Amoday, Altman, all of the AI CEOs, they do want to create the impression that like a great disruption is coming that makes it easier for them to sell more automation software, right? Like that's it's a it's a product in the array of products that they're selling is enterprise AI automation.
SPEAKER_01Now you might also remember Brian from a podcast we were doing together earlier this year called System Crash, or just from hearing him on the show before we were actually doing that show in the past number of years. But since Brian has been talking to so many workers about the impacts that generative AI and the rollout of those tools in various companies in many different sectors are having on their professions, on their work, on basically the sectors that they work in, I figured it was a good moment to have him back on the show so we can discuss, you know, not just the effects that generative AI is having on work, but also this broader narrative of AI that we've been seeing over the past number of years, these questions that many people are posing now about the state of the AI bubble and the AI hype that we have been experiencing and whether we're finally starting to see it deflate. You know, I think it's still very much an open question and we'll have to see where that goes. But also to try to understand, you know, what the longer-term consequences of generative AI might be. Even if this bubble does eventually burst, that doesn't mean the technology is going to disappear. There are still kind of remnants of the metaverse out there. Of course, cryptocurrency has turned into a political force. So what might happen with generative AI in the future? So I think there are a lot of interesting discussions that we have in the show. And of course, you know, you'll hear that we get on pretty well when we're talking together because we've known one another for years and we also hosted a show together for quite some time. So I have little doubt that you're going to enjoy this episode with me and Brian, where we dig into these issues that we've both been paying so much attention to, writing so much about over the past number of years, but also to look specifically at the work that he has been doing recently. So if you do enjoy this episode, make sure to share the show on social media or with any friends or colleagues who you think would learn from it. If you want to support the work that goes into making Tech Won't Save Us every single week, so I can keep having these critical, in-depth conversations that help you better understand the technologies that pervade our lives and the tech industry that pushes them on us. You can join supporters like Christopher from Stockholm, Lavinia in Geneva, Switzerland, and Steve from Indy by going to patreon.com/slash Tech Won't Save Us, where you can support the show as well. Thanks so much and enjoy this week's conversation. Brian, welcome back to Tech Won't Save Us.
SPEAKER_00I have to say, I it's just it's such an honor to be here. I'm just such a fan. I'm listening to the podcast for so long. I was just hoping one day that I would uh you know get this invite to join the show and spend some time with you, my favorite podcast host, Paris Marks.
SPEAKER_01I appreciate all this high praise coming from you, a person who I don't know very well, uh coming onto the show like this.
SPEAKER_00Right. Practically strangers who didn't spend the most of a year talking weekly for hours at a time.
SPEAKER_01Yeah, I was just talking to myself, as as we both know, right? Just with an AI filter on my voice.
SPEAKER_00That's right. Uh it has never been proven or disproven. Uh so whoever it was in the Tech One Savus community that put that theory forward, uh just say uh you haven't been proven wrong. Definitively. I might not exist. So oh, it's good to talk again, Paris. It's good to don't you miss this.
SPEAKER_01Oh my god. Absolutely. You know what? I do miss like for what uh nine months, eight months or something. We were like chatting every single week. And every now and then I'm like, man, haven't heard from Brian in a little while. Wonder how he's doing.
SPEAKER_00I know it was a nice way to it was a nice way to process all the all the shit that was going on. Unfortunately, it was also a lot of work producing these things. And it was my it was, you know, first time podcasting. So now I know what goes into the sausage a little bit.
SPEAKER_01More respect for the podcaster community.
SPEAKER_00Yeah, more respect. And I'm sure there's some overlap between you know Tech Won't Save Us and System Crash, or maybe like a 100% overlap. So to all those who've been asking us if we're gonna come back or say we missed the show, just say like we thank you so much for for the kind words and for reaching out. And uh, we still still don't know. It was honestly, it was just like we both have so much going on, and Paris is what you're halfway through the writing a book, or how far are you almost done?
SPEAKER_01Uh about a bit past a quarter, I guess. Yeah. Yeah. Slowly chipping away. The first quarter is the hardest quarter. Yeah.
SPEAKER_00And then the ball's rolling, and it's like we got momentum going downhill.
SPEAKER_01Totally. I'm like we were saying before you know we got on the pod. Like, I'm I'm feeling good about the momentum and and where things are going. Hopefully, people are gonna like the book.
SPEAKER_00That's no small thing. Anyone out there who's tried to write a book, it can be very daunting. There's just like you're just like wading through a swamp of words, and it's hard to corral them into any meaningful sort of shape or direction. And so I am not kidding when I say that first quarter is probably the hardest quarter.
SPEAKER_01Even just getting into it like entirely was just trying to get myself in the headspace to be able to start writing it was you know daunting in itself, right? And took me weeks.
SPEAKER_00Yeah, and now's when I should probably say that I'm ghostwriting the whole thing. So I, you know, I I it's really Paris on the podcast, but it's Ryan.
SPEAKER_01We weren't supposed to tell anybody that's but it's really me writing the books.
SPEAKER_00So I'm writing Chris's books, he's being me on the pike.
SPEAKER_01I'm gonna be I'm getting to pretend to be stressed out. Yeah, while you do. But no, it's like you're saying, it's great to have you back on the show. You know, now that we're we're not doing system crash, I'm sure you'll be making more regular appearances back on this show as we talk about what you're up to and you know these big issues that we're both talking about now that you know we're not talking about every single week.
SPEAKER_00Yeah, if my you know, if my schedule permits it, I'm pretty busy these days.
SPEAKER_01Right, sorry. I know you're in really high demand and you know, just a little old tech won't save us. Might not be able to catch someone like Brian Merchant too often, but we appreciate when we can get your time, you know.
