455. The Anti-Anti-Anti-Data-Centers Social Club

It's sh*t, I can't hold it over there. In this episode, 455 of this machine kills, I'm Jathan, joined by Ed and producer Jeremy as usual. And for our episode today, we've got, I think, a couple interesting stories we want to talk through that are all kind of linked together by the theme of opposition to technology. How should you do it in what ways? Whether the most correct and leftist ways to oppose technology for what reasons? And I think, in actually a bit of a surprising turn here, we have an essay from Jacobin and we have an essay from the New Yorker. And if you just had to pick blindly, which of these is the most vehemunately opposed to technology? And which is the most kind of like third way, hey, we shouldn't oppose technology in these ways, but rather we should softly govern technology in these ways. Right, if you had to pick blindly, which is the strongest more radical argument? And which is the more a third way milk toast argument? You might not pick that Jacobin had the third way milk toast argument for opposing data centers. And New Yorker had a real fucking full-throated argument against AI in schools. So, you know, to Comrade New Yorker, I salute you. Trader Jacobin, through the struggle session with you. Interesting, right? Because I feel like, you know, we've talked about this. And our thoughts on the eco-modernist analysis of tech, but specifically the most advanced frontiers, of capitalist technologies, like, you know, renewable energies, material extraction, artificial intelligence. And, you know, the listeners in case they forgot, and what is eco-modernism? Oh, my God. What is an eco-modernism? I mean, one way to think of it is it's that we can, with enough technological progress focused on ensuring we develop our, you know, technical ability, capabilities, and productive forces as much as possible, we can make sure that humans are able to have the most advanced, most developed societies possible. And this will not have a negative impact on the environment, right? We can have an ecologically modern civilization. Think of it as we can have everything that we want, you know, we can ever kick and eat it too. I mean, that's a, maybe that's a bit simplistic a way to think about it. And there's a lot of the Tonganchi, but half serious argument presented by Lee Phillips about how, you know, in an eco-modernist world, you might want to articulate for or defend the ability of everyone to have a private jet. You know, and the argument ends up being, well, you know, everyone should have the most advanced capabilities available to them. The most convenience with which they navigate the world. The most access to resources is, you know, a bunch, a true abundance and everything. And it should not be circumscribed or un, or encumbered by any sort of external force. Because we will liberate, we will decouple, we will separate development from the historical zero-sum relationship it might have had with the environment, where the more developed something gets, the more extractive it becomes. And so, you know, in the cities becoming like, you know, Toronto houses for the earth essentially, right? Because you require all these extracts, you require inferred all hell in the terms of the minds, right? You need, the thriving masses, you need to throw at them, and you need more masses to throw at other people. And you need, you know, you need masses of people to build arguments that will destroy and steal land from one another. Land that you can also build these productive forces on and to buy products of with and just your pollutants will, you know, destroy the world. We can create, if we develop technology far enough, a society that doesn't have this sort of relationship to the environment. On its face, I will say, you know, look, I get it. You know, I think this is obviously, that is, make sense, that's something that we want. The ego modernists, we disagree with them in the extent to which sometimes this feels like this is an argument that converges with or dovetails with the capitalist rhetoric, which is that, you know, listen, this is an amazing system. It can provide for people even more if only we were able to hand off more of its infrastructure, more of everyone's individual creative energies, more of their time to capitalism to solve all the various problems. Of course, you know, you know, you know, you come on, this will be a first to tell you, especially if they're socialists that they're not doing this. But I think this piece is an important one to tack on to because it's from Holly Buck. Holly Buck, you know, I read her work in Jacobean before, followed her work also, she's a professor, you know, University of Buffalo. That's a lot of interesting stuff from sustainability. You know, the question here that she's trying to ask, right, is, how are we going to make, how are we going to make an intervention in the development of artificial intelligence that will reoriented towards a purpose that benefits us. Right now it's being deployed for destructive monstrous, potentially world historical purposes that might close a bunch of political horizons for us, might open a door for devastating environment, might devastate people's minds, might devastate, you know, people societies, how do we steer to development of artificial intelligence away from that. And uses the, the, the, the, the March towards data center moratoriums to say, look, like maybe the solution isn't we ban data centers and their proliferation, but that we figured out how to equitably distribute access to them. Now, you know, on tmk, I feel like I have at various points either said, you know, we should ban them all or maybe if we don't, I, you know, and we'll talk about this more, you know, especially ever since the debate that more is of and better and whether we had I've been thinking about this and in my head, a more to him is useful for reasons for certain reasons, maybe it's not useful for other reasons, but her argument that we can develop democratic ways of governing AI, I agree with, I find a lot wanting because when we actually dig into what's being articulated here, I think that there are a bunch of assumptions. And I think there are a bunch of shortfalls and the argument that end up leading us to a place where when we might say, okay, we of course want to democratize artificial intelligence is going to be governed, but we don't actually, you know, have much in the way of real tools or aspirational political visions or radical breakaway points or tactical suggestions that will get us to a different political economy than one that we have now with like, well, you and you know, representing you or something like that, I mean, you know, as we'll go through is a little bit more than that, but I think that the ego modernist position we can think of as capitalist productive forces in line with Marx, some of Marxist thinking can be developed to hire can be made to sing at a higher level that allows them to yield more abundance and more material comfort to people without trading. Trading off the environment, whether or not that is true is another question, right, but that is the kind of core thrust the core move, I would say that the ego modernist are making and assuming yeah, and I think in inherent to an ego modernist position is as you've been saying this kind of like. The only way out is through right the only way out of climate change is through you know, accelerating technologies that will solve climate change right and so like the you know, an ego modernist position is very much show you. Premist on like, you know, a belief in techno optimism in a lot of ways right that like it gives the gives technology the benefit of the doubt right and I think in a lot of ways it can be a little disconnected from the plague economy of that technology that politics of that technology to various ways but it comes from a position of giving technology the benefit of the doubt having a kind of optimist view and also pointing to being like no like what we need is things like, you know, I mean, you know hard line. You know hard line the ecom modernist would you know point to things like geoengineering right as a solution now, Holly book is written a book about geo engineering right and I think she has fallen, you know, the author of this Jacob MP switch for listeners is called democratic. the product governance of AI is the real solution. We'll have a link to it in the description, of course. Well, I think what's funny is that like, I know in some of Holly's previous work, she's been in