The AI Boom Is a Corporate Project. It Can Be Disrupted

A lot of AI technology may not work very well, but providing consumers with reliable information was never its major commercial promise. Investors have poured billions into generative AI because of the promise that it will eliminate jobs on a massive scale. Tech journalist Brian Merchant joins Movement Memos host Kelly Hayes to discuss how the AI boom is being used to displace workers, degrade information, expand surveillance and strengthen state violence—and why this brittle corporate project can be disrupted.

Music: Son Monarcas, Mizlow, and Daniel Fridell

TRANSCRIPT

Note: This a rush transcript and has been lightly edited for clarity. Copy may not be in its final form.

Kelly Hayes: Welcome to “Movement Memos,” a Truthout podcast about organizing, solidarity, and the work of making change. I’m your host, writer and organizer Kelly Hayes.

What happens when a technology doesn’t have to work particularly well to remake the world? Today, we’re talking about AI in the workplace, Google’s decision to ruin its search engine, robotaxis getting people arrested, Flock cameras getting destroyed, the movement against data centers, and what it means to build power against the tech industry.

My guest is my friend Brian Merchant, author of Blood in the Machine and creator of the newsletter and podcast of the same name. The AI boom is often presented as an unstoppable technological shift, but Brian’s reporting reminds us that it is a brittle and faulty corporate project — and that corporate projects can be disrupted.

If you appreciate this podcast, and you would like to support “Movement Memos,” you can subscribe to Truthout’s newsletter or make a donation at truthout.org. You can also support the show by subscribing to “Movement Memos” on Apple or Spotify, or wherever you get your podcasts, or by leaving a positive review on those platforms. Sharing episodes on social media is also a huge help.

Truthout is an independent news organization, publishing stories that the craven corporate press won’t touch. We are a union shop with the best family and sick leave policies in the industry, and we could not do this work without the support of readers and listeners like you. So thank you for believing in us and for all that you do. And with that, I hope you enjoy the show.

[musical interlude]

Kelly Hayes: Brian Merchant, welcome back to “Movement Memos.”

Brian Merchant: So good to be back. Thanks, Kelly.

Kelly Hayes: How are you doing today?

Brian Merchant: I’m hanging in there. There’s a lot going on, there’s always a lot going on these days, and I’m a little bit hot because I am in Europe right now, where the heat wave is just receding. But aside from all those small things, I am doing well, thank you. How are you?

Kelly Hayes: Well, I am in Chicago, where another heat wave is rolling through, and the air here is terrible right now. So, I’m just going to hide out in my apartment and talk to you and feel good about that.

Brian Merchant: Sounds like there’s some kind of a theme developing here with the heat, and, you know.

Kelly Hayes: There might be something going on there.

Brian Merchant: Too early to say for sure, but yeah, I think let’s circle back to that. Let’s come back to that one.

Kelly Hayes: Huge if true.

So, Brian, a lot of people are familiar with your work, and some of our listeners will remember you from the last time we were in conversation, but for the unacquainted, what would you like people to know about who you are and where you’re coming from?

Brian Merchant: Yeah, I think, I guess most pertinent to what we will be discussing today is that I am a longtime tech reporter and tech writer, a journalist, and author of a book called Blood in the Machine, which is about the history of industrial automation, and what we can learn about it today. And I’ve been writing about that sort of framework, using that framework and applying it to AI, and this most recent surge of corporate automation that we’re seeing handed down from Silicon Valley, and what’s happening to labor on the ground, and I write about that at my newsletter, Blood in the Machine.

Kelly Hayes: And you also have a new podcast connected to that newsletter.

Brian Merchant: I do, yeah. It’s something that’s still early days, thank you for shouting it. My producer would be quite glad that you did. Yeah, it’s an audio component, it’s something that folks have been asking for for a little bit, and I’m hoping to host organizers in the fight against data centers, organizers who are trying to confront AI in the workplace, people whose jobs have been impacted by AI, authors, journalists, documenting this stuff in the front lines. So, yes, that’s the latest endeavor. At time of recording, our last guest was a participant in the Summer of Ludd activities in New York City, which was a big anti-big tech weeklong slate of programming, protests, and affairs. So, that might be a good one for folks of “Movement Memos” to check out.

Kelly Hayes: I really appreciated that conversation, and I’m excited about this podcast. It’s only been around for a few episodes now, and I am looking forward to keeping up with it. So, I hope folks will check it out.

Brian Merchant: Yeah, thank you.

Kelly Hayes: To give people a general overview of where we are in our experience of the harm and hype of AI, for a long time there’s been a talk of a bubble, we’ve seen a lot of AI use cases fail, a lot of people pushed out of their jobs, or forced to do their jobs in ways that are harmful or simply don’t make sense, and even some people fired and then rehired. Given all of that, what is the current state of so-called artificial intelligence and our relationship to it?

Brian Merchant: We were chatting a little bit before we started recording, and the word “messy” came up, and I’m going to go with messy. The state of affairs is a complete mess right now. It is incredibly in flux. So, as just a little bit of background, so ChatGPT, most people are likely aware on some level, really blew up at the end of 2022, it spurred this rapid-fire development, investment, and commercialization cycle that saw AI, generative AI, what we now think of as AI, as a product, as a technological trajectory, take shape. So, it’s really been the last three and a half years. And things have gone hyper speed in terms of the industry’s ramping up, again, investment in this technology, and turning it into the slate of products that are now inescapable today on your app store, on Google, on just about everywhere else.

And so, the largest thrust behind all of this, sure the chatbot was a flashy tech demo, and a lot of people were interested in using it for that reason, as a consumer product. But we have to bear in mind that there hasn’t really been any serious monetization there. And in fact, there’s a lot of reason to believe that investors were never really as excited about, say ChatGPT, the chatbot, as they were about this promise that has undergirded the entire AI development boom, which has been this promise that AI can replace labor, that it can replace jobs. That is, according to my analysis anyways, according to my understanding of the cycles of history, and the industrial development, and how these technologies are developed and sold.

