unfathomable amounts of money at their disposal to invest year in, year out. I mean, we'll see how long before that comes due, maybe SoftBank and whatever is kind of the final boss of the BC system. But so far, that's why I think the question remains so murky. Is it going to be economical for a lot of these firms to continue to use AI at scale, even that they've been doing now, where the drive again is to automate labor to replace a lot of labor.

Jon Bateman: You know, Uber is an interesting example. A lot of consumers prefer Uber over taxis, so you could still kind of view that as like a worker versus consumer issue. But stepping back a little bit, you know, I am thinking about, I can imagine a world in which Uber was never allowed to exist because if regulators said no, like we want to maintain this like regulated, like medallion system, we want to maintain taxi cab drivers as a skilled trade. I sort of feel like that, I could imagine that timeline. It's harder for me to imagine the timeline where textiles were still made in the home through these kind of cottage industries that you describe in the book versus factories. There are sort of types of technological and social organization that are so powerful, so productive, so good at fitting what ultimately consumers and society seems to want that it's going to be hard to hold them back. And maybe we can kind of. Soften the blow in modest ways versus others where the old way of doing things, the kind of more comfortable, cozy social contracts is kind of quasi sustainable with effort. And is that is that a fair way of looking at it? And then kind of where do we put AI in that?

Brian Merchant: Yeah, so I think that there is the the answer is is that like it there is absolutely a middle ground. There is there is a there's a third road where you know and and inevitably that road is just is taken only by giving workers like true empowerment and say over their conditions. And under the conditions into which the technology is being integrated. When they have no power, when they have no say, you get a situation like the Industrial Revolution where. Those early days of the Industrial Revolution where industrialists hit the gas as hard as they could. They immiserated workers to the bone. They stripped away. They got Parliament ultimately to sort of conclude the story we began earlier with at the beginning of the 1800s when cloth workers start going to Parliament saying, you got to at least make them follow the rules, follow the laws and here like here's what we we just want a you know, minimum wage and we want to make sure we want to laws in place that'll prevent them from paying workers in truck or goods rather than actual money because that's common at the time for for the industrialists to do. So they had these things that they wanted to have in Parliament said get lost, right? They said you get nothing. Not only that we're going to tear up all the all the regulations that are on the books. And then we're going to ignore and laugh out of time all of your proposals that you have.

Jon Bateman: OK, so the middle way back then, it wasn't like, let's not do factories. It was like we can go to factories, but we can impose, you know, maybe protections around like child labor and sort of how you're paid, things like that.

Brian Merchant: Yeah. Or, you know, I mean, it would have, it seemed somewhat, you know, novel at the time, but there were a bunch of rather progressive proposals that were written down. You can go read them. They have, there were, there are proposals basically for VATs for value added tax. There are proposals for like, what if you tax, tax like the extra length of yard? I mean, what if he tax the extra length of, of cloth that was produced by the, by sort of the machine. Uh versus was produced by by a skilled trader and then use a a portion of that to fund, you know, programs to to to support workers. Those were like on the books back then. Um or you know you can imagine yeah a a a middle ground where the where there's there were already it's not like there were no sort of factory-esque buildings. They just had completely different working conditions. There were workshops, right? Like if if machinery were integrated into these workshops under worker control then we could maybe we don't get the same level of mass production but i think you could get production at a level that would have satisfied because the problem is is that we start getting into a bunch of other you know rapidly multiplying questions as well because they're they're big markets at the time are export markets right including the united states and the colonies to which uh you know England is is is sort of fighting to protect or fighting to entrench at the time with France and so the Napoleonic Wars are are a big issue and in fact one of the big reasons you get the light uprisings is because England imposes one of first ever trade embargoes and prevents them from selling or exporting their goods to um the United States. Which if I'm

Jon Bateman: If I could say, that actually does have a lot of parallels to AI as well that on two fronts. One is we're now in a fully globalized economy. And so if we restrain and regulate the deployment or the conditions of use of AI in our country, that the argument is that others will deploy it in an unrestrained manner somewhere else outcompete us and it'll be a lose-lose. And then, I guess... The new version of the Napoleonic Wars is geopolitical conflict in which AI is understood not just as an economic tool, but a military tool and intelligence tool, a strategic asset. And so there's fear that global competitors like China, again, if they're able to kind of win this race, that could be to our national disadvantage. I wonder what you make of those arguments that seemed... I guess to carry the day in the time of the Luddites, should we listen to those arguments today? Is there any credibility to them? I understand sometimes people making these arguments actually have self-serving motivations, but nevertheless, they kind of need to be still reckoned with too, right? Like how would we deal with those now?