microgeoengineering work, she's been critical of eco-modernism for kind of overlooking the exploitative and extractive nature of a lot of technological modernization. But those critiques are not present in this article, right? And so it's like, you can't give someone the benefit of having a position in the past when their current, when their present position is not representing that, right? And so like, the basic assumption here at the Bay, in this article is that it's opposing two things. One on one hand is opposition to data centers. And on the other hand is, you know, a gesture towards democratic oversight of how AI develops or democratic governance of AI. You know, and those two things are held up as opposed to each other, right? Not things that like one may lead into the other, but right now they're, but in this article, they're kind of, they're put up as like two separate and opposing positions that one could have on AI. Either you wanna oppose it and stop it or you want to build democratic oversight of how it develops, right? And so the assumption here is that it's either a, it's a techno negative argument, right? To put it in the language of Thomas Deck-Cellar's new book. It's a techno negative argument or it's a techno positive argument, right? And the techno negative says, stop it and it is, in the essay, you know, that's NIMBYism, right? Like she explicitly in the second paragraph says these efforts to use the power and machinery of familiar NIMBY or not in my backyard politics, local opposition, tying up projects and red tape and so on. To confront the multiple perceived threats of galloping, energy demand, carbon emissions and job loss. So on one hand, you have a techno negative NIMBY position, stop data centers. On the other hand, you have a techno positive democratic governance position. Let's just, let's have better oversight of how AI develops, right? But that doesn't, but that doesn't mean anything about stopping it. And in fact, as we'll get into the position of Jacobin in this piece is that actually, we should be actively expanding the development of AI. We should be expanding access to AI. It's very much set up as she calls it, right? A counterintuitive position because a moratorium quote, "A moratorium on AI data centers is a terrible idea, one that poses serious equity concerns. A moratorium springs from the desire to stop the concentration of wealth. But ironically, it is likely to exacerbate it. It's a massive strategic blunder for the left. And we should think through the global justice implications and follow on effects." And then so she kind of then wants to walk us through what she sees as these global justice implications and follow on or spill over effects of a moratorium. And this is where the piece like immediately just loses me. 'Cause I'm like, the arguments that I put forward here are reaching and strange at best. I think they really come from a position of saying, how can I oppose the moratorium on data centers and then building an essay around a position, right? So in other words, rather than coming to the thing with an analysis and seeing where the conclusions, the analysis takes you to, it's coming to the thing with a conclusion and then trying to build an argument or analysis around that position. And I think when you do it in that, and it's a little, it's evident to me that when you do it in that way, you end up reaching for whatever you can find to help support your pre-descend conclusion. And you do so, there's a tendency to do so in ways that are a bit pretzeling, right? You're twisting things and knots, you're bending over backwards, you're intersecting and interloping things that maybe shouldn't actually be intersected and interlumped with each other all as a way to, say, look, I've reached the conclusion that I have set out in a pre-descend way to reach. And so the kicking off here around like the first argument against a moratorium is that it's one about off-shoring burdens, right? So it's essentially saying that we should be wary as Holly Buck writes quote of proposals that would send burdens elsewhere. And then she, you know, said, well, you know, under neoliberal capitalism, industries offshore these environmental harms to places with weaker governance, cheaper labor costs, fewer environmental safeguards. In other words, the undeveloped or developing world, the global south. And so she, you know, it says, well, if we stop building AI data centers in the US, you know, place a place that has stronger governance, you know, higher labor costs, more environmental safeguards. I mean, TBD on how true and how that is now and how true that will be in the future, right? Because the point of building a lot of these data centers in the US is that they are dependent upon the rolling back of governance and safeguards and things like that in the US. But let's just take the point, right? That like if you stop ban, if you ban the building of data centers in places where the data centers can be built more fairly and cleanly, then those data centers are just going to be built in places where they are built more exploitatively, extractively and in an environmentally dirty way, right? So the assumption here, and hey, maybe there's something to it, right? That like there's a global demand for data centers. And if you shut off the US as a place that can supply those data centers, that demand is just going to look elsewhere and that's going to off shore the burdens and may actually, you know, result in bigger impacts to environmentally and socially. So, Ergo, we should actually be welcoming data centers in the US because at least we can do it the right way or a better way than a place like the UAE or Southeast Asia or Africa. I mean, that's essentially the argument she's making here, right? About like one of the global inequities of the data center moratorium in the US. - That first section, right? I think it's also interesting, right? Because there are a few things that it is, you know, that it is trying to do that I are moving at different registers, right? Because there's one part of the argument that's trying to go, you know, like you said, this off shoreing, which is like, okay, we're going to, more Torians are just going to redistribute. They're not actually going to pause development, but I think this is a bit slippery because it suddenly shifts from, you know, like the argument was once moratorium, you know, they're definitely not going to halt development, moratorium are going to off shore harms. These are two very different things, right? In one scenario, you know, if we're being concrete about what we're saying in one scenario, what happens? A moratorium, a moratorium in the United States is going to push hyperscaler buildouts to places with these weaker jurisdictions that you were talking about. And, and, and, you know, the idea then is like, okay, we're just going to have a, you know, this is going to be the off shoreing pattern that we had for manufacturing, right? Where the benefits, you know, flow back maybe, we'll get compute access, we get AI services. And in this other scenario, right? The moratoriums are going to slow down the global pace of development and the global pace of the AI buildout because America accounts for the vast majority. Of frontier lab, AI compute capital expenditures, right? So, off shoreing is partial. And, you know, in part of it is coming maybe from grid constraints, part of, you know, and outside the United States, part of it is maybe from grid constraints. So, export controls or, you know, geographical stickiness, they're going to be having from these operations, but that off shoreing is, you know, partial in this scenario where the United States still exists is most of the, most of the, most of the site of capital expenditures, right? These are kind of, bulk kind of marriages, these, or we're bugging saying, okay, look, there's some constraints on off shoreing, right? Right? Which is that, you know, some places, they've got wheat grids, some places, have chip exports, you know, restrictions. And then says, well, you know, like those constraints are not set in stone. But the constraints, of course, they're not set in stone, but they are set, right? They are, not accidental, right? They are consequences and they're downstream of very specific geopolitical and industrial configurations, right? And to change them, right, to get different chip export rules to loosen them, to build grids in the Gulf, right? To, you know, start working on how we're going to implement power provision for some of these, you know, data centers. That's