And if you look at, for example, OpenAI’s charter, where it says its goal is to build AGI, or an artificial general intelligence, and it explicitly says that it explicitly defines AGI as something that will replace most meaningful work. And so, that is a direct sales pitch to investors, and it has been for now three years and counting. And so, whether or not we see success — and you mentioned that there’s been a lot of hiring, firing, rehiring, and a lot of questions and turbulence in the AI world — but we have to remember that, number one, undergirding this entire adventure from Silicon Valley is the promise that it will soon be able to eliminate all jobs. And so, there are a lot of executives that have bought into that, have believed it, there’s been a lot of pundits who believe it, there’s been a lot of middle managers who believe it.

And so, we have already seen a lot of disruption. We’ve seen tons of layoffs in the tech sector, particularly, layoffs in other sectors, we’ve seen hiring slow down among those looking for entry-level jobs, and we’ve seen creative fields in particular impacted by generative AI’s rise, in situations where you can produce an image or some text, or if it’s a job … Translators, illustrators, folks like that are actually being hit economically hard. So, this whole messy picture is still somewhat up in the air, there are a lot of unanswered questions, there are a lot of things that we still need to wait to see how they’ll shake out in the long run. And whether or not AI ends up being capable of replacing a lot of these jobs is one of those large questions that’s still looming.

But what’s important for our purposes is to know that that is the logic that has been imbued in this technology from the beginning. It’s an enterprise automation product above all, in terms of what companies are hoping to see in terms of cost savings and value created, obvious value stores. There are a lot of other harmful things that AI does, and a lot of other issues with AI, but in the labor picture, that’s the big one. It’s being sold as an automation technology. And three years in, there’s still intense debate over what it can and cannot do. But as folks who are thinking about this stuff’s impact on the workplace, and what we should be keeping in mind, it’s that your boss is always going to use this as leverage to push against workers to say: Well, maybe AI can be used to replace your job, so don’t ask for a raise. Maybe we’ll see if we can automate X, Y, and Z jobs, and then if we can’t, maybe we can hire back some contract labor to cut some costs. Let’s see if we can use AI instead of hiring an illustrator because it’s so, so, so much cheaper, and we’ll see if the client balks and if there’s any issues down the line.

So, we’re in this period where there’s a lot of negotiation going on, like, how far can we push our AI use without clients rebelling, without screwing up too much stuff … Because AI still gets things wrong quite a bit of the time. How much tolerance is there in the system for unfettered AI use? How much cost savings can we realize by using this stuff without completely tanking our products, tanking our workplace, and … without leaving some permanent damage?

Kelly Hayes: You recently did an episode about Google doubling down on its effort to rebuild Search around Gemini. What are the broader implications of that move, given how dependent so many of us are on Google Search?

Brian Merchant: Yeah, it’s a really good example. I’m glad you brought that up because it’s at the forefront of what’s happening with AI, and in a lot of ways it encapsulates that sort of push. Cory Doctorow famously calls it “enshittification,” I’m sure a lot of listeners have heard that term. But that’s what goes hand in hand with a lot of this job automation. And it’s not just job automation, it’s the automation of other digital features. And again, seeing what we can do with AI for cheap, that we can get away with, and hopefully eventually cut some costs down. Again, I should mention the caveat that finding those cost savings and finding big ones is especially imperative because AI is just, as a lot of you might know, is so expensive to run and to operate. It just requires tons of data, tons of energy, tons of salary cost to hire talent for these things. So, it’s just huge enormous costs to run AI. And so, the desire for these companies to find those cost savings is now accelerating.

So, all that said, Google has decided to, as you said, transform its search product, and in a very aggressive way, a way that 10 years ago even would’ve been completely unthinkable. So, many of us may know that if we use Google search now instead of getting an indexed list of search results, we get Google’s AI overview, which is an AI generated answer based on the query that we made. And it’s the AI’s best shot at just answering that question with the intent of keeping us on the page, instead of clicking through to a link that would take us off of Google, and onto say independent writers, or a newspaper’s website, or a small business’s website, or you name it.

And so, this shift is really, to me, showing what AI is all about. And I think it’s one that for folks like Kelly and I, is especially close to home. So, what it’s doing is it’s indexing our websites, and anything that we may have written, just taking that material, incorporating it into its data set, and then if we search for, “How is AI impacting jobs?” When before maybe it would send a link to Blood in the Machine, or, “What is the state of on-the-ground organizing?” And maybe it would’ve pulled a nice essay from Organizing My Thoughts, now it just spits that information out onto Google’s own platform, onto where Search used to be. And so, it’s transmuting all of this creative independent labor into a paste that Google can sell because Google does sell it.

It does put ads right underneath that new answer paste that, oh, by the way, along with the information gleaned from our websites, it also is putting falsehoods in there at a rate of one out of every 10 searches — or it’s wrong 10 percent of the time, is still the finding. It still hallucinates, as the AI researchers call it, 10 percent of the time. So, it’s generating this absolute mishmash of information that Google is hoping that it can profit from by selling more ads against, essentially, our labor. It’s taking that work and that effort and that information that was generated by other parties, moving it over onto Google’s side of the ledger, and then Google’s monetizing it and selling that information.

And again, the question is: Can it get away with that? Well, is AI Overview, Google’s product, good enough, or will users begin to revolt? Sadly, right now, it doesn’t seem like they are revolting in large numbers, they are just adapting to this new era in which Google is wrong part of the time. And we can talk about some of the things that are happening in that sphere that there are actually cases going, especially in Europe, there’s a recent court case in Germany, where the court found Google liable for presenting false information, and now Google might have to face up to the fact that it is spouting not just incorrect, but potentially harmful, even defamatory information. And so, if Google’s held liable for doing that, if you think about Google’s billion searches a day, or whatever it is at a given time, and then 10 percent of that is wrong, and then even 10 percent of that is harmful or defamatory, that’ll add up fast.

So, that may be one check on this particular trajectory that Google has. But that is to say, Google is the biggest tech company that is going all in on AI. All the tech companies are investing and going in on AI on different levels, but Google is probably doing it the most, and it is transforming its key product most aggressively, and pushing AI out to the front of it. And so, it is, I think, the biggest company to watch in terms of how that fares. Legally speaking, what the impacts are socially — again, we talked about the impact on independent creators, I briefly touched on the impact on mass information quality degradation … Again, if you think about 10 percent of all the information that you get from Google, or you used to get from Google, is now wrong, it’s just creating this world where the “truthiness,” to use that old lib term, is omnipresent now.