SPEAKER_00Of course. No, always uh always always a pleasure to be here.
SPEAKER_01But you have been doing a ton of writing and reporting, and you know, we're doing it while we were doing system crash as well on AI and and the broader effects of these things. And you have this great series that you have been writing called AI Killed My Job. And so I wanted to talk to you more about that. But if we're talking about AI, I think the big thing that we have to start with is obviously all of this discussion that we've been having about an AI bubble for the past little while, right? I think it's pretty well established that these AI companies are overvalued, that they are making claims about their products that are not really supported by what the products are actually doing. And it feels like in the past few weeks, we have reached this point where it's kind of been like okay to acknowledge that there is an AI bubble and to question whether that bubble is finally going to burst in the near future. You know, we've had seen some difficulties in the stock market. Different companies have been pulling back on certain initiatives and things like that. Obviously, we saw Sam Altman come out and basically acknowledge that AI is in a bubble. So I wonder what your vibe is at the moment and how you're feeling about where this kind of AI market, where this AI bubble, where the AI hype is in this moment.
SPEAKER_00There is a lot going on, and I think there are a few major developments that have sort of changed the conversation, perhaps permanently. Number one is that when GPT-5 came out, which was this long-awaited, mega-hyped product from OpenAI that was supposed to be sort of like the next incarnation of almost AGI or artificial general intelligence. It was supposed to be this amazing transcendent moment.
SPEAKER_01Yeah, supposed to be this like massive leap that Sam Altman has been talking about for like months and months and months, right?
SPEAKER_00Months and months and months for two years almost, because you know, it was three when when uh Chat GPT first came out and kind of made its first splash in short order. They went to GPT-4, and then sort of everybody was kind of in the AI community was like, okay, well, GPT-5 is going to be the one because from three to four was a pretty noticeable, or the performance was much better of the models, and it just kind of like felt more like an actual artificial intelligence in terms of a product and interactivity and all that. And then five just uh you could kind of now in hindsight, it's pretty clear that or it was the the question constantly haunted them like, is this going to be enough? And it seemed like the answer was always no. So as they would iterate and like release new models, they started to get into you know the point fives and then like the letters and then 4.0 and then Orion or whatever, and it was uh very in incremental progress. And that sort of complicated matters we can see now, because if you release an incremental product update and then say this is you know the next coming of AGI, then people are bound to be disappointed, which is exactly what happened. So my sense is that Altman and his C-suite at OpenAI were kind of like just like, well, like we got to pull the trigger sometime. We can't, if we it's only gonna look worse if we wait another year or whatever. And it was like kind of like this safe distance from when they secured their last round of mega funding from Softbank. And it was just like, okay, maybe we can just release it now and maybe we can get away with it. And they couldn't, right? It was like users had already sort of baked in a host of assumptions, other users were quite sort of uh addicted already to the previous iteration of the product, and it was sort of just on the terms that open AI set for itself a failure. And that's how that's what I think is important because some people are saying, like, oh, like this is it's it's it's amazing, and the critics are you know being too harsh or whatever. But I'm I'm judging this by the terms that open AI set out for itself. And you can look back at Sam Altman's comments himself that he published on his blog just in February, where it's like, we're getting close to AGI, right? Like it's in the air, like it's gonna be very close. And then what happens after the launch, six months later, of GPT-5? Suddenly, AGI is not really a useful term anymore. It's not a super useful term to quote Sam Altman. I was like, Are you kidding me? Because you have been banging this drum, and maybe it's not super useful for you right now because you're going to be criticized about it, but it has been incredibly useful to you as a fundraising tool, as something to tout as you go to Microsoft or go to, you know, prospective enterprise clients and say, AGI is around the corner, we're building it, give us $10 billion, or you know, invest in this next round, or or you know, buy a suite of uh GPT for business or whatever. And so that disappointment, I think, finally solidified the fact that the level of improvement is not going to continue. I mean, we can debate the actual sort of benchmarks or how well that the model did in this context or what it's good at. But the bottom line is that, like, you know, critics, you know, folks like Gary Marcus, most notably, probably have been talking about how sort of that just scaling, which is just feeding more and more data into the systems, into the models, had hit a limit. And now it's pretty clear that he was right about that. And that means that whatever else happens, it doesn't mean that like, you know, AI isn't going to be able to do interesting things or different, but it means this model where you're just getting more and more and more data, getting the LLMs to train on more and more data and then to produce output based on just more and more and more and more. That ethos has sort of reached its limits. And there's gonna have to be new interjections of symbolic reasoning or different configurations. There's gonna have to be something else. And so that I think has permitted the business press, the tech press to sort of take stock of what's actually happened on the ground so far. And I think it's also worth noting that into this sort of environment came this study from MIT that showed that 95% of businesses that have adopted AI have essentially struggled to do so and had have not showed major gains. And so you have like, well, the business case is iffy, the model sort of improvement is iffy, has slowed down. And the future all of a sudden seems very uncertain because as listeners of this pod know, that like this is an incredibly capital-intensive technology where it's not just like oopsie, this didn't work, let's try something else. It's like you have already sort of baked in massive contracts with data centers, with cloud compute providers, chip purchases from Nvidia, where like it really, really matters because again, it was all predicated on scale. It was all predicated on scale. And so you have to take a hard look at the through lines. And so now we're at this sort of cloudy moment where it's like, oh, wait a minute. Meta, which was just like a month ago or even weeks ago, like paying a hundred million dollars, signing bonuses to get AI researchers away from open AI, is going like, actually, maybe we're gonna pause our uh super intelligence team that they're calling it. Yeah, we need to reorganize uh in this moment.