going to take, of course, is that will take years. They'll take a lot of capital. That itself will become a pressure point. I mean, this is kind of what I was talking about with, you know, and what we've talked about with Paxillica, right? The point of Paxillica is that these, it's difficult. It's politically costly to construct this offshore architecture. It's not just that we're interested in inking bilateral deals, is that the bilateral deals are imperial architecture. They're being negotiated with very specific entanglements such as Gulf, sovereign capital, such as, you know, United States military security guarantees, such as, you know, infrastructure for, you know, moving dollars around in the global financial system. Here, you know, she, you know, she sites, you know, the US UAE example. But this is, I think it goes more against her point than she might realize where the framing is that this is, of course, like the USA UAE relationship is evidence that clearly, you know, companies are just going to go around the moratorium. I don't actually think that's what it's evidence of. This is, you know, evidence that the offshoring options are fundamentally entangled with the geopolitical questions of how to build an architecture that will sustain and regenerate to rejuvenate American hagemony. It's not simply a substitution. You have, it's a geopolitical commitment. And it has its own contradictions. It has its own constraints. You can't just simply, you just can't simply frame it as this is something that is going to be substituted. I think there's, you know, there's another argument that is in here where it's just kind of the idea that we have compute scarcity is going to lead to higher prices, which is going to lead to smaller users getting priced out, which is going to lead to incumbents capturing the access that remains, which is going to lead to this digital divide that will widen. And I feel like that's a causable chain that I think leads on, leans on, but we can contest that each step. You know, so we have this first section where we have compute scarcity. And the idea here is we're going to assume that for compute scarcity, the thing that determines it that it hinges on is consumer pricing and not enterprise margin or not the capital expenditures for any part of training the model before you actually get it, right? Is that true? You know, what are the actual binding constraints right now on consumer AI access? Is it compute availability? I don't think so. I would say actually the real binding constraint is that there are business model decisions about what subscription pricing is going to look like. What services are going to be available at the free tier? How are we going to limit the rates available to individuals and to firms that are interested in using our product? And then pick an open AI. I mean, especially now people are insisting on a topic might be running up on a scarcity of catbacks and open AI, but they have plenty of compute. And they have enough compute to serve all these free users, but they don't. Why? Because it's a unit economic. I mean, Google just announced like a tens of billions of dollar investment into anthropics. Yeah. I mean, yeah. So what is the problem is, you know, matching Amazon, right? Because Amazon just did 40 billion into open AI. And so Google is like, well, we're going to do 40 billion into anthropic. And it's like, you could you could do it. You could service them. But why would you? Because the unit economics don't work at that pricing, right? A moratorium. What is it actually going to do? A moratorium that is constraining the rate of compute expansion. Yeah, maybe it might affect training cycles for next generations frontier AI lab models. What is it actually going to do anything like what's what's what's underlined the relationship between that and what people throughout the world are paying to access cloud? You know, the that paid the relationship between, you know, this and what students are paying for that. It's mediated by pricing decisions that are made on other grounds entirely that you know, that are made independent a lot of times under what the compute scarcity is. And if there were compute cloud, what is there is the idea here that the pricing is perfectly rational? I don't know. And then the next trunch where it's like the idea that small businesses, that academics at nonprofits, as she says, you know, are accessing these frontier AI at a low marginal cost. And they're going to lose that access under tighter compute. Probably not. I mean, the academic APIs access. I mean, you told me, Jayden, you got a university. Is it, is it rate limited? You know, is it, is it, is it not rate limited? Is it cheap? Yeah. It's also like these are deals that are struck by universities, right? I know. We're from major financial institutions, right? And so it's like the, like, you know, the idea here that academics and not, but also the rate limited, subsidized companies have forever limited. Yeah. Companies have forever placed limits and constraints. I mean, I have a, I have a piece in nature with Salona, Villan, and Meredith Whitaker about this exact thing that like tech companies have always used their monopoly over data and compute to exert forms of soft and hard control over what kinds of research happen and what ways, right? That exert asymmetrical relationship to academics and universities, right? Like, this is not new. And like, you know, a tighter restrictions on compute through, you know, limiting the build out of data centers is not going to like suddenly change the unit economics or the power, the political economy, right? The power relations between the suppliers and owners of the thing and the people who are, you know, renters and users of the thing. Like, I think fundamentally what this is based on is, you know, there's a lot of a misunderstanding of the political economy of the AI value chain. Yeah, because there's kind of a presentation as if like frontier AI is already democratized, you know, right? Exactly. And it's going to be de-democratized by more, more Toriums with it. That doesn't make sense. Again, you know, the academics that we're talking about here, it, it find me an academic partnership with a frontier AI lab that is not heavily subsidized, but it's not, there's not subject to rate limits, right? And it's not, it's not part of some reputational package or ploy by these firms. It just, it doesn't, it doesn't make any sense to look at this as something to, to preserve. And then it also, you know, and if that's the case, then this third level where we're saying like incumbents, I'm just going to buy remaining access. You know, yeah, okay. I mean, that might be plausible, but that kind of, that also undermines her argument, right? And then you have incumbents always buying access, no matter what the price is, then isn't the concern with more Torium that you, you know, have marginal access for non-incumbenced? And this is governed by pricing decisions, by access programs that public policy can direct. So it isn't then the implication we don't really need to be concerned with the compute constraints is that we have to actually constrain compute and mandate its access to provisions for non-incumbenced users, right? Isn't that the, isn't that the conclusion you would actually follow if the idea is that incumbents will always buy it at any price? If that's the case, you have to limit it. You have to limit their ability to buy it at any price because they're, they're doing that, if they're doing that right now, that's something we should be regulating. And if they're doing it later, that's something we should also regulate, right? It's not, it doesn't make clear to me that we should try to gin up a market environment in which they don't because the implication here is they always do. And unless we want, again, unless we want to, you know, go big. But if we go big, then what are we talking about? If we go big, then doesn't this undermine the idea of a moratorium? It's because it's just, it suggests we probably should do a big, musely move that it seriously constrain compute capacity like banning or pausing or slowing down compute expansion, right? And all of this, like, this ideas that these ideas in this section, where we're talking about how the incumbents are who we need to kind of be concerned with because we're trying to develop their productive capacity or technical capacity at the frontier so that everyone can benefit from trickle down effects, you know, and it treats it like it's some scarce good, it's distribution is, is, is