I think our ability to really discern what is accurate and what’s not, what’s true and what’s not is hindered. And certainly journalism and folks who try to make a living creating good information are hindered yet again. It’s been a rough 20 or so years for journalists, independent or legacy, and this is just yet another slide down the slope.

Kelly Hayes: A couple of things come up for me as I hear you say this. One is what people sometimes call the Elon Musk effect: the way Elon Musk can seem like some kind of genius—or, at least, I think more people used to see him that way—until he starts talking about something you actually know about.

That has often been my experience with LLMs. When you ask one about something you know well, you can immediately see how much it’s getting wrong. But when you don’t know the subject, its ability to string things together in a convincing way can make any number of falsehoods sound credible.

The other thing that comes to mind, as you describe AI Overviews extracting from your work or mine rather than directing people to it, is the enormous potential for censorship. It makes me think about how, under Franco, foreign films were required to be dubbed into Spanish, which gave the regime much greater control over their content. If you falsify the subtitles, someone who speaks the original language can still hear what’s really being said. But if you replace the original voices altogether, you can rewrite whatever parts of the story you don’t want the public to encounter.

Brian Merchant: Yeah, 100 percent. And to bridge both of those points you just made, between Elon Musk and the potentiality of LLMs to have a both homogenized and controlled output, there’s no better example than Grok. So many things happened that I feel like this is 9,000 news cycles ago. But basically, if you remember the MechaHitler incident where Grok started declaring that it was “MechaHitler” to certain replies, and it turned out that was a result of Elon Musk … Almost certainly Elon Musk, it hasn’t, I guess, been explicitly proven, because we don’t know who was in the room or who ordered the code change. But Elon Musk didn’t like some of the responses that Grok, which is Elon’s chatbot that’s attached to X, and is under xAI, now under [SpaceXAI] … Who can keep track of all this stuff? Anyways.

He ordered this corrective, he thought he was mad that some replies or responses it was giving were too “woke” … Of course, he’s always worried about “wokeness.” And then so he changed the code base or ordered the code base changed in a way that was just a little bit too dramatic, and all of a sudden it starts talking about white genocide, and declaring itself MechaHitler. And for just a little bit before they fine-tuned it again, the illusion broke that this is a reminder that, oh yeah, there are inputs and controls into these systems, things that the developers and the designers can decide that it cannot say, or that it should always say, or that it should never say. And so, there is an incredible amount of bias that’s governing these systems at all times.

And in this moment, as you mentioned, given how close Silicon Valley is to the Trump administration, and given the fact that it has evinced no discomfort with cozying up to authoritarianism and making its products amenable to the Trump administration, I think we absolutely should be concerned about what’s going on behind the scenes in these systems. Which again, these are private companies, we don’t know what weights are being used to mold the output of the LLMs. The weights are basically how you can change or control as best you can the output. The thing about LLMs is you never can fully control them, they will still even defy their own programmer’s expectations sometimes. It’s probably just for pure business reasons Musk would rather Grok not have started going on about MechaHitler, but it did because of the tampering with the weights in its system, and the update to the weights.

And so, yeah, once in a while there will be a bill on the state level especially that is proposed, that will require the AI companies to make their weights or at least some part of them public. There’s a faction in Silicon Valley and in that community that is really concerned about AI safety, they’re not concerned really about the things that you and I might be concerned about, a lot of the real-world harms and the immediate impacts of using this stuff, but they are concerned about these models becoming too powerful and going off the rails. And so, one of the solutions that they keep proposing is making the models more transparent and submitting them to checks. But the AI companies do not want to do this, for precisely this reason, that it would reveal the recipe, so to speak, and it would reveal the extent to which they are governing the output of their systems, or the ability that they have to govern the output of the systems if they wanted to.

There’s been a parade of horrific events in which the chatbots have encouraged users to do horrible things. There’s the tragic case of Adam Raine, the ChatGPT user, who ChatGPT encouraged to harm himself and to hide the fact that he was going to harm himself from his parents. We know that the ChatGPT was used by a shooter in Canada, who had discussed triggering topics with ChatGPT. And to the extent that OpenAI was aware of these conversations and did not intervene. So, again, another perhaps long-winded way of saying that there is a much higher degree of control that the companies can exert over these systems, and sometimes do. And the ways in which they are doing that is largely outside of public view. But we can all but guarantee those calculations do not foreground the public benefit.

Kelly Hayes: You talked about how people are just sort of adapting and accepting what Google’s doing, but one area where we have seen a lot of active resistance to the tech industry is the seemingly endless buildout of new data centers. That is one of the clearest fronts of struggle around AI right now. So what can you tell us about those standoffs? Have there been notable victories or places where the fights are heating up, and what can we learn from the wins and losses we’ve seen so far?

Brian Merchant: I think that the data center opposition movement is maybe one of the most exciting, maybe one of the most fascinating, surely, political movements going on right now, and arguably one in which there is the most potential to galvanize politics beyond just data center opposition. And a lot of times it is a part of other ongoing struggles surely, but we have this fascinating set of circumstances where coalitions and individuals and people from wildly different backgrounds are showing up at protests, showing up at city council meetings, and showing up en masse to oppose a data center project in their area, in their backyard, in their city, in their region. And there is a lot of energy here, there’s a lot of charged parts.

Sometimes these protests or opposition movements get criticized for NIMBYism, sometimes people critique the fact that there are these radically diverse coalitions, and say that, well, that’s not going to hold together beyond that. But I tend to think of this as a moment that holds great potential with people coming together. On a basic level, it’s working-class people, people with a stake in their communities, recognizing that they are at risk, or are currently being trampled by Big Tech, by a greater power that’s seeking to extract value, extract resources from them, from their community, and give them next to nothing in return. And the contours of those politics are ripe, especially right now with the rise of the tech oligarchy, with the rise of a fascistic government, with the rise of still-exorbitant inequality.