SPEAKER_01We need to reorganize. I think that what you're saying is so important to understand, right? Because the big assertion around generative AI for so long has been like it's on this exponential curve, like so many of these other like tech products, right? And so if you have GPT-5 come out and it's not showing that, then all of a sudden, like the whole thing that this whole boom, that this whole market, that this whole like supposed business and and business venture is built on is being called into question because these products are not actually getting so much more powerful on this exponential curve. It's like, okay, you had this moment where it came up, but now it looks like we're on the S-curve that we've seen with you know AI for so long, where you're gonna have this advancement and it goes up, but then it plateaus again for a long time until you, you know, maybe 10 years down the road or something, there's this next development or this next kind of series of research that results in this next level of advancement. And then on the narrative side of things, it's like, like you were saying, you know, you have this MIT study, you have just the general things that these companies have been saying, the way that GPT-5 comes into all this. But then you also have these increasing like stories from employees talking more about how AI is not making them more productive, is not making things better. And I think the big thing, like the past couple of months, has really been like just the growing wave of these stories about the mental health consequences, about people committing suicide after having talked to chatbots and gotten really concerning series of dialogue from them where they're basically egging on their suicidal ideation. And, you know, there's even this story about this guy who's like high up in open AI who is apparently kind of having mental health consequences as a result of this. I don't know the best way to describe it, but you know, it feels like kind of the the stories about the health and human consequences of this technology are just growing so rapidly that you know, more and more people are like, what is going on here? Yeah.
SPEAKER_00One thing that has always been sort of unique about the AI boom is that it has sort of required all this forward motion, all these promises of AGI and things like that to sort of overtake any critical backlash or introspection. Because it's always been there from the beginning. This is a technology that has never been sort of a majority of people surveyed, for instance, have never said, I love this technology. From the beginning and up until recently, you know, Pew has done polling, tech equity has done polling of California. There's lots of polling, and time and time again, it you find that consumer sentiment and worker sentiment is more negative than positive. People are more concerned than excited by significant margins over AI, and they have been. And that's always been that even some industry insiders have have pointed out, like been kind of the risk of touting this technology as so powerful, right? Like the Doom hype was, I think, taken as a tactic because it was working for a while. But now, as you're saying, we might see some of that sort of backlash come because it used to be like, well, if this technology is so powerful that like it's going to take over the world, at least it'll help me sort of replace my workers, or at least it'll be um sort of addictive to users. So we all better invest in it. And then if that pitch can't bear any fruit, if it turns out like, well, actually, it's just like a semi-successful automation technology that is only useful in a few key contexts and with a lot of oversight and work, and we have to hire other people to make sure that the AI works and we have to pay the, then all of a sudden that whole calculus is thrown out of whack and that those criticisms can then sort of shine through and take up more of the space. And I think you're right. I think we'll start seeing that happen more and and sort of you know those very real, and that's also not to say that, you know, those criticisms haven't been more developed and become sharper over the years as we have more data to point to, more kids whose lives have been ruined by, you know, AI addiction, more educators just completely exasperated by the way that AI has sort of taken a wrecking ball to the classroom. And then uh yeah, what we could talk about, which is which is labor.
SPEAKER_01Before we pivot to the labor question, can I just ask you one final thing on this AI bubble? And then we'll get into your series and the work that you've been doing on that. I wonder how you feel about the state of that bubble at the moment. Because for me, I feel like there's certainly questions in this moment, right? Questions that are being asked much more publicly. There is clear evidence of the vulnerabilities and the problems and kind of the lies that this market, that you know, the valuation of this technology was built on. I think there's still energy incentive to try to keep this bubble inflated, not just from investors. But for me, I always look at how the technology has kind of become this geopolitical football where you have all these countries trying to pretend that they are going to be leaders on AI too. And I feel like even if these vulnerabilities are becoming clearer, these issues with the narrative that the valuations were built on, I still think it's entirely possible that the bubble remains inflated, at least to a certain degree, because of that kind of geopolitical aspect, the aspect that is beyond the business case. But I wonder how you kind of feel about where that stands at the moment.
SPEAKER_00Yeah. This it's this has always kind of been been my sense as well that I think for a number of reasons, AI is essentially at this point, you know, too big to fail. And I, you know, I've actually I've had some really, really good arguments about about this with folks like Ed Zitron, who, you know, can also persuasively make the case that there's a house of cards quality, especially to a lot of the companies, and that once, you know, things start going south and investors pull out, you know, a lot of the what the AI companies are doing is is unsustainable. And I, you know, I think that's that can be persuasive too. But we've already seen precisely what you gestured towards, which is that we're already in this new kind of era where there is a new sort of formation of Silicon Valley and the federal government in the US, especially, this new sort of silicon state here that is still much more maybe insulated from a lot of like sort of you know the market activity, and that the state can do a lot to prop up a company, as we're seeing, right? Like we're seeing like the state is taking an actual stake in in Intel, for instance. It's making weird deals to exempt NVIDIA from export controls, and it has you know close relationships with a lot of the AI companies and their their architects and their executives.
SPEAKER_01I will say it was interesting to see the exemption that NVIDIA got, and then like Howard Lutnick basically turned around and made some comment on like how they were gonna have to treat China differently or whatnot. And China like immediately was like, Yeah, we're discouraging anyone buying these chips at all, even if they're now available.
SPEAKER_00Even if they're for sale. Yeah.
unknownYeah.