Brian Merchant: Yeah, so, you know, it, it's a little the comparison falls apart a little bit in that there wasn't another sort of industrializing nation or nation state that was all that they're like, we've got to be the first to make, you know, the power looms and get the factories set up. Rather, the government itself imposed this ban as like a punitive measure, right to the United States for siding with France in the war. Um so I think I mean personally I see the the sort of the looming specter of of of sort of China getting AI I I I see that as 90% plus specious as a way of for the AI industry to to sort of um you know can pursue deregulation and or or in some cases, perhaps with like, sort of anthropics approach more like sort of where the you know, safety friendly, you know AI approach and as a you know maybe before this latest dust up to with the with the administration as a means of sort of winning contracts, right? Like, I think you know

Jon Bateman: Can I can I push you on the 10% that's left over though, because I'll fully grant that the AI companies when they're waving the flag and pointing to China, of course, there's all sorts of political and economic motivations for them to do so. But setting aside the companies and putting yourself in a position of a voter or policymaker who then has to take seriously like, okay, are, are we worried about this is someone who's kind of like takes like a left sort of worker oriented take on AI. Is that, does that person still have anything to fear from China racing ahead and developing better AI than us? Is that?

Brian Merchant: I mean, I could I could tell you like, nothing causes me to lose less sleep. I mean what like what what are we worried about? Honestly, like, I mean they have a they have a nasty surveillance state already. Like they, you know, what are we great stewards of AI, like all the reports that I've read say that Claude may or may not have been involved in a in organizing and facilitating the accidental bombing of an Iranian girls school in which, you know, dozens of girls were killed. I mean, Claude, we don't know the extent to which it was, but like, here's how we're using AI already. It was Claude was involved in the campaign to, you know, abduct the Venezuelan president that came out. It was involved in the military campaigning in Iran, this unprovoked war in which we're just sort of like, you know bombing another nation. Like, I mean what- I just I guess I don't see like what what are we so what are we worried that China is going to do with this AI? Yeah, that's any worse than

Jon Bateman: Well, it gets back to how we started the conversation of like, does it matter if AI is useful and powerful? Right? I mean, we're we're in this kind of infantile stage, where we're experimenting with this technology, there's a certain kludge or crud factor to it. There's also things that it's doing that are remarkable. And there's there's belief that many have that it could improve. And so if that's true, one would expect AI to be productivity enhancing. For kind of strategic industrial competition, and then also enhancing the lethality or the efficacy of a military if we ever had to go to war with China. So I think that's the argument, that actually this technology is useful and it's going somewhere. And then therefore it could matter if an adversary becomes dominant in it.

Brian Merchant: Well, I will say that it already matters, but again, for reasons that are a little askance to, I think what most people are worried about, I think, you know, Kevin Baker had a great piece in The Guardian and on his newsletter, you know about ways that you know that that Claude may actually have been used or you know when we think about AI and sort of being used by a bureaucracy a military bureaucracy it's it's again it's not that it's becoming super powerful but it had but it's enabling all of these different sort of decision making processes to sort of either be accelerated when they shouldn't be or sort of taking away accountability when there really should be accountability in a kill chain right and similarly there's been some you know really interesting reporting done on the way that the IDF for instance has used AI and its project lavender and they don't need the AI to be great for it to have a devastating outcome in fact there's a lot of evidence that the AI has been used as a target selection tool and it's not that like Oh my god, it's so powerful that it's finding the right person all the time. But no, it s it's serving as this rubber stamping mechanism where it's it's just running through all these it's, it serving up even maybe more targets

Jon Bateman: Yeah, totally, totally grant that I mean, listen, you know, supposedly AI allowed us to prosecute like 1000 times more targets than we could have before in Iran. The war in Iran has been a strategic disaster. Maybe you can argue that right that the temptation for many military technologies have kind of tempted leaders to believe they could accomplish more than they can you could even kind of argue actually that it's a strategic disadvantage to have some of these technologies in terms of how it changes leader technology, leader psychology, let me let me just try one more one more.

Brian Merchant: I do. Just one quick point is it also can lead to more death because it can lead what I think is the Dan Davies calls an accountability sink, right? If you're pressing the button, then you don't feel as connected to making that decision as you would have before. So there's there are cases when it can Lead to more, more and sort of gratuitous violence by by sort of the weapons of war. Because these systems are sort of letting off letting you know, people in the kill chain off the hook by pressing them. So I think that's a real concern too. And again, I'm I lose a lot more sleep over over that the way that it's sort of facilitating violence right now, rather than sort of, you know speculative, you know, what will happen if China also, yeah, you know, get a big automation tool.