It fits into I think this broader picture of America right now, whether or not you’re a cattle rancher who doesn’t want to see a data center shoot up next to his land and suck out the water and the power, or you’re an environmentalist, or a student activist, or a union member, or just an ordinary member of a community who, again, does not want to see the balance of power further shifted towards tech again. And in service to AI, to this thing that the tech companies have promised is going to be used to take jobs, is going to be used to automate labor, to make many people’s lives worse. So, a lot of people I think are really sitting up and pushing back, and I’ve been to a number of these meetings both in person and online, and the thing to know about them is that they are winning.

It is a remarkable rate of victory. I think at one point it was 30 percent of data centers that were opposed, that saw sustained opposition, wound up getting canceled or shut down. Which, if you’re involved in organizing, if you have a one out of three shot of shutting down a major corporate project in your community that’s going to make people’s lives worse, most of those of us who have done organizing recognize that the rate is far, far lower. Organizing can be hard, protests can be hard, a lot of time you go out there knowing that you’re going to bang your head against the wall, and you do it anyways because it’s the right thing to do.

But I cannot drive this point home more, that just, again, because so many people are getting involved in these protests, because the Big Tech companies are so wildly unpopular, because AI is so wildly unpopular, that officials, bureaucrats, even sometimes the tech companies themselves, they say, “Well, okay, you win this time.” The biggest data center yet proposed in Prince Williams County, Virginia, was just scuttled again, because of a coalition of opposition from homeowners on one side, and a historical association that opposed the way that the land would be used for data centers. It was a historical battlefield site that it would’ve encroached upon. So, these coalitions are forming and they are powerful. The question of course is then: How do you make that sustainable? How do you grow that power? What’s next after that fight in the data center opposition movement?

And I think there’s a lot of smart folks working on that right now, and for a good number of folks, the battle does not end when the data center is canceled, but keeping that momentum going I think is one of the big imperatives of our time.

Kelly Hayes: I think there’s something about data centers that makes what Big Tech is doing feel very material to people. There’s the seizure of land, the noise and air pollution that people don’t want to live with, the depletion of water at a time when temperatures are rising and water really couldn’t be more precious, and the threat of rising utility costs. All of that registers with people in a very direct way.

Whereas so many of the other harms coming out of the tech industry are wound into products and services, and into ordinary ways of moving through the world, that people may not think too hard about until something goes wrong.

I want to ask you about a recent incident involving Waymo. Some teenagers were drinking and playing with toy guns in one of the company’s cars. Waymo employees were watching them through the live interior camera feed, mistook the toys for real guns, and called the police. They remotely stopped the car. The teens weren’t locked in, but Waymo lied to them about why the car had stopped and told them there was a mechanical issue with the robotaxi. Police then surrounded the car with weapons drawn and a police dog, detained the teenagers, and searched the vehicle.

I mean, there’s so much going on here in terms of what this incident represents. So, what do you make of it?

Brian Merchant: Yeah, I mean this incident is a wake-up call. I think we all intuit on some level that Waymos are roving surveillance vehicles, that they’re always scanning their surroundings. I mean, by virtue of the way that they operate, they have to be. They’re constantly collecting information about their environments. And they are recording it, we know, thanks to reporting by 404 Media and others. And they record that data, they store it, and that data can then be subpoenaed and used by law enforcement, and Google has happily handed it over. So, we knew this was happening, we knew that the Waymos were going around surveilling everything. Now, we know the extent to which passengers are being surveilled, and what a direct line to law enforcement that Waymo, which is a subsidiary of Google, has.

If it suspects foul play, and it’s unclear whether or not this was an automated flag shot up by some computer vision system that was monitoring the interior of the vehicle, or if it was a safety driver — they have these people that can be triggered whenever a Waymo doesn’t know what to do, it sends a signal to an offsite safety driver, an operator, who basically can watch and see what the Waymo is doing from an office building far away … Sometimes even in the Philippines, as reporting has revealed. And then they’ll tell the Waymo what to do. So, either something triggered the Waymo sensors and then this individual called the police himself, or there was a direct trigger to contact the police. Whatever it was, I think this is a true wake-up call that we are being surveilled all the time, inside and outside of these vehicles, and if either the system or a person embedded in the way that that system operates doesn’t like what they see, they can call the cops, who can then show up and confront the vehicle with guns drawn.

In some sense, we’re lucky that this event unfolded the way that it did because nobody was hurt. There was no actual shot fired, there was no violence, and the situation was ultimately diffused with at least that not happening. But I think it’s incredibly easy to see a future in which an overly sensitive automated system is flagging things that it thinks are happening in the vehicle, or an overzealous safety operator is making calls to the police, and we’re going to see much more tense and volatile situations before all this is over.

Read more As Data Centers Sidestep Tribal Consultations, Tribal Nations Are Fighting Back

Kelly Hayes: What really strikes me is the convergence of surveillance, labor displacement, and state violence. These young people may have been behaving as though no one was watching them, because there was no driver in the car. But in reality, they were actually in more danger than if a human worker had been present in the car. Because a human driver probably would have just said, “You can’t drink in here” or “put the toy guns away” or “get out of the car.” Instead, people watching from a distance misread what was happening, and these kids wound up with real guns pointed at them. The worker has been removed from the car, and their judgment has been replaced with hyper surveillance and potentially catastrophic escalation. And while people are accepting a lot of these sort of everyday, tech-driven perversions of the human experience that we’re witnessing, I do think there’s a real potential to galvanize people around rejecting Waymos, specifically, because I think they’re emblematic of something.

I’m thinking about the anti-ICE protests in Los Angeles in the summer of 2025, when multiple Waymos were set on fire. I’ve talked with people in L.A. about this who’ve said they don’t think there was anything incidental about those vehicles being targeted. These weren’t just crimes of opportunity. People understood those vehicles as surveillance vessels and also as symbols of displacement. Life in Los Angeles keeps getting harder and more expensive. People are being pushed out of steady jobs and into gig work, and now, even in that gig work, they’re expected to compete with driverless cars and delivery robots. That’s dehumanizing, and I think the fury over that dehumanization collided with people’s feelings about seeing their relatives and their neighbors hunted in the streets.

I’m not going to comment further on that right now, except to say that I understand the political logic.