SPEAKER_00Yeah. Which is to say that, you know, I think so that's on one layer. Like the state, like, whether that whether it's through direct contracts, whether it's through actually taking a stake in a comp, which is, I mean, we're that it that's interesting. That's not really something that I necessarily would have put on my bingo card seeing like Trump want a 10% stake in Intel or anything else. But now, you know, I we have Have to understand like how interested it the the state is in AI. This is a point I think we made on System Crash, but I'll make it again here. And that is just like everybody should be asking themselves why it is that the one technology, for all intents and purposes, the one non-overtly military technology that the Trump administration is interested in is AI, right? Like he's cutting subsidies for clean tech, gutting electric vehicle supports, pulling the rug out from under, you know, health sciences and investments in vaccines and things like that. And yet here we like AI. We're pro-AI. And that's because it's so well suited to be a technology of control, of domination, of surveillance. It is, it can produce shitty propaganda that the White House can put on its Twitter feed. It can, you know, be pumped into government agencies in hopes that it can do the jobs of fired public servants and all the while sort of concentrating control under a fewer number of officials and and sort of allies. It is, and it can be a tool, you know, at least as the way that Silicon Valley has pitched it, for uh military might for on the geopolitical stage, as you said, whether it's to conduct sort of uh you know hacking uh malware attacks or or you know help guide more conventional weapons or do target selection as we've seen the IDF do in Gaza. So the state, it as at least right now, is is also quite invested in AI. The American state is. So that's one factor. The second factor is like whether or not this is comparable to the dot-com boom of 20, 30 years ago, or AI is in any way, shape, or form the next internet or anything. I do not think that it is, of course. But that's what the industry is treating it as. That's like this is their idea.
SPEAKER_01Everything is the next internet. Crypto is the next internet web three, AI is the next internet.
SPEAKER_00But more than crypto, more than the metaverse, more than NFT, there has been a convergence on this and investment in this idea that has kind of made it the only game in town. And that's not to say they can't, you know, scatter to the wind and pivot away or try to afterwards. But that that's number two. Silicon Valley, I think, is too sort of invested in in this idea and propping it up. So I think we'll see some some interesting things happen if if and when that bubble bubble, or rather, when that bubble does start to burst. I think it will burst. What what happens next will be the interesting thing, whether there will be government intervention or how the companies will react and this the scale of that bursting. And number three, and I think the dark horse factor here is just that it's such an alluring idea for the clients of this technology to have a tool like AI that can automate and labor and surveil uh the smaller workforces that remain, in theory. It's a much more appealing pitch than like than crypto was where you know if you're whatever Walmart, you're looking at crypto and going, like, how does this? I don't care, you know, and how does this affect me? But you know, there's it's that there's a reason why like almost every organization has been like, How do we do AI? Like, how do we get AI? Every like CFO in the world has been like, all right, bring on the AI, you know, let's uh let's cut labor costs here. So I think those three factors are gonna make this uniquely sort of resistant to a bubble. It also might make it all the more cataclysmic if and when that bubble goes full burst.
SPEAKER_01No, I I think you've put that so well. And I think it pivots us really well to start talking about labor. But there's one thing I want to tell you before we start talking about your labor reporting and what you've been hearing from workers and what you're seeing. You know, obviously on System Crash, we used to talk a lot about what's going on in Canada, and you know, Canada has a new AI minister.
SPEAKER_00I'm sorry, remind my remind me what is that like a city in Europe or what what yeah, that's your 51st state.
SPEAKER_01Don't you remember? Like oh, that's right.
SPEAKER_00That's right.
SPEAKER_01Not Washington, DC. It's uh Canada. Um but uh so the AI minister gave an interview the other day, and he was like, there's this bill from the old parliament that I needed to understand. And so I ran it through Google Gemini, and I had Gemini make a 15-minute podcast for me about the bill to explain it to me. And I listened to it on the car on the way to work, and it was fantastic. Let me pull it up and and let you listen to it. It was great. And I was like, Man, tech policy in Canada is so fucking screwed if this is what we have. It's bad up here, man. It's bad. Not as bad as down there, I know, but like it's it's not good.
SPEAKER_00I think that's explicitly like there's a Bloomberg profile of like Sacha Nadella, I think. And he that's what he said that he did. He would like download books into chatbots that he could then talk to and ask questions about. It's just like it's so deranged. Like, yeah, what what even is that piece of information anymore? It's just it's like already been, you know, uh regulated and processed through the mass of human knowledge into something that like could probably not even be discernible, like as the book anymore. So you're just like you're just talking with nothing. It's just like it someone might as well be just like blowing hot air on your face. Like it's just so ridiculous. But yes, you know, the UK, Canada, like I I continue to be astonished by the the openness and the eagerness that you know states around the world have, you know, not all, not all. There's plenty of there, there are some good, you know, exceptions who have basically said fuck off. So uh, but yeah, it's Canada screwed, UK screwed, we're obviously screwed.
SPEAKER_01But man, the Anglophone world, we're oh a mess. It's you know, unfortunate. But yeah, let's pivot and talk about your labor series because you know, I think you set us up really well to get into it. You know, you have this series AI Killed My Job, where a bunch of people have been sending in their stories about how they're seeing AI affect their professions, their workplaces, what the effects of that are. Of course, you've published two pieces, you know, kind of directly telling those stories so far. And, you know, I know you have more kind of in the pipeline as you're going through more of these things. And for me, it's been really fascinating to read through that and to see the types of things that people are saying about their workplaces, about their work. Of course, you know, the ones that you've published so far on tech workers and on translators. But even then, I think there are so many things that just feel so much more like broadly applicable, potentially, of the things that they are talking about, even you know, going from some of the things that you were you were just saying. So I guess to start, like, how did you decide to to do this series? And was there anything that you were surprised about when you started to get these stories coming in from people telling you about what was happening in their work lives?