really supply signed. But I think it reveals the number of assumptions at the heart of an eco-modernist argument. This all just sounds like Uber to me. I mean, just like, I mean, you would have ran a great piece about this, yeah. You know, but like, I'm also thinking about it in the sense to like, I stopped really kind of following this stuff closely because it just, I don't have a lot of hair to pull out anymore and I'm tired of pulling out what I do have. That's why I'm called. But what I'm saying, like, you know, it looks like Uber, to me in the sense of like they want market capture, right? They want to be able to take in all the compute that's available so that you have to rent it out so people can't own it anymore. And then tokens are going to become so expensive. It'll be cheaper to just hire people to be your AI agents. And because people are going to be so in need of work, they're going to be willing to accept that work at lower and lower prices to be exploited. It just sounds like Uber all over again. I think the analogy that Uber is a really interesting and app one because that was also a lot of the same ways that people advocated for Uber from the left, right? Or at least like nominally from the left was like, you know, saying, well, this is democratizing access to a necessary service, you know, this is, hey, we owe what if, you know, people don't have money to pay for taxis or there's not a taxis, you know, there's not taxis services in their area. What if they're, you know, it's also, you know, with the platforms, a lot of times that would happen was a regular like this, you know, willing out, you know, pun intended, willing out disabled people as a shield to, you know, say, well, if you want to attack Uber or taskrabbit or, you know, what, you know, one of these platforms, then you're actually attacking the disabled people who rely on those platforms for transportation or access to meals and stuff. And so it creates this kind of like the exploitation is worthwhile, the extraction is worthwhile because it means that this, this, this group gets access to a service, right? When they're in reality, like, that's never the purpose of the service, right? That's always like a trickle down benefit for these services, which, you know, Uber and Facebook and everyone is like happy to, to use as a shield against any kind of criticisms, right? They, they get the equity shield out whenever they need to use it, but they don't design systems for with that in mind. But also it points to a like, yeah, but like, the, the problem here is not this or nothing. The problem here is not this and something else, right? Like, that's the solution. And so I think the, this gets to the heart of a lot of assumptions at the core of an eco modernist to take, which is that there's an assumption of, of inevitability, unstoppable ability and autonomy of technological modernization, right? Technological modernization is going to happen whether you like it or not. It's inevitable. It's unstoppable and it's autonomous. But if we channel it and direct it in the right ways, it can be beneficial to us. But also the assumption here is always that wider access and wider use of these technologies is a good thing, right? That, that's also another core assumption that of the kind of eco modernists meets techno optimists. A argument here is that it's inevitable. It's unstoppable. And it's autonomous. And the best we can do is try to direct where the train is going. But we cannot stop or change the train. And also the more people that get on board the train, the better, right? Like that, that's the argument here. And I think the train metaphor is the right one because it's like, that's also why there's this basic assumption that there's an opposition between moratorium on data centers, techno negative and a democratic governance of AI is development, techno positive. Think of it like a train. It's saying, we cannot stop the train. And if you try, but what we can do is direct the where the train is heading, right? It's somewhat. We can throw some switches and change tracks a little bit. But if you try to stop the train, then there's no way to direct where the train is going because the train stops moving. And so therefore you shouldn't stop the train because we actually like that the train is moving. We just want to change where it's going slightly. So it's the same thing with AI. It's the same thing with, you know, same arguments with platforms, right? It's the same arguments with these technologies is do not throw the emergency break, but instead throw, you know, hit the switch to change the track a little bit. As if these things were like mutually opposed to each other. And because again, it comes down to this assumption that this is inevitable. It's autonomous, but also it's progressive. It's a it's a progressive force. If if we can just like exert a little bit of guidance and direction on that progressive force. And but I think that completely misunderstands as we're saying, they're like the geopolitics and the political economy of AI supply chains. What's the point of Paxilica, right? The point of Paxilica is not to say this is well, the rhetoric of Paxilica is to say this is inevitable and there's nothing you can do about it. But if that was the case, you wouldn't fucking need Paxilica, right? Because it would just happen whether you want it to or not, whether you like it or not. The point of Paxilica is a recognition that this is a fragile thing in the making. And we need to create a coalition of the willing to make sure that it happens and that it happens in a way that we the US, you know, hegemonic sphere has control over, right? In other words, the point of Paxilica is to say that like actually all these things that this article is assuming is going to happen whether you like it or not, right? It assumes that there is a a static and immutable amount of like AI demand and supply. And the only thing you can do is choose where that demand and supply is located. It assumes that like the progression of these frontier models is inevitable. It is, you know, all these things, the article assumes the existence of something like Paxilica, the existence of something like the US UIE partnership is evidence that these things are not inevitable, but are instead highly, highly, highly socially constructed and also highly fragile geopolitical partnerships that depend on a lot of fucking power and money going into keeping this thing from popping, right? Keep as we've talked about before on the episode on the show, it's a game of musical chairs and there's a lot of fucking money going into making sure the music never stops. Because if the music stops, then the whole thing falls apart. Everyone realizes there's not enough chairs for everyone to sit in, right? And that's so the assumption of this article is that like, there's these immutable, eternal dynamics of neoliberal capitalism that there's a, you know, a static supply and demand for AI and all you can do is move pieces around on the board, rather than like change the shape of those pieces or take some of them off the board or recognize that the board itself is a highly fragile and flexible environment. Like all of these things are actually true about the geopolitical nature and political economy of the AI value chain. And so it troubles an analysis that is trying to talk about equity and talk about development, but do so from this position that you can't change any of the parameters, you can only change how you respond to the parameters. And this goes to like the piece continues by talking about like, well, if you limit the, you know, that access to like US-based models, you know, what people are going to do, they're going to go use Chinese models like Deep Sweet and Quinn. And you know what, that's bad because they have different, quote, "approaches to data privacy protocols, human rights and economics." Right? So, I mean, there's an assumption here as well that there's like, there's a universal and static demand for AI usage rather than something rather than that demand being a highly manufactured product, right? Like demand for AI is not organic. Demand for AI is like a totally synthetic thing. It is manufactured by the suppliers of AI. But the assumption that there's a static demand is that like, well, if you limit access to American AI models, people are going to go use Chinese AI models. And that's bad for various reasons, right? And so again, better to keep them at home. Keep them using the US models that are at least, you know, have a more American sensibility baked into