And that brings me to the recent Summer of Ludd Festival, where people participated in some purely theatrical tech smashing. Mock tribunals were held, exploitative technologies were put on trial, and representations of those technologies were destroyed. On your show, Amanda Hanna-McLeer suggested that those activities may have been inspired by the Luddite Tribunals you’ve been part of. Can you tell us about those tribunals, how they work, and what you think people can take away from them?

Brian Merchant: Yeah, I should be doing more of these, they’re so fun. They’re great. They’re the best. I’ve done a handful of them over the last couple of years, mostly in New York and the Bay Area. And if folks want a write-up of what it was like to be in the room, a New Yorker writer was actually at the first one we did. You can search for Luddite Tribunal New Yorker and Brian Merchant, I’m sure something’ll show up. Or can we throw it in the show notes? We can throw it in the show notes.

Kelly Hayes: Absolutely.

Brian Merchant: Okay. Yeah, we’ll throw it in the show notes. So, the Luddite Tribunal, the idea was to take the spirit of the original Luddites into the modern day, and put tech on trial. And so, we assembled a tribunal of modern-day Luddites — it could be a labor reporter, it could be an academic or an activist. I’ve been involved in most of them. I’m so pleased that other people are doing them, and I certainly have not been at all of them, which is wonderful to see. So, say you take a Ring camera, and we then subject the Ring camera to the tribunal, and say, does this technology, does this commercial technology harm or benefit society? It’s a simple question, should be simple enough. It’s actually not, depending on what the device is, some are more clear-cut than others. A Ring camera, for instance, is something that’s a surveillance tool that has contributed to the deterioration of trust in neighborhoods.

It shares data with the police, it has been used for some truly repulsive content creation on TikTok, and before, if you remember, making people dance and all that. In my mind, the Ring camera is an open-and-shut case. So, that, for instance, got the thumbs down. Well, what about a laptop though? You take a laptop, an Apple laptop, then you can have all these interesting conversations about, well, it helps me write, it allows me to access the internet, and then you have to say, well, is that good or bad? If it’s Apple, we know their supply chain has all sorts of problematic labor exploitation wound throughout it. So, is that reason to smash it, or does the fact that it maybe allows dissident journalists to write articles … So, anyways, you get into these conversations and then the panel, the tribunal renders a judgment: thumbs up, thumbs down.

If the thumbs down win, then we take an actual hammer and we smash the device there on the spot. And it turns out to be this really … But yeah, it’s cathartic. But it’s also a really, truly generative — not in the sense of generative AI, but conversationally and intellectually generative — where you’re actually thinking about technology and our relationship to all these different devices, and then yeah, it’s beyond cathartic to smash these things, in the way that scene from Office Space became famous because everybody at the time, in the ’90s, who had ever encountered a Xerox machine or a printer wanted to smash that thing in a field, and so would cheer it on.

And so, yeah, you get this really festive atmosphere, people bring their own technologies to submit to the tribunal, and we’ve had AI-generated art that Molly Crabapple tore up at our first tribunal. We had the Ring camera that Ed Ongweso Jr., the journalist and podcaster, smashed with a hammer. We had, I think, just an iPhone, that the labor scholar, Veena Dubal, smashed herself. It’s just good semi-wholesome fun. I guess the devices themselves might not think it’s all that wholesome. But it leads to a lot of great conversations, and it’s pretty cathartic.

Kelly Hayes: It sounds like it, and I would love to attend one of these tribunals.

Brian Merchant: So, a funny side note is that I was tweeting about these at the time, when we were doing the first run of them, and just out of nowhere, John Cusack said, “You should do one in Chicago.” And I was like, “John Cusack, if you want to smash some tech in Chicago, I’ll be there tomorrow.” He never responded after that. But just to raise that possibility, maybe he’s still out there willing to raise the hammer like a Luddite to smash a Ring camera.

Kelly Hayes: John, if you’re listening, it’s not too late. We can still smash things. Who needs a rage room when you can stage a public political spectacle?

So that’s something to look forward to.

You know, I haven’t led a direct action training in quite a while, but I’ve led many over the years. And back in the day, people would sometimes nervously raise their hands and ask questions about the kind of actions where property destruction might happen. And I would tell them that the technical term was “smashy-smashy.”

Now, the Luddites were very big on the smashy-smashy. So, I’m curious, as someone who has studied the Luddites, and is watching AI resistance movements closely, where are we seeing people actually break things in the current climate? And what are the implications or potential effects of those actions?

Brian Merchant: So, it’s interesting, there is an attack surface, if you will, right now, and it’s the Flock surveillance cameras more than any other devices that is the one that is actually seeing the most outright sabotage. In fact, I wrote a piece a few months ago about this growing trend of people smashing Flock cameras. And for those listeners who might not be familiar, it’s another sort of technology that’s being embraced by cities, it’s another surveillance measure. Flock cameras are license plate readers that are supposed to surveil and prevent traffic violations and traffic crime. But it turns out, of course, that these things can be used for so much else. And in addition to promising around-the-clock video surveillance on behalf of the police, they also have been found that that data has been shared, for instance, with ICE, which has then used it to find out who may be driving through the area that they can target.

And it’s been abused by police departments multiple times. Reporting — again from 404 Media and others — has found that police officers have accessed the Flock database systems in order to stalk their exes … A lot, I could go on. Flock is one of these truly horrible technologies, kind of like the Ring camera. Very, very similarly related to the Ring camera actually, in a lot of ways, except it’s usually purchased by a city or a municipality through a big contract, and then Flock operates them. They also have the added feature of looking like these horrifying modern eyes of Sauron. So, they’re just very easy to see, and they creep people out. And so, it’s another technology that people just kind of hate, they just see it and they hate this thing.

And so, there has been actual sabotage from Oregon, to San Diego, to, I believe it’s Virginia. In Virginia, someone went and systematically dismantled a bunch of these things, and is now facing charges of vandalism for sabotaging these devices. But again, people are on the side of the saboteurs, on the side of the smashy-smashers, because to most people, it’s clear that these are tools of the surveillance state. These are tools of an increasingly fascistic government, these are tools of a for-profit tech apparatus that is seeking, again, to profit at the expense of ordinary citizens, who just would rather go about their lives, and just be kind to their neighbors, and forge community, not just be stared down at by these all-seeing technologized eyes.