SPEAKER_00Yeah. So and I'll and I'll also just add a note to say that I think that the talk of the AI bubble and that some of the mythology and the more sort of pie-in-the-skyness of the AGI conversation sort of beginning to evaporate kind of allows us in a lot of ways to see generative AI and the generative AI tools being sold by these companies for what they are, which is either you know socially mediated kind of entertainment products like chat bots that people talk to, or it's enterprise, workplace, or personal automation. It's just it's it's sort of you know souped up productivity software. And so, you know, with that in mind, which I mean, that's kind of how I've always approached generative AI, as you know. That's I think to me, the most useful way to look at a technology like this that is being sold as something that is going to disrupt the workplace or transform work or, you know, beget a jobs apocalypse, in the words of some of these AI CEOs, is to just like look at history and look at all the times when similar pronouncements have been made and other technologies have been sort of, you know, entered into working life. And so the best way to do that is just look at the material uh conditions on the ground and who's doing the introduction and the adoption and how it's changing. So the AI Killed My Job series came about because I spent a lot of time talking to workers and I have since the beginning. And I part of that is just because, you know, that's just sort of like where my beat has naturally been talking to before it was AI, I was talking to Uber drivers, talking to Lyft drivers and Amazon workers, and trying to understand, you know, what was happening on the ground on the other end of Jeff Bezos's or Travis Kalanak's pronouncements about the transformation of this, uh, you know, of the workplace or the future of work or whatever. And it seemed especially acute to me that during the AI boom, so few people were really just going right to the workers. So, like, okay, great. Dario Amode from Anthropic says that whatever, 10% of all jobs are gonna be gone, maybe half of all young collegiate. So, okay, great. He's a CEO selling a product. So what what's actually what's actually happening? Like, where is the technology actually like hitting the pavement? And that's usually you can find that out by talking to the people that these tools have been thrust upon or that are using them voluntarily or that uh are parts of organizations.
SPEAKER_01And when you say when you talk about the tools there, do you mean the the AI products or do you mean the executives? Sorry, I couldn't help myself.
SPEAKER_00They're both uh tools in their own way. Yeah, I'm finally reading Why We Fear AI by Hagen Blix and Ingeborg Glimmer. And an important point that they make is that you know, that these companies and the the AI salesmen are all just being, you know, like motivated by the same capitalist forces that are animating the whole the whole to do so in in a sense, they have to say that, you know, like our product is gonna put even more people out of work than yours. And uh and then it becomes a sort of arms race to to see who can scare people more, but but I I digress. Another point that I want to make without digressing too much before we talk about the workers, is that talking about AI and labor replacement or or a jobs crisis is is so fraught because on the one hand, Ammoday, Altman, all of the AI CEOs, they do want to create the impression that like a great disruption is coming that makes it easier for them to sell more automation software, right? Like that's it's a it's a product in the array of products that they're selling is enterprise AI, automation, productivity software. And I think there is a tendency, even on the left, to sort of like push back, completely saying, like, this stuff is bullshit, it sucks. And giving that idea any credence is just like playing into these corporate narratives. And then some of the, I think some of the better critiques, like from Aaron Beninov, who's great, looking at sort of the sort of middling results of like sort of previous sort of in industrial, you know, mass-scale job scares and saying like that just it isn't borne out. I do want to caution against minimizing too much. So, and I think that's part of this project is that I think in no way are we gonna see anything like a mass-scale jobs apocalypse. It it it's it's not gonna happen. The AI is not suited to do enough jobs. It's it requires too much oversight, it's too expensive. But that said, there are still a lot of use cases where a management can either use it as a tool, use it as leverage to immiserate or yes, sometimes even replace workers or freelancers, especially, where there are workers in more precarious conditions. And so I do think we want to be careful about swinging the pendulum too far the other way, because I have been talking to probably hundreds of workers at this point. And that's not like a, you know, obviously a meaningful sample size if you're looking at the global economy or the American economy, but it's enough for me to get a sense of what I think is happening in particularly sort of vulnerable industries where executives can use AI maliciously or aggressively to cut costs. And in a lot of cases, it can still be very pernicious in the way that it is used to sort of reshape a job or to take away parts of a job that people think are meaningful and replace it with like with button pressing, or where like your job used to actually be to translate the text, for example. Now, because some you know person in middle management was uh susceptible to a pitch from some tech company, now the part of translation is outsourced to a machine, but you still need a human to go over the output and correct it. And sometimes I heard over and over in my survey of translators, that job was just as time consuming sometimes, but it's just far less like you're not actually doing the translation, which is considering meaning and considering context and place and person and painting a picture of a game or a piece of art or prose and then translating that. Instead, you're taking the automated output and trying to see if it lines up because somebody somewhere on the supply chain got convinced that that's more effective and it can save the firm a few bucks. So there are a lot of impacts like that that are still rolling out and that I'm hearing about. So yeah, I I really just wanted to hear from the workers. I guess that's a long-winded way of saying that, like, I wanted to hear all this corporate Silicon Valley AI speak. How's it playing out on the ground by the people who have to deal with this stuff every day? And yeah, I started, I decided to sort of separate it by industry for now. I might, I might do other things as I as I move along. I've started with tech workers because they're in a very interesting place. It's one of the more, you know, obviously management at a lot of tech companies is the most uh sort of gung-ho and aggressive about deploying AI. And then so it becomes interesting to see like, you know, the disparity between uh an AI-loving executive and a senior software engineer who really knows what they're talking about and is just going, like, I can't believe we have to use this stuff. Or, you know, and so there's a lot of great stories that came from that one. A few where it was, you know, we think that the AI, boom, sort of convinced our our executives to close down a department or to f or a fire me as part of a layoffs, as part of an AI first strategy or something. Um, but more often than not, yeah, it was like, I work for Google, I've worked here for a long time, and they're automating sort of like the AI generated coding process, and they're just just injecting it directly into our code base, and I think that's a disaster waiting to happen. So that was a really interesting one, you know, stories like that.