them. It's another kind of bizarre argument to make here. I mean, don't want to get into the whole like, you know, she also buys too much into the cloud, mythos, myth, right? That like, did you like anthropic is not releasing this model because it's so dangerous? And that shows that we need a world where access to these models is not stopped or hidden. You know, I don't know. Like it, it seems, it feels like waiting into a debate. where you're like, you're understanding of the terms is based on what the companies are telling you. - It's like on some level, right? We can get it or understand, because the move on the basic levels look like tech corners of our internet are freaked out about this new model. They say it has advanced cybersecurity capabilities and the topics withheld it, and they withheld it from a public release. They gave some access. Two folks to help patch vulnerabilities. Experts deemed the capabilities real. And so the world that we see now where adversarial agents or actors can develop such capabilities, that's dangerous. And so US development getting paused is also dangerous, right? But yeah, there's a bunch, I think there's a bunch of things going here. One is like, you know, the alarm itself is not evidence, right? Even though, you know, luck insists as much. It's a very, like we've, for years now, this has been a clear, common marketing of rhetorical move and that has been made by these firms, right? That, regularly, especially over model releases, there's a lot of noise generated by who? By the individuals and firms who have a strong financial and ideological incentive and investments and insisting that capability claims are a little bit stronger than what they actually are and that they're taken seriously as such. So citing the alarm as evidence is, would we like take what pharmaceutical firms say about their new drugs completely at face value without scrutiny, of course, enough, right? We might, you know, of course, their statements you can and can't make depending on if you're a private and public company, but like we would not immediately believe a firm saying that its product is the best thing and that it's the, and it's going to solve all these problems or that it's gonna be dangerous thing or that, you know, well, I mean, you know, to mirror orthopedics and it's incredibly dangerous thing. A farm of drug being marketed as the safest thing, right? You know, alarm is not the same as what the actual evidence on the ground is. What's the evidence on the ground, right? The assumption here is that anthropics release decision, this is face, we should take it at face value. This is evidence of capability. I think, you know, we just said, look, like we just, went through a version of this with the anthropic pentagon debacle. We're supposed to go and look at a lab and say, their read of the dangers accurate. We have to trust that the limited release is responsibility and not a business strategy. We have to believe that the consortium access they're offering is a governance solution and not a way to keep capability claims, you know, limited to individuals and relationships that want to preserve them and affirm them, right? These are a bunch of things to trust. Whereas, you know, as we saw in the pentagon, there were real concerns that were being raised by anthropic, for example, about commercially available data being integrated in surveillance and about autonomous weaponry, but also at the same time, anthropic is still building the device that it claims to have these concerns about and is on some level because it is both and some dimensions more, you know, not really a threat to fundamental beliefs that they have about what we shouldn't have and they're fine with proliferation of these tools. And also on another level because they may not be as dangerous or disruptive as they claim that they are, right? And that, or maybe a better way to put it is that, you know, in the case of autonomous weaponry, four years we have had kill chains that are incredibly compressed and relying on little to no human supervision, right? And this has been used as a cultural to learn about how this is around the corner and this is why we need to invest in anthropics vision for how artificial intelligence is going to be integrated to military chain. And while they've been accelerating and driving us towards the sort of future that they're warning about, but you only get their future. - This is a classic case of, you know, pointing, point using, you know, using alarm about a problem, whether true, whether overblown or not, but using that alarm about a problem has evidence that the only way out is through, right? It's the eco-modernist logic again. They point to, I think rightfully, concerns about climate change and then they say, well, the only way out of climate change is by pushing the development of technology further, accelerating it faster, deploying it more widely, right? That will solve the problem. So they use the alarm as evidence for accelerationist arguments. And so it's no surprise that the same eco-modernist logic is being applied here to AI. The alarm is used as evidence for accelerationist arguments about the need to develop AI faster and better. And I think this gets us to the next kind of like tranche of arguments here. And then we can, you know, kind of use this as a way to wrap up our discussion of, of, because in classic team K fashion, we're going like an hour on this, on this article. But I do think it's worth spending the vast majority of our time on this article. I want to point to the New Yorker one at the end of the episode. It's just the kind of like, hey, this was a really interesting piece. We really enjoyed it, unexpected as well. But I think this is a good kind of like struggle session episode. We spend more time, you know, critiquing the comrades, being like, guys, you're getting it wrong here, right? But the next tranche of our arguments about the data center moratorium is on one hand, one that I, I initially resonate with the most, which is that, you know, big pockets of data center opposition are rhyming with NIMBY politics, right? Like there is a bit of like, how much of the data center opposition that we're seeing in some places, you know, like in the suburbs of Northern Virginia and so on, how much of that is actually premised in a like, like a political economy of technology, probably not a whole lot. Some of it is for sure, or just a kind of like more leftists, like, you know, critique of big tech companies. I think that's a lot more of it than like a deeper political economy of AI. But there is also a lot of it that is a kind of a developmental politics, right? It's a NIMBYism. It's a like, I don't want this thing in my backyard because it's like gonna make the air quality worse. It's gonna, you know, it causes a huge amount of noise and traffic. It's bad land use policy, right? Like it's gonna affect our property values. Like there's definitely parts of it that are there. I think it's an open and empirical question that Holly Buck admits in the essay that they don't really know. So there's a lot of assumptions going on here, but how much of NIMBYism is the core of data center opposition in some places. But the article kind of takes as the assumption that that data center opposition is NIMBYism and then advances from there and article and article and argument that I think is a huge threat. This is like one of the more, this is even more pretzeling than the misunderstandings of like geopolitically economy of AI value change, which is to say that if the, you know, because data center opposition is just NIMBY politics of affluent middle class and upper class suburbanites, therefore data opposition is class warfare, but not in the way that you think it is, right? It is class warfare of the middle and upper class affluence of urbanites against the lower classes who because of data center resistance, you know, here I'll quote from them, the class-pilot particulars matter. What if the picture that emerges of data center resistance is one of educated middle class people, including ex-Zerban and rural residents, but also professionals who work in knowledge jobs, mobilizing consciously or not to protect their class positions from the threat AI poses. How many of these people will block data centers, but end up paying for a subscription to a frontier model once it is clear how useful it is to navigate daily work in life. It's not fair for affluent environmentalists and property owners to try to stop development of this infrastructure before most people in the world have even