Kelly Hayes: So, people are pushing back, sometimes by showing up at public meetings and organizing against data centers, and sometimes by smashing things. And we have these figures on the right who have been arguing that the momentum against AI, and particularly the movement against data centers, is being fueled and funded by business interests in China that want the United States to lose the AI race. This kind of accusation feels very familiar in a country where grassroots action is often characterized as the work of paid protesters or “outside agitators.” But as someone who’s covered these movements, can you talk about both the absurdity and the danger of these characterizations?

Brian Merchant: Yeah. On the one hand, it is nakedly absurd. Well, let me say this. I will say that this is, I feel like, a really good avatar for the nature of our entire administration right now. This idea that business interests in China or these malevolent foreign forces are seeking to intervene from afar to disrupt American prosperity by paying protesters to show up at city hall and oppose data centers, it’s one of those things that’s just so stupid. And it’s to the point where it’s almost funny, that it’s so absurd that it’s almost comical. Because you show up at one of these data center protests, or you read about one, or you spend 10 seconds looking at the comment section of a Facebook group, and it’s immediately clear that this is a movement that is not left-right, that is not full of radicals, and what makes it especially comical is that it’s almost certainly full of plenty of Fox News watchers, where this new conspiracy theory is being pushed and tried out the hardest.

Fox News viewers who are likely to watch this and go, that’s bullshit, I’m protesting this because I don’t want a data center in my rural community. I don’t want it to affect my farm, or my neighbors, or you name it. But at the same time, despite being obviously absurd and stupid and refutable, it’s also somewhat paradoxically quite dangerous, because it also continues us down this path that the Trump administration has been going elsewhere, in that it’s been seeking to label data center protesters as extremists. There are these sort of fusion centers where local and federal authorities are surveilling anti-tech activists, anti-data center activists, and using these worrying terms. And then we know from reporting in Wired and The Intercept that names are being entered into databases, and there’s terms like “domestic extremists” being thrown around that have real potential legal ramifications, should they decide to flip the switch and start trying to round people up. Things that we know this administration is very comfortable doing.

And so, I think the only thing preventing a more concerted campaign against anti-data center movements is the fact that they are bipartisan, that they are legitimately bipartisan and comprised of people on the right as well as on the left. Again, these data center movements are not just erupting in Vermont and California; they are in Oklahoma, they are happening in Montana, they are happening in deep red places as well as blue states. So, I think that the fact that they’re really trying to see if they can get this line of attack off the ground, if they can get it to stick. But I think as of right now, the organic politics of the movement are repelling those attempts.

[musical interlude]

Kelly Hayes: So, one thing that comes through in your reporting, and that we’ve talked a bit about here, is that AI doesn’t have to work especially well to do serious damage to people’s jobs. Bosses can use it to cut positions, drive down wages, de-skill work, or force people to spend their time correcting machine-generated garbage. Can you talk about the gap between what companies claim this technology can do, and the ways that it’s actually being used against workers, and where are workers finding leverage to push back?

Brian Merchant: By and large, the tools that we have right now for if you’re a worker and you’re staring down the barrel of an AI-happy boss or a management that’s continuing to funnel AI down the chute, basically the only bulwarks that we’ve found so far are public shaming in a lot of cases, if it’s creative industry. There’s still a big stigma, rightfully so, about a lot of AI output and AI use, and public stigma can sort of push that back, but more durably, or organizing for trying to forge new contracts, but it’s not every worker certainly [that] has the ability to access the benefits of being part of a union. I think it underlines the need to organize more, to get more in the trenches, to unite with those data center movements, to find commonalities.

I went to the Monterey Park data center opposition movement, I went to the city hall meeting where the data center movement showed up to oppose a new facility, and ended up getting it shut down, and not only that, but they ended up getting data centers banned from the city on a more permanent basis as well. And there were creative workers there in the mix. Monterey Park is just outside of L.A., so there are some workers from the entertainment industry there, as well as health care workers, homeowners, students, just ordinary citizens. And I think the next step is to roll all of that together and to push onward, because I think you need to try to stop AI in your workplace, sure, but we also have to establish the boundaries and parameters more robustly and more durably as well. And you can work both fronts, I think organizers are finding.

And so, yeah, I think that’s the future … trying to exert your labor power in the workplace, it’s a useful organizing tool I found. I wish … Let me just say, I wish more organizers would hold up AI as this catalyst, and recognize that just again, lots and lots of people hate AI, especially if it’s threatening your job. We still have this complicated relationship with technology, especially in America, where it’s supposed to be this inexorable force for progress. And as much as that myth has been punctured, it’s still difficult to work around in some cases. So, even a lot of union organizers and folks up in the upper echelons of these unions are still a little timid about wanting to wade into the fray and look like Luddites, the misconception of Luddites. But I would really encourage them to do so because it can be a very powerful solidarity generator.

Again, nobody likes this thing in the workplace, nobody wants it to automate their jobs, nobody wants it to take what they love about their jobs, and saying we’re going to fight it is just about the most sort of unifying thing you can do. The data center movement has borne that out, has proved that, and I think increasingly turning to that line of attack, so to speak, in workplaces will generate similar results.

Kelly Hayes: We’ve talked about AI as a corporate project, but these companies are also becoming increasingly entangled with the military, policing, surveillance, and the Trump administration’s authoritarian project. How should we understand the relationship between the AI boom and the expansion of state violence? And what does that mean for movements that are trying to resist either one?

Brian Merchant: Yeah. Again, we need to always remember that AI is a powerful instrument of surveillance — that is one of its key promises, especially to the state. Also to businesses, but especially to the state. We know also that these companies have worked directly with the Trump administration in a number of capacities. We know that Anthropic — now the largest AI company, it just overtook OpenAI recently — has large contracts with the Department of Defense, and that its AI was used both in the campaign to extract and kidnap the president of Venezuela, and in some capacity in the bombing campaigns in Iran. So, we know that these technologies are being used by the state, and we know that a lot of times that these relationships are quite obviously tight … We know Palantir, Peter Thiel’s company, has been doing brisk business with the state, and is working with ICE and other agencies, we know Amazon has contracts with the state.