SPEAKER_01I completely hear where you're coming from with that kind of distinction between what jobs are being destroyed and kind of what jobs are being transformed, and often transformed in a way that is degrading them, right? Making them more precarious, lowering the pay, making the work more frustrating, I guess, maybe in order to have to deal with. Like I think that these are important things to understand. And when the executives are just coming out and talking about productivity and replacing jobs, like the actual material consequences of that can be abstracted, right? But when you go in and actually talk to the workers, you can see what is happening there. And my view has always been in part informed by like the last AI wave. I've talked about this in the past, that the actual job destruction is often minimal and is often focused on particular tasks. That's my view. And what we see much more of is the use and kind of the weaponization of these technologies by bosses, management, and executives, you know, things that you've, of course, written plenty about through your career, in order to try to change the work to make it so that workers have less power, so that they're being paid less, so that they have fewer abilities to really intervene in the work process. And I feel like this was something that really came out in some of those stories that the translators and the tech workers were talking about, where it really felt to me like some of them were saying, like, you know, in the translators' case, that with OpenAI and with LLMs, the quality of the translation has not actually gotten significantly better. But it seemed like more like the hype of the past couple years provided a justification for a lot of these companies to adopt and roll out these tools and change the profession of a translator in a way that wouldn't have been justifiable in the past, but because of the hype, it was now okay to do it, even though the quality wasn't there. And that to me seemed like a really significant kind of bit of detail to come out of these things that you were talking about. So I wonder how you reflect on that piece of things after and and the way that executives in particular have been able to take advantage of this after talking to the workers about how they have seen it actually play out in their in their professions.
SPEAKER_00That's absolutely correct. As I put in a in a previous piece, in one that actually helped spur this project, and one that I spoke to a laid-off Duolingo contractor after that company pivoted to AI. And this was the same time that sort of like the doge clearings of houses was at in full effect. So I wrote a piece that argued that sort of the real AI jobs crisis is sort of the cultural logic that it allows executives to embrace and to you know impart onto their organizations. But it's it's less that, you know, AI can actually do any one of those jobs of the civil servants that have gotten laid off. It's just it provides sort of like the window dressing, the cover, the idea, like the futurity necessary to at least sort of gesture towards this concept of replacement, or that there's going to be the same level of functionality even after these people are gone, when it's just really what management wanted to do, anyways. And I think there are some cases, the Duolingo case is interesting because I think this is just one of those guys that really seems like he's just really does like either believes in the hype or maybe he had been itching to get rid of all of his contractors for years, anyways. I mean, we we we can't know. He certainly seems very credulous about the the capacities of AI, or at least he did until everybody sort of revolted and and and started pushing back. But I think that's a lot of it. But there is there there is this layer. So, like, I also don't want to minimize the experience of the people that have said, like, my work is gone, right? Like they my work is gone. It's dried up. Like I that that cultural cover allowed my boss to select the good enough option, which is sort of, you know, auto-generated code in the case of the tech workers that, you know, in some cases can be good enough. Again, if you have somebody, maybe you can hire somebody who's less expensive to sort of spot check that output. Or in the case of translators, maybe it's just like, you know what, consumers of this particular like Japanese video game that we're putting out, maybe they don't need a good translation, or maybe this is good enough. They can still get the gist and they'll still, you know. And so it facilitates like those trade-offs and it does sort of again provide cover to management to make these decisions. And because like ultimately you just got it, like it's AI in these contexts is an automation technology. At the end of the day, it's just like it's what management chooses to do with it. And usually that is just again, yeah, squeeze, surveil, control, or or replace tasks or jobs that management think it can. So it's gonna, it's gonna be deployed in those same contexts that automation technologies always have been. There's nothing particularly mystical or or you know, befuddling about it when you actually get down into the details. As such, it still stands to be a pretty potent force because it's been imbued with these properties, right? Because logic has become powerful enough. And I think that going back to what we were talking about at the top uh about the bubble, one really interesting thing to see will be whether or not that sort of wipes away some of the eagerness to use this as an automation technology. Is it gonna be like, oh, like actually we were overzealous on this? Maybe we, maybe it can't do everything that we we were sold on it being able to do, and now we have to change tack. Or there's a potential, you know, fork in the road where it's like, well, we've sunk all these costs, anyways. Everybody's just gonna have to deal with subpar output, subpar cultural products, subpar customer service experiences, and AI is going to win the day because it's a little bit cheaper and we've already bought the enterprise contracts.
SPEAKER_01Yeah, or in the case of a country like the UK, apparently they're looking to just buy an open AI subscription for like the whole government or something. Like it's wild. But I think that's really interesting, right? Because as I was reading like the translation piece in particular, I was also thinking about how I have seen this being used in like other parts of the world as well. Like I spoke at an event in Amsterdam earlier this year where they were using basically AI like live translation of speakers. And I was like, this is weird. Like, who knows? What that thing is like claiming that speakers are saying when they're on the stage. But again, like it's in place of where in the past you would have someone actually like doing that, right? Or maybe you just wouldn't have it at all. I don't know. And then I was speaking to some publishers who operate outside the English language, and they were like frequently using Chat GPT for like correspondence to English speakers and stuff like that. I don't know. I I like gave them a bit of shit for using Chat GPT, but I also like kind of understood it like, you know, if English is not your first language, that makes it a lot easier to potentially converse with people outside of that. And so like I feel like I've been picking up on a lot of how I'm seeing people using these tools and normalizing these tools and, you know, not super comfortable with it, obviously, but you know, sometimes I feel like it's a bit more prevalent than I expected it to be. And that makes me wonder like what kind of the the you know, say post-bubble burst kind of use cases of this technology are are going to be.