had a chance to work with and learn from these models. That is the point at which I'm like, huh? Well, so the argument here is that this is class warfare of upper middle class affluence sub-urbanites against the majority world. This is upper class people saying, I'm protecting my own class interest as a knowledge worker and by banning data centers and therefore preventing the majority world from ever having access to frontier models. This is the most pretzily argument that I can imagine for against the opposition to data centers. So many assumptions built in here. What if this happens and then this happens and then this happens and now people in Africa and Southeast Asia, they don't get access to the latest Gemini model and their life is worse for it. I guess you're just a your ANOVA supremacist. You believe Northern Virginia nights should rule the world. Yeah, like we're now for the end and like CK three or something. I'm building out my I'm building out a nice beautiful compute to Republic and the in Virginia. Virginia is for Virginia is not for lovers. Virginia is for for users for compute. No, Virginia is for the noters who want to prevent the rest of the world from being users. Right. Yeah, yeah, yeah, yeah, yeah, yeah, yeah, take them to Norfolk and we'll see how that goes. But you know, it's it's on some level I'm interested in this class competition argument, right? This idea that it is a class warfare argument or something that's going on this question that you that you raise right where it's like, OK, well, how much is an imbi's? When should people preserving themselves? But I don't think that OK, so like the section does yeah, of course, there's some mindful and participants. There's some cross ideal ideological solidarity there real AI access my Heism distributed effects. But I you know, like we were talking about here, the section doesn't really make it clear. Doesn't really argue or establish. I think satisfactorily that resistance is class protected, right? And and and the essay itself, it doesn't really. It's all assumptions and it nor does it, nor does it establish that like it all just says what if wide scale worldwide access to frontier models is good, we do, but we'll never know unless unless we roll them out to the whole world and then see how the see how the dice roll, right? It's it's such a strange argument to say because again, it is fundamentally a techno optimist argument that says we don't actually what if sure there might be consequences, but what if there what if utopia's on the horizon and your smud and your smothering it, we don't actually know until we do it. And so it's an argument in favor of again, it's a techno positive. You should do the thing, see how the cards fall and then make decisions after rather than doing anything that's precautionary or proactive. And it's worth noting these are very big claims, but they're they're not really defended with evidence so much as rhetoric, right? You know, is what is the evidence that the resistance is class protective? What is the evidence that AI access is largely determined by US compute capacity? What is the evidence that subscription based access is our equity question that we should hinge the debates on public policy about? What is the evidence that opposition linked to affluent, you know, participants is going to be the binding constraint on global access? I mean, there's not really much evidence or articulated for these. And I think this also kind of gets to the deep one of the deepest weaknesses of this action, right, which is like the political economy is obscured, you know, what are the firms? What role do firms making pricing decisions play here? What role do regulatory frameworks play in shaping access? What role do you, you know, public institutions could up public institutions play, right? And the provision of compute. What role do labor markets play in determining who needs access? What role could global infrastructures or global political economies in the AI value chain or in the distribution of AI play in determining this? All that is talked about is the AI build out and the infrastructure. And, you know, partly that's because, you know, this is a tradition of this like, you know, Eagle monitors or, you know, the maybe buck might dispute that. They're a liberal technocratic tradition, you know, also, I don't, you know, know whether or not buck with concern or stuff. That. But, I mean, this is the sort of question where it's like, we're concerned with the infrastructure because the infrastructure is the platform under which we can, you know, ensure that outcomes are what we want them to be. All there is is this build out, you know, affluent workers or owners who are interested in blocking it and non-subscribers who deserve it, right? What? This is not really a good material vision, you know, and it doesn't also help us figure out what to actually do because if without the political economy, this doesn't really tell us what kind of interventions we should make to on the build out itself, right? Okay, let's say we do want to, you know, ensure that the build out is continues. Is it is, is it hard to imagine there's any partner who we might want to constrain? Is it hard to imagine there are specific outcomes, paths, you know, developments we might want to pursue or might want to reallocate capital towards or might want to reallocate, you know, talent towards, you know, these are questions that I think are, you know, are left unanswered to the detriment of the essay because they are interesting questions. I mean, I want to know and I want to think about it. I want to talk about, you know, what is democratic governance going to look like? But again, we have to be clear, you know, this piece is not really talking about that. No, the piece uses democratic governance in the way that so many leftist articles do, which is as a, as a conclusionary gesture. Yeah. And, you know, to solve this problem, democratic governance period, you know, Q, Q, E, yeah. Like, you know, there's no actual like material vision or particulars. And you can say, well, you know, it's a Jacobin essay. And so you can only do so much and stuff. It's like, yeah, but like you, that was the whole premise of the article was setting up two points in opposition to each other. But then the article just goes on to give, you know, strange, disingenuous, and at times interesting arguments against one point. And then ends by saying, and the solution that we should be striving for is this other point, which I will know, which I will not detail at this time, you know, that disingenuousness of the, I think you're exactly right where it's like the argument is made by rhetoric, not evidence here. And you see this in the disingenuousness of one thing that really struck me as like, oh, come on now is on one saying in one paragraph here, there is a narrative of vulnerable community standing up to big industries illustrated by cases that do not fit this model like XAI's exploitation and air pollution at its South Memphis data center. However, the reality on the ground and then the anti data center Facebook groups points to a more complicated picture, AI data centers face a left right opposition as likely to post gatson flags as in this house yard signs. So on one hand, they say, there's this narrative of vulnerable community standing up to these data centers. But if you look at Facebook groups, it's a lot more nuanced than complex. There's left, there's left and right wing people who are opposed to data centers. And then they go on to say in the very next paragraph, the stuff about like, it's all affluent, exerb and professional knowledge worker is opposing. So it's really disingenuous. So it's not the vulnerable communities are just a narrative. The politics are more of a left right joint opposition that you see in Facebook groups. But it's actually all just upper class affluent suburban knowledge workers. Like you can't have every position at one time, right? You can't say that's a narrative, but this is not right. That's complex, but this is, but it's also not complex, right? No. It shows a disingenuousness that goes back to what I said before. You've established a point that a conclusion that you're trying to reach and then you do anything that you can to reach that conclusion. So you end up with sophist, sophist, you end up with rhetoric, right? Because also like saying that the, the narrative of, we're saying that vulnerable community standing up to AI data centers is a narrative is also not a very leftist position to take here. When we have a lot of