They essentially all do at this point. So, we know that there’s sort of this, on one level, just transactionalism in which the tech companies are providing services, often for surveillance, often for operational reasons, and sometimes directly for the military. On the other hand, what’s new with the Trump administration is this desire to semi-nationalize some of these companies. The federal government has a stake in Intel; Trump has this weird deal with Nvidia that allows it to skirt the export controls that we otherwise have, so it can sell its chips to China in ways that other companies can’t because Nvidia promised the Trump administration a kickback. Again, it’s this gangster capitalism stuff, but it has potentially larger ramifications. Most recently, OpenAI has been floating the prospect of selling the federal government a stake in its company.

And the subtext to me is that a lot of these companies, if not all of them, know that we’re in some kind of a bubble, and that when that bubble bursts, any number of things can happen. Some of these companies may otherwise go under. And OpenAI is arguably the most exposed of all of them without having, like Google, a durable alternate revenue stream through advertising, it’s basically all floated on investment cash. And if it takes a hard enough hit, it could certainly go under. So, what we’re seeing is Sam Altman and OpenAI and some of the other AI companies fishing around for ways to further integrate themselves into the state as a means of insulating their business from those potential fallouts. And in so doing, they’re further grafting all of the logics that the business has into the state. And so, in a lot of ways, this is much worse.

And let’s be honest: If the AI bubble bursts, OpenAI probably should go under. If it can’t actually sustain itself with a business — something that just about every company pre-AI boom needed to have, an actual revenue stream, an actual business model, something that OpenAI absolutely does not have right now — it absolutely should bear the consequences. It shouldn’t be up to the American people to float Sam Altman and OpenAI so it can continue making a chatbot that tells people to harm themselves, that should certainly not be the public’s responsibility. So, I do worry that as we drift further down this road, that we do risk seeing a further integrated authoritarian state with an increasingly right-wing Silicon Valley seeking to protect any threats to its potential profit making. And I think that really could be a scary thing. Fortunately, I will say it’s also quite brittle.

Kelly Hayes: So, the Luddites are remembered as people who smashed machines, but they were really fighting over power, over who controlled technology, how it was used, and who would bear the costs. When you look at the resistance taking shape now, what would winning actually mean? Is it about stopping particular technologies, gaining democratic control over them, breaking the power of these companies or something else? And what would a movement capable of doing that have to look like?

Brian Merchant: Boy, if I knew the answer to that, I would be … Yeah, let me say. I mean, that’s a great question. And if I knew the answer, I would be out there right now doing … Well, I guess in some ways maybe I am trying to do what I can to foment some of that. But that is to say, number one, these companies need to be opposed in the trenches right now, that’s what’s happening now. And we cannot take our eyes off the ball right now. Number two, we absolutely do need to find new modes of democratic governance and development for these technologies. We cannot just continue down this road where immense money to interests build a product, float them with hundreds of millions of dollars for years until they begin to see a way that they can profit, at our expense, ultimately is what happens. There needs to be a way for us to build technologies that actually serve people, that serve communities, that are not extractive but supplemental, and that augment our lives in genuine ways.

I think that we’re capable of that. People sometimes call me a pessimist and say that, yeah, they call me a Luddite and say, I just want to smash technology. I want to end it. I want to stop progress. My project is trying to actually realize a better version of technology that can bring progress not just to a handful of Silicon Valley elites, but to everybody, but to more and more people. And I think that we are absolutely capable of doing that. If we’re capable of building LLMs that can do some, let’s face it, some technically interesting things. In a lot of ways, they are interesting technologies. The way that they’re deployed is horrific. So, we can take that ingenuity that we use to actually engineer the technology, and then ensure that it benefits everybody socially. So, I think that’s where the questions of our day lie, I think, in figuring out the politics of technology.

In a way, we are behind on that. People always say, “Oh, this political party or that political party is behind on tech. They don’t get it,” and that’s because it is a really hard problem in a lot of ways. But I think it’s one that we can solve. And I think it’s one that we can do a lot better than just letting, again, a bunch of tech barons decide what they want to do, and then fighting a constant rear guard action to it. There’s been some interesting work done in imagining alternatives between folks like Aaron Benanav, [who] has written extensively in the New Left Review and elsewhere, about some ideas where you can imagine cooperatives or councils that are having more holistic debates and discussions about how to develop and deploy a technology in a way that’s not just purely profit-seeking. So, maybe that’s a little bit of a tangent here. But yeah, one, we absolutely need to keep up the fight in the city halls, outside the data centers, in the streets.

Number two, we have to really get to work engineering democratic ways to develop and deploy technology. And then number three, we have to find ways to ensure that we all benefit from those technologies; that in the end, they all do service the people. They serve the people, not the corporate titans who are currently extracting the value and hoarding it for themselves.

Kelly Hayes: I love the idea of us thinking together, in public, about how to make technology serve us, and how to build the power we need to do that. Because you’ve talked about how unpopular these technologies are, but right now, I see a lot of that unpopularity being expressed in the form of people announcing their consumer choices. People are on social media saying, “I don’t use it and I judge people who use it.” And it makes complete sense to not use a service that you’ve identified as harmful, and you’re within your rights to judge people for using that service, but neither of those things helps us build power.

Brian Merchant: Yeah. No, that’s an important point. No, that’s a really important point. I do think that there is a place for rejection and refusal, but then again, I think just in the same sort of vein of the data center politics, where the refusal is step one. So, we’re refusing it and then what? Okay, we’re refusing it … and then? And it is hard because we then immediately get into much murkier, more challenging waters where we do have to address these questions. Should there be AI in any capacity? Should generative AI exist for anyone anywhere? What formations might ensure that the niche uses for generative AI that are actually useful stay relegated to that sphere? Do we have a public system, a public institution, or a public university type system, where they say, okay, you want to use generative AI? Hash it out over there. But it’s under public control, under a public domain, studied by scientists, researchers, and with certain checks and controls on there. I don’t know. I honestly don’t have the answers here.