Jon Bateman: And in fairness to you, Brian, your focus is on labor and the workplace. And the only reason why I'm bringing up all this stuff is, it just seems like it has to be it has to be put together into a larger sense of where this technology is headed. But you may you painted this picture earlier, but there's a third way. And as I understand this third way, it's like, just like the Luddites weren't against technology, we don't need to be against technology. If we don t want to, we d have to be against AI, I guess that the the way there would be we can innovate, we can use AI, but then what you would, I think, call our attention to is under what conditions is it deployed? Like who owns the factories, so to speak, right? Sort of who controls the technology and who benefits from it. Maybe... Maybe one way to ask the question would be, do you have a kind of greatest hits from the history that you've looked at as to like times with it when a technology arose that we kind of did this or sort of came close or like what can we be inspired by to not just feel like history has been this kind of grim march of like technological exploitation but that actually this third way is viable. We've done it at times. We can do it again.

Brian Merchant: Yeah, I mean, it's a great question. And unfortunately, industrial history kind of goes one way. And, you know, it hasn't I would say things have been getting progressively better since the first industrial revolution. Again, it was very stark. It was very, you know, that was completely the definition of completely unregulated for many decades. And it was an unmitigated disaster there have. And I think you could one way of looking at sort of the history of industrial automation since then. Is, you know, workers trying to sort of build on what was learned through that, you know, I, you're organizing officially was a great leap, winning the right to sort of, you know, organize labor unions and collectively bargain was a huge leap. And that immeasurably improved millions of lives at that point. And also, fun fact, that drive to sort of that would collective bargaining was banned at the time of the Luddites. But one thing that came out of the Luddite rebellion, as argued at least by a lot of pretty smart scholars and historians was that class consciousness was developed at the first time and a lot Of the reform efforts gained steam out of sort of those those Luddit days, a lot of the reform pushes sort of really took root took shape. And some of the luddites themselves would go on to sort of helping to overturn the bans on collective bargaining. And so you get unionizing for the first time. Out of that push. And then so, yeah, I mean, you could look at then the rest of the industrial revolution is an incredible struggle, but you finally do get some, you know, some constraints, you do get, you know, child labor kicked away, you get working, you know, hours limited when when they went before.

Jon Bateman: I want to ask you about unions and all of that. I almost want to asked you the bigger question of, what do you think about the Industrial Revolution? Are you happy we did it? Because in one way, it sounds like your take is, overall, we have failed over the last 200, 300 years to consistently pursue this middle way. And so if that's right, we could then look at the Industrial revolution and say like, was this good like are we happy that we have these technologies with the exploitation that you would argue that came with it? Or would it would other parts of it that you would like go back and undo it? Does that make sense?

Brian Merchant: Yeah, I mean, again, it's really hard to answer that question because history did happen this way. I do think there are always alternatives available to us. I do you think that there are other ways that things could have gone. I'll direct listeners to go read if they're interested. There was a volume, I think, put out by the MIT press of sort of, sort of like... Alternate, like speculative economic histories. And one of them in there is like, what if the lead? What if the lights have won? Like, right, right? Right? It's not that like, yeah, technology would have been shut down, and it would have been stopped, but it would have taken a radically divergent sort of social organization, as you say, and so it's by Miriam Cherry. And it's a great provocation to where you say you could see technological development, continuing a pace, but sort of you would have maybe more worker collectives, you'd have maybe, you know, like that I alluded to earlier, maybe you'd have more sort of, you know co ops or factories with, you know, with with with less of a sort of a gap of between the the owner and the worker on the floor. And you can imagine all kinds of different cooperatives and stuff. I think those I think the key when what I always say to kind of dumb it down in a podcast quickly is that it just comes to it comes down to the level of democracy that the worker has over the technology, and if they have true democratic input into the conditions again, into how this technology is unveiling, then things are going to turn out. So we do have select cases, and you have some interesting sort of experiments here and there. By and large, no, we have not gotten this right yet. Through our 200 years of history, we've had periods where things have been pretty good, right? You go back 60, 70 years to sort of the auto factories where people could afford to buy houses and had good pensions. And that was a more even distribution of the technology that automated the assembly line and the production of automobiles and distributed some of those gains to the workers. So. Yeah, I mean, right, right now, I think we are being asked that question again, if you're not by the AI companies themselves sometimes. And I you know, I would also suggest that like, now is a very good time to try to answer it.