fucking examples of it, it's not just Memphis, it's also Indiana, it's also, you know, the super fun sites in, in the Bay Area. It's a super fun sites in Indianapolis. It's a super fun sites in Memphis, right? It's a super fun sites in Arizona, right? Like are you saying all of these places that are historically places of environmental racism, deep structural systemic, impactful environmental racism? That's just a coincidence. Those are also places where AI data centers are being erected, right? And that those vulnerable communities are actually, well, you know what, their opposition, doesn't really matter or it's more complicated because they're not all ideologically peer socialist Marxist Luddites, you know? The reality of everything in the world is that most people's opinions about politics are a piecemeal, incoherent mishmash of stuff that they've picked up from various places. This crediting their positions and their experiences doesn't just because of that, because they're not ideologically peer is a very strange and disingenuous way instead, actually go in those communities and ask them, why are you doing this? What's going on here? Look at the histories, look at these structures, look at the environmental racism, right? Like, you can't point through northern Virginia, the excerpts of northern Virginia and Columbus, as your case studies, while also excluding the places of Memphis, Indianapolis, and Mesa, Arizona, and saying, "Well, those don't matter because they don't fit my model." I think this is also why that final section, the real concerns, I think, ends up being, I think, the most disappointing section for me, because I, you know, again, I want something where people are articulating what democratizing governance is if we're going to be using those words, and this is not that, right? Here, like, for example, the opening, part of why the moratorium pushes such a dead end is because the disparate right-left coalitions, not of emerged around stopping data centers have different interests when it comes to other issues. It doesn't follow that stopping data centers will lead us toward a clean energy build out, or the social policies needed to address job displacement such as healthcare for all. I mean, I think that this is, I understand this argument because in one way or another, this rhymes with the argument that we might have about, of course, why you should never ally, which like do some sort of green, brown, red, brown alliance, right? That, uh, fascist, for example, are so diametrically opposed to what a socialist might wonder, liberal might want, that you work with them and you're going to have your lunche-edit. What I think that, that is a very different creature from what we're talking about when we're thinking about the data center moratoriums, um, because here, the, at the, the move is, I think, like, yeah, of course, of course, that stopping a data center is not going to get everyone to agree with Medicare for all. It's, but it's, and it's also the case that, you know, none of these individual political moves are going to lead to some sort of comprehensive program. But, you know, there's a lot, there's a different question here. We were asking, what are the tactic? What are the conditions that it can create? What are the leverage points that can realize? What is the political momentum that we could direct towards a larger broader aim? And, you know, as is kind of jestered on at the end of the section, a lot of people are trying to, as a first step, say, what if we shifted tax burden from, you know, recalibrated it such that it was not capital versus labor, but shifting towards compute since clearly compute is going to be used to zero out labor. On this question, it's not really, that if you want to argue that a right-left coalition is not going to be able to lead to a shift in tax policy and industrial policy that looks at compute instead of labor, I'm, all years listening to that, but that's not what's being done here. It's that the moratorium is sort of a residue of a naive move, which is that if we all band together on the norm, we can make anything happen. But the fragility of the coalition is not in of itself the just qualifying thing. Yeah, it's a constraint you have to work with. But again, we should, you've got to step back and ask, what is the thing that you actually want to organize with for this left-right coalition, if you're going to have one, you know? And then from there, you say whether or not that broader demand is possible. If we're simply going to stop at healthcare for all, I feel like that is a bit of a low-hanging fruit. And also, if we move on to the next part, we're saying, like, you know, she quotes a person, not person, who says that. Do you mean that the. Siding as evidence and article from Astric magazine? Yeah, I know. The Bay Area rationalist effective altruism magazine. That was the point where I was just like, "Oh, you let it show it. You let it hang out a little bit too much, where your bread is butter." I know. It was also surprised by that one. But this, you know, saying, look, like, pause is not a substitute for an actual AI governance and attempting to tackle all the issues through single move makes it less likely they get addressed. Yeah, sure. But why is it that we have to chew. Why are we given a false dichotomy between pause and governance? You know, we can have a pause and govern. I mean, we have done this. In fact, it's probably easier to do democratic governance. If you fucking pause the train that's running away and gaining steam and building more power, the longer you let it go, right? It's probably easier to shackle the beast if you say, "Hey, can you stop moving for a second so I can put these shackles on you?" This is also how we've done things. You know, it's not, of course, not say we've done things great. But this is a model that does work. You can actually pause doing something to develop it further. You know, for example, the financial sector, right? You know, I'm trying to figure out how we're going to regulate. Maybe we stop certain products and to rev an instruments from being widely used and accessible while we do a settlement, right? And figure out what sort of a range of financial regulatory framework we're going to do, right? This doesn't really make sense to me. And the idea that then it almost immediately jumps into this thing where it's saying, "Look, you know, we can't pause and we can't govern. We have to either pause or govern and pause is bad. Governance is good." And the people who are trying to suggest it any other way, you know, this is a little too proximate to like this sort of conspiratorial environmentalist, political violence channel that's opening up now, right? And that, you know, these movements, which are concerned with bringing down massive infrastructure and political violence, they're going to be, they're with us now and they're genuine concern that's going to emerge now as people maybe try to target additional executives, like they did with like this guy did with Sam Altman, right? And I think his house was targeted twice, right? But what's the implicit argument? You know, under this. The idea that organizing against data centers happens to lead to this outcome in extreme and so we should discourage it. I mean, what are we talking about? The whole history of mass movements in this country suggests that, you know, if someone's every movement, every successful mass movement has had a flank that is going to apply some pressure either the capital property or the visuals. The idea that the main movement should denobilize because some people may be engaged in property damage or some people in political violence is nuts. And I'm not, I must be misreading it because it's like there's no way that's what's being implied. You're not though. That is what she says. Like she's literally saying that that opposition to data centers, you know, pro-moratorium position puts you in bed with, quote, conspiratorial, parro environmentalist politics rife with concerns about electro-manetic fields and cellular damage. Or it might lead to political violence and subsequent crackdowns on activists, right? And so in other words, it says the existence of a wacky friend. So, you know, I guess from what we're learning here is that a pro-moratorium political position is composed of affluent ex-surbanites who are protecting their own class interests as knowledge workers and also conspiracy people who are afraid of electric magnetic waves and also a violent eco-terrorist who are Molotov-ing. Why are we pretending that the guy threw