SPEAKER_00I mean, it is pretty pervasive. It's being used by hundreds of millions of people every week. There's a pretty big user base, I mean, which uh is gonna open up a whole nother can of worms because a lot of those use cases are extremely unhealthy and concerning. I think, like, was it Harvard Business Review did a survey of like the most common AI uses, and right at the top was like therapist. People were, you know, treating it as an AI therapist, and you know, that's just like just like red flags just came tumbling out of the sky for me on that one. But yeah, it it is it is terrifying. So it is, and and look, like I think that you don't have to deny that there are some genuinely like interesting context and use cases. I remember when like you had a computer vision that could, you could like take a picture of like a road sign in a foreign country and then it could translate that. And you know, that was something that you just, you know, you could try to ask somebody what it meant. And but and a lot of times you maybe you can't find somebody who speaks the same language. So there are like utilities where people are are finding it useful. There's also people just like falling into the trap because it's so useful or so it's so much easier for them to do this than, I mean, famously homework, right? It's so much easier to just like have ChatGPT generate answers for you than to than to actually do it.
SPEAKER_01Oh, I definitely know people who turn to it to answer like any number of questions instead of just having to think about it themselves or even turn to Google as maybe they would have done in the past. And now it's just ChatGPT instead, right?
SPEAKER_00Yeah. I mean, absolutely my my, I think some of those use cases will be like, I think, filtered out. Some of the mass automation stuff will will eventually be be filtered out. The youth, yeah. I mean, I think uh in other cases it's we, I think we were gonna be stuck with hard questions about like what we're gonna fight for and what we're going to, because there's that famous line about how AI and technology were supposed to like automate doing, you know, the dirty work, doing laundry and dishes and giving us time to do art and music. And instead it's automating art and music and forcing us to spend more of our time working, doing the groundwork. And I like the the in the impact on creative industries is something that's just gonna have to be negotiated against, fought against. Same with, I think, translation. Like, do we value translators? I I think we do. I do. Like doing this piece really has underlined my my sense of the importance of this work. And in in this sense, I'm really grateful for having done the this piece and really thinking about like, oh yeah, like how many like translated works of uh you know, books have I read over the years? A lot, you know, like how many translated documents, a lot of them, now recognizing sort of like the art and the labor and the toil that goes into that process, and and really, you know, spending time with folks who and their stories who love that, who love the act and the art of taking something that somebody else said, thinking about it, contextualizing it, and then making it accessible to a whole nother culture, creating an inter intermediary between cultures. This stuff is so important. I mean, it it it probably wouldn't make the top 10 list of things most people are concerned about in in AI or automation, but now, you know, this the prospect of automating that process feels incredibly sad to me, you know, instead of actually, you know, humans uh putting cultures in touch with one another and negotiating those meanings together, discussing them and sort of ensuring that things are as best accounted for, the details, the nuances, the color, the, you know, you name it, it's all intact. You know, I feel like something really stands to be lost if we just automate these processes and it's like in one side, out the other, and we just have these like tubes of content production that are going each way. And I want to take this opportunity to shout some of the groups that are that are kind of standing up and trying to trying to fight against this. And I would love to spend more time talking about them. There's a there's a group with a name after my own heart, translators against the machine. They're a group that's sort of gathering stories and data about what it's like to work in translation right now in order to sort of build solidarity and to fight the the encroachment of tech companies in into their professions, because they it is really important, I think, what they do. And it is one of these areas that I think Silicon Valley companies do stand to sort of grind away, you know, whether just for a few extra enterprise automation contracts or just as sort of like a thoughtless byproduct of this rush to to build and and and release these products. And there's also, if you're a translator who's worried about or interested in organizing around the impacts of AI, the National Writers Union has a translators organizing committee. And you should you should check them out at nw.org slash chapters slash T O C. So there are folks who are out there doing some stuff about this, and and it's I I think it's a space, like I said, that it's not going to go away. If the AI bubble bursts, there are still gonna be these automation products that are widely available and in use, and you're still gonna have executives and clients who like who still will want to use them. And and it's going to have to be, it's going to be a fight, as it is with, I think, you know, uh people in the arts professions, illustrators, copywriters, graphic designers, screenwriters who already, you know, who won the first round of their fight, you know, and there's going to be many more. And finally, I would shout the work of uh Lucille Danilov, who's a translator who's written a lot about games localization, um, and has uh has a website called lockedandloaded.net that you should you should check out if you're a translator or interested in this stuff.
SPEAKER_01Awesome. Yeah, we'll we'll put those in the show notes so people can more easily find them instead of having to remember what you were what you were saying there. But I think that's fantastic that that you laid those out. And I just wanted to pick up on a couple things that you were saying, right? Like, you know, we think about the ways that these technologies are rolling out for language and and translation as well. It immediately brought to mind what I heard from Maori uh speakers and people who advocate for that indigenous language in New Zealand and how they're worried about, you know, how AI will continue to hamper efforts to kind of, you know, renew, restore, uh enliven that language and keep kind of the older pronunciations and things alive as it just becomes jumbled and treated as this translation of English and and be related more to English rather than what it previously was. But also, you know, as someone who is from like an officially bilingual country, I think having that kind of back and forth, that proper translation and and you know, the understanding of the context between the two languages, even if you are a monolingual person and are trying to engage with like the whole of French and English culture in Canada. And, you know, I I believe there were a couple people in the article who were from Canada who were kind of talking about things like this. Like I think that there's such a huge loss if instead of having these translators who can translate that context, who can who can actually like have the meaning there instead of just you know replacing words is gonna be such a loss for you know a country that still has kind of linguistic divides and identity issues around language and things like that, right? I I think it potentially harms some of those kind of like national unity questions there as well, like, you know, these kind of bigger issues that we talk about. And, you know, obviously you guys in the states have, you know, English and Spanish. It's a bit different up here where it's like, you know, officially bilingual on the government level and and things like that, right? And you know, just to close off our conversation, I wanted to ask you, you know, you have been talking to so many of these workers, you have been writing about, learning about, speaking to workers for so long about this. But I wonder after doing this project, AI killed my job, after hearing so many stories from people about this latest wave, you know, has this changed how you assess the impact of AI on work after doing this for so long? Like, what has been kind of the main takeaways that you've had from this experience?