Jon Bateman: I think that's right, actually. I mean, people are using this term the next industrial revolution. And of course, the part of the segments of the anti-AI crowd will kind of roll their eyes at that, right? And say, like, this is just the ultimate marketing mantra. But I take it seriously as like a plausible outcome. And so then it does it is kind of relevant. Like, what do we think about the first industrial revolution? I was thinking about this because you brought up co-ops as an alternate organizing structure for the workplace. So. In my neighborhood, there is this fantastic grocery co-op. It's owned by the community members who shop there. Um, they, I think are unionized. They certainly pay a living wage and benefits to their employees. I've lived there for three years. I haven't seen any turnover at this grocery store. Everyone there is super friendly. It's cozy. I absolutely love it. It's one of my best places to shop. It's also incredibly expensive. And I told my wife the other day, I was like, we just can't go there as anymore because the the sort of giant foods or Safeway the kind of picture of like, austere, it's like gross in there, the workers aren't paid well, we've got these like fluorescent lights is totally dehumanized. And yet, I kind of I'm actually increasingly choosing that in part or at least like doing a mixture like a better mixture of the two, this can get back to the worker versus consumer side. So, so is if the The third way is saying, you know, we're going to be more like my neighborhood grocery store, like as an as a kind of economy and less like giant or Safeway. To me, that's actually not it's not clear which which one is better.

Brian Merchant: Yeah, I mean, I think you're, you know, corner store co-op is is kind of an extreme example. They're probably like going to the greatest lengths that they can to sort of, you know, serve a very niche market and get the most organic goods. And you know it's like the sort of the Erawan effect or something. That's you know that's kind of a kind of a product of just sort of you know like they're again they're serving that need and use also gesture towards the potentiality that you know we might see sort of like artisanal you know, human made goods, uh you know sort of come down the pie or there be a market specifically for people writing like human made, you know. Definitely. Yeah. Novels are yeah and I I would yeah I think it's clear that the premium is already being put in a lot of ways on on human made stuff but II think it'll probably be a little more expansive than your um uh than than your your neighborhood co-op but I think the short answer is that like I do you know it's I I it's it's ultimately going to be up to us, right? Like I don't think we would need to take such an extreme position as like, I don't everything needs to be or, you know, organically sourced to the extent that you're, you're persimmons are or whatever. But we but we can sort of negotiate these questions in a way that's again, not just adhering to this top down. Uh mentality that's sort of governed just about everything right now. So I, I, you know, I also think that we are also in this age where we are like a wash in cheap crap, like, and so to me, the prospect of having less cheap crap in some ways, it seems great. Like it would be like if we have some checks on on Matt. I mean, I think everybody would Would you love to see less slop? On the internet like I think we would if we could produce less uh less slop then everybody would vote for that the option um and we you know again like I don't I I should spend you know more time writing out a political manifesto about how we how we achieve this uh uh this middle ground here but I do think that that you know the in the short term um it comes not through any sort of like radical utopian manifesto or program but just like building up union density again getting. I mean, that's when you see actual working sort of functional condition. It's it's it's a slog. It can be but like real democracy often is so the best cases where you have like, you know, AI use being mediated between like an AI happy boss and workers so far that I've seen have been union contracts. You know, a good example is the the the pen guild has a has a union contract that you know says if you're gonna use AI that's gonna impact our work in a serious way, you gotta give us a heads up. We have to have we can deliberate about it and management has you know has has violated that. It's been taken to the NLRB. They've been sorting it out. Obviously, it's not ideal but it is a mechanism, right? And if we normalize more of those mechanisms again, I think that's the first step is just is just getting. Getting more workers to understand that there are means of actually addressing AI use in the workplace. There's just not as many unionized workplaces in the United States as there were 50 or 60 years ago, back when, again, we were reaping the gains from automation that was happening during the auto-production boom. So I do think that there models that we can draw from, I hope... We can improve upon that and, you know, create something even more durable than obviously union power has been corroded and eaten away over the last half century. But you know I think that's it. That's a place to start.

Jon Bateman: Well, I think the discussion of unionization is actually a fitting place for us to end this conversation too. We've, we've talked a lot about this term social organization and Brian, what I appreciate and admire about your work is that you continually direct our attention back to the choices that we're making as a society as to how and where to deploy technology, who controls it, who owns it, who benefits from it. That is a distinct set of questions from whether technology exists or is allowed to progress. Send. We're going to continue to face these. I do hope you write your political manifesto. In the meantime, people can check out Brian's outstanding book, Blood in the Machine, where he gives more of this Luddite history. And Brian, I think that's the name of your sub stack as well, which is one of the best places out there to follow these issues.

Brian Merchant: Wonderful, yes, thanks so much. It was great chatting, I really enjoyed this and let's do it again sometime.