a Molotov because he was mad about data centers? He's mad about the anchor. He's angry, you know, he's mad about the fact that the tech industry has so much power because he's anxious about how it's explaining these other things. You know, it's a disingenuous way of painting the position by saying the existence. In other words, what it says is the existence of a perceived wacky fringe of a political movement discredits the entire political movement, right? But by that measure, no political movement is credible because every political movement has a wacky fringe in it, right? That does not mean it's the center, it means it's the fringe, but also who you count as your fringe according to this, who you count as, you know, bringing the fringe into the core or whatever, and the, according to this argument as well, those categories are flexible depending on who you're trying to discredit, right? And so, it's the same kind of analogy that's being used to say, you know, the existence of a more militant opposition in your political movement is a way of discrediting the entire political movement. Look, I'm sorry, but do you know what that rhymes with? Do you know what that makes me think of? Is it's all of the people that said, of course, I support the civil rights movement. Martin Luther King is up. He's a saint, but they got Malcolm X, you know? And I can't really support that. Like that is to me, like, it's the same kind of argument. It's the same kind of analogy. It's saying, of course, I support in principle, the position that you're posing here, but I cannot support it because, you know, there are more militant wings in that position that discredit the entire flank. Folks, we don't want any more Malcolm X's, okay? We don't even want Martin Luther King's. We want Bay Area rationalists at effective altruists leading us hand in hand toward the next capital fundraising ground, right? Wherever that's gonna be in the Bay Area. I mean, look, and it's also, I don't like all of this because it's, I also want to think that just kind of bothers me. I don't really accept a framing that if we, you know, bundle up all this stuff together, it is going to lead to worse outcomes, right? These are, if we're trying to figure out what to do about climate, about water, you know, about financial risk to the economy through the displacement of labor, right? You know, through, you know, all of these, it is fair to handle them all at once and bundle them and try to find a single mechanism that can give us leverage or that can great conditions, right? Or that can force issues to allow us to then do other negotiations because they're all downstream of, you know, a driver, right? You know, why are these, why are we concerned about water, why are we concerned about environmental impact, why are we concerned about finances, why are we concerned about labor displacement? Is there something that could be possibly at the center of all of them? I appreciate the attempt to kind of disaggregate them, right? But I don't think it properly appreciates that, you know, the hyperscaler capital expenditure cycle, it is, what is it doing? What is it doing, right? It's producing these climate impacts. It's producing, strain on water. It's producing systemic financial risks, it's producing labor displacement. These are not separate issues that have happened to be co-occurring. It's a single accumulation process. In terms of political tactics, it also doesn't make sense because a high profile problem is always a way to get people to care more about the cause, right? Yeah, people care about data centers right now, then you should be using that as a way to talk about water issues, about about emissions, about about land use, right? Like you should be using, but instead what she's saying is, why are we talking about data centers when we should be talking about water resource management and de-carbonization? You, come on now. (laughing) Like, let's say we do, let's say we do what she's recommending and we say, okay, look, let's disaggregate them. Let's send each issue to these institutional contacts where they can be handled, climate issues. Let's send them to the, let's keep it up state, bubble utility commissions, let's handle it with clean energy mandates, right? Water, state permitting, financial risk. Well, she doesn't, you know, offer something for that one. That's no hope. A full labor displacement, you know, pause, industrial policy debates, right? You know, each of these are framed in size in such a way that they handled the issue at the scale of mitigation for sure, but not in terms of being able to attack the underlying driver. And this is the sort of technocratic move that I think is like emblematic of this, you know, this large umbrella under which equal modernism is one and techno optimism are some of the things that we take issue with, right? You take a very big structural issue. You break it down into manageable pieces. You're route them into, into, you know, smaller institutional frameworks which can mitigate them and you call this governance, but we're not actually handling the structural driver that is going to generate more of these dysfunctional things, right, or these, these, these functions that are going to lead to these conflicts where we're then going to say, "Oh, well, you know, everyone wants to make this "like this left and right coalition." And we got it just, we just got to disaggregate it. And in the, in the process of disaggregating it, bring it down to levels at which people have a better handle on bringing it in to the sort of policy visions that will allow us to address each individual thing. I don't think, I don't, I don't think that's the case, right? I mean, look at housing. And in, you know, look at the housing crisis in New York, for example, I mean, the solution to housing, for example, you could try to solve it as a series of discrete, separate issues handled by series of discrete individual institutional context, or you could try to create like a coordinated affirmative program where you have multiple levers doing specific work in service of a larger, ambitious coherent strategy. And I think that disaggregating makes it feel like that's what is happening, right? It has the form of this sort of differentiated institutional analysis, but it doesn't actually have a coordinating mechanism in structure that would make this an actual strategy. - No, 'cause it's fundamentally a wonkish view that believes that if you have the right committees and subcommittees, you'll solve the problem, right? Instead, no, what you need to do is when people, when people latch onto a highly visible, highly impactful, highly concrete issue that they can reach out and touch as a way to understand a bundle of otherwise abstract and hard to touch issues, when they have one thing like data centers, for example, that might be an in-road to caring about, discussing, learning more about, acting out against issues like water resource management and carbon emissions and land use and concentrations of power and wealth, when, but you do, is you go up to them and you slap them. You say, "Why are you caring about data centers? You should be caring about the disaggregated bundle of structural causes and issues." And they're like, "I thought that's what data centers were helping me do." No, wrong. (laughing) - Right, right, right. I mean, that's called good politics. That's called not blundering your leftist strategy is when people latch onto a concrete material issue that represents larger structural abstract causes, you slap them in the face and say, "Stop caring about that thing. Care about this other thing." - Yeah, it's, I don't know. I mean, we could go so much more because I think like, there's a whole other section talking about the climate passages, about the water passage, about the financial risk passage, about the open AI report. I mean, this, yeah. The article also starts waiting into what is now becoming a increasingly more common counter argument is that like, actually all the environmental issues of data centers are overblown. And when you aggregate them, they're just a, they're such a small minority of carbon emissions or water use that doesn't actually matter, right? So it's like, it was a little disappointing to see, buck kind of weighed into the, the, and by the way, all of the environmental concerns about data centers are overblown and should be deflated because it, you know, in the grand scheme of things, it just drops in the ocean, therefore it doesn't matter. Disappointing to see that as the, the, the ending coup d'état of