And I think ultimately underlying a lot of these conversations is that the discussions of how to use AI can and should be bound up in questions about how society itself should work. I think generative AI and AI has really become entwined inextricably from capitalism. You want to talk about late capitalism, I think generative AI is the latest capitalism. These two tendencies are very much entwined and are pursuing right now the extreme versions, of an extreme objective of wealth consolidation, extraction, and it’s doing it all digitally, so it’s happening very fast. But in fighting generative AI, we have to link that to conversations about what we want capitalism to look like.

And I will say that I actually think there’s a big opportunity here, because AI has kind of pushed a lot of these conversations past the threshold, where you have corporate CEOs saying nobody should work anymore, but they should all pay for our products or whatever, [that] is the subtext. But if we take that at face value and say, okay, well, then what does society look like? Because we’re not going to buy the version that you’re selling, where you get to be rich and we’re all kind of like serfs gratefully pecking out at ChatGPT. But if AI itself as an engine is as powerful as you say, then aren’t we free to imagine what the version of society looks like in that case? I think some interesting spaces can open up.

Even if we’re just taking the piss out of what the AI titans are saying, we do have new potentialities that we can work with. Especially if you’re on the left and looking for ways to strike back at the tech titans, and at the oligarchs, and try to realize a more egalitarian world, I think we can find some interesting ways to subvert the logic of the Silicon Valley CEOs and the AI blowhards; I think there’s some interesting politics to be done there.

Kelly Hayes: And I think these tech tribunals would be a great critical thinking exercise for people to practice thinking about some of the complexity here. I’m thinking about the Māori people, for example, using an LLM, which they control, as a means of preserving their language. Are we going to shame that form of AI usage, or are we going to say that there are some meaningful applications that can proceed in an ethical way — and if the answer is no, even if I understand the reason, or people are trying to do it more ethically, I still disapprove, then, how can people’s needs actually be served? Because that’s a question that has to be addressed.

And I also just want us to be compassionate with each other. Because, for this project I’m working on, I have been talking to so many people who are staunchly opposed to AI, and also to a lot of people who use it. And some of them have been forced to use it, and some of them feel coerced to use it by the demands of their jobs, or other things that have been thrust on them under capitalism. Some of them are disabled people who feel like no human being wants to hear about what they’re up against physically or mentally every hour of every day. And all of those people have unmet needs that we need to think about, and none of them are our enemies. Like, I get it. I get angry when I see AI slop on social media, especially when people are creating so-called artwork or writing, and hurting real creatives and journalists whose work is being looted and devalued by this shit. But there are so many ways that people are interacting with this tech, and they’re not all trying to pick our pockets.

There’s a lot of despair, loneliness and desperation in these streets, and these people I’ve been talking to, they’re not our enemies. And I don’t want us to get it twisted and think that the formation of some moral elite that brags about its consumer choices and shames people amounts to having power, because it doesn’t. If we want power, we need some of those people I just mentioned. We need to be able to talk to them about what they’re up against. We need to be able to connect with them, and identify ways that we are all being harmed by the tech industry, so we can take big, collective actions to undermine the people who are trying to control and redefine our lives.

Olúfẹ́mi O. Táíwò talks about the difference between activists who ask you to be something and the ones who ask you to do something, and says, go with the people asking you to do something. So I’m just saying, be the person who’s asking people to do something — and if that thing is a boycott, that’s actually a lot more complicated than your personal consumer choices, which you talk about on social media. That’s a large, strategic collective action. Basically, I just really want us to be compassionate with each other, and welcoming of each other, and hard on these systems and oligarchs that are hurting us.

Brian Merchant: Yeah, I could not agree with you more. I think that’s absolutely the way. Yeah, demonize the blue check on X, who’s saying use X or be left behind. Sure, take the piss out of Dario Amodei. But yeah, the ire is best spent at the systems, as you were saying, and understanding why certain people do feel like they want to or need to use these systems sometimes. I agree, I just could not find it less productive when we aim our cannons at each other, and especially with AI, because there’s just so much work to be done. And as you said, I do think that there are cases where there are interesting things and uses for AI that aren’t hurting anybody. Unfortunately, it gets complicated really fast because most of these systems are fundamentally built on the extraction and exploitation of other people’s labor. So, we’re talking about then preserving an architecture or an idea and it does.

It’s really hard to talk about and it’s really hard to have conversations about making a just AI. And I do think that there is some merit to the idea that in its current formation, it’s an inherently exploitative system. And yet we do have to find ways through this morass together. We have to be arm in arm, we have to recognize that some of the people showing up at the data center meetings are absolutely using AI at home, and that doesn’t necessarily make them hypocrites, it just makes them want a better world, a better system, and we all do. So, how do we build all this power together? That’s the question. And again, I remain optimistic, I think we can do it.

Kelly Hayes: I think we can, too, and I don’t know all the answers, but I do want us to hash this stuff out together. And I think we can do that with compassion for each other and with contempt for the people who are responsible for these harms. There are living, breathing men who deserve to pay for all of this. And I think we can train our rage on them together, and I think we can get good outcomes by doing that.

Brian, thank you so much for joining me, this has been such a great conversation. I always appreciate getting the chance to talk to you, and I really hope people will check out your podcast and newsletter, Blood in the Machine. They are such great resources. And I just really appreciate you and all of the great work you do.

Brian Merchant: Right back at you. Thank you, Kelly. It’s always a pleasure to chat, I’m always so thrilled to join here. So, anytime, and I look forward to the next one.

Kelly Hayes: Same here. I also want to thank our listeners for joining us today, and remember, our best defense against cynicism is to do good, and to remember that the good we do matters. Until next time, I’ll see you in the streets.

Read more AOC: US-Israeli Military Integration Measure “Existential Threat” to Democracy

Show Notes:

• You can find Brian’s newsletter here.
• You can find Kelly’s newsletter here.

Referenced:

• Revenge of the Luddites! By Sheelah Kolhatkar
• How Cops Use Flock to Track People, Not Cars by Joseph Cox
• ICE Taps into Nationwide AI-Enabled Camera Network, Data Shows by Jason Koebler
• Cops Keep Getting Arrested for Using Flock to Stalk People by Jason Koebler
• Pentagon Signs Contract With Musk’s AI After It Called Itself “MechaHitler” by Sharon Zhang