SPEAKER_00You know, I wrote a a year and maybe even a year and a half ago, like before I was really even fully doing the newsletter. And I would just kind of like jot out some thoughts on it occasionally. I wrote a I wrote a post, I think it was called Understanding the Real Impacts of AI on Jobs or something like that. And I was just randomly going over it again because it it popped up and when I was looking for going through my archives, looking for something to link to in a recent piece, and pretty much everything that I predicted would happen has more or less been borne out so far. The fact that, you know, it's really going to be a question of management using AI as, you know, as a tool to sort of cut labor costs when possible, to sort of to concentrate their their power or gain control in an organization, to use as leverage, which we've seen, you know, happening to some extent for sure, where less than, and I also, you know, I didn't think from the beginning that we were going to see a jobs apocalypse either, or that it would be sort of this mass unemployment event. Um, and you know, no, I I I think I I've been a little surprised by, or at least a year or two ago, I would have been surprised at sort of like the pervasiveness at how many corners that companies have been determined to just ram AI into. Just, you know, part of that's just like FOMO that like everybody's saying AI. AI is the buzzword. Like, if I don't, if I'm a middle manager and I don't find a way to like have some kind of an AI program, my boss is gonna think I'm stupid. So I better get it in there. I've been a little surprised by the extent to which, like, in education that a lot of the teachers have been adopting AI in certain contexts. So, like, I feel like there's case-by-case instances where I'm a little bit surprised by a certain use case. And even when it seems like it's obvious that this isn't a great idea. And as that MIT study found 95% of the time, it's just like not gonna generate any real savings or advantages for your firm or your institution. So yeah, I the the zealousness, maybe a little bit. So I should also say that the project is ongoing. We have at least four more installments to do. And so if you, dear listener of Tech Won't Save Us, have had AI kill your job in any way. And I should say, like, I I'm a little, I'm still to this day ambivalent about because AI killed my job. It's supposed to be, you know, AI has like made as changed, transformed, made unpleasant, immiserated. In whatever way, blanket. It's not, I don't mean to give the impression that like AI is an autonomous force that's just killing jobs. That's the antithesis of everything I'm about. It's supposed to sort of help explode that that myth as AI is a sentient thing.
SPEAKER_01But it wouldn't be as catchy if all the additional context was in there, you know.
SPEAKER_00I tried. I was like, you know, the AI that my boss bought from an enterprise AI company killed my job. So it just wasn't flying. So so if AI has killed your job, or you know somebody who's dealing with AI in the workplace or who has seen their work fall off, it's AI killedmyjob at PM.me. It's a proton mail account. I'm particularly interested in the next two installments are going to be healthcare workers. So if you're a nurse, if you're working as a therapist, if you're working in a hospital, an admin, if you're a healthcare worker, I would love to hear from you. And and secondly, I'm I'm looking at illustrators and graphic designers, artists, people whose work has been impacted by the rise of uh services like Mid Journey or um or or Dolly, which is not just Chat GPT, but the the image generation side. So, but everybody uh where there's gonna be more installments after that. Those are just that likely to be the next two. So I would love it. If you have one to share, we'll throw that in the show notes too. Or Paris will. I'm no longer in control of throwing things in that show notes.
SPEAKER_01We'll see. Maybe I'll put it in the show notes. Yeah, that that's great, Brian. I think it's so important that you're that you're doing this. I think it sheds such important light on, you know, this facet of what we've been seeing with this AI hype over the past few years and really kind of grounds it for us, right? So that we can actually feel the tangible effects that this is having, which, as you say, is not always destroying a job, but can still have, you know, massive, terrible repercussions for people's lives, people's work, how people are living in this world, right? And often it doesn't get near the attention that it deserves, especially when we see these statements from these executives absolutely everywhere. So I think it's a great series, and I've really been enjoying it so far and can't wait to hear uh and read the next uh installments coming. Installments, thank you.
SPEAKER_00Well, it's certainly been really eye-opening to me to talk to all these workers and hear their stories, hear your stories. And I'm so grateful to all of the all of the workers and translators, tech workers, and everybody still to come who has submitted stories, answered my questions over email, and and you know, started bringing to light these issues and the reality of AI on the ground. So thank you to everybody who's participated so far.
SPEAKER_01Absolutely. And Brian, great to speak to you as always. Thanks for coming back on Tech Won't Save Us after such a long period where you weren't here. But that was, of course, because we were talking every week somewhere else.
SPEAKER_00We were doing our own show. Oh, Paris. It's always good to be here. You know that. Always a pleasure. I'm sure I'll uh see you again before long.
SPEAKER_01Brian Merchant is the author of Blood in the Machine and writes a newsletter of the same name. Tech Won's Save Us is made in partnership with The Nation magazine and is hosted by me, Paris Marks. Production is by Kyla Hewson. Tech Won's Save Us relies on the support of listeners like you to keep providing critical perspectives on the tech industry. You can join hundreds of other supporters by going to patreon.com slash tech won't save us and making a pledge of your own. Thanks for listening, and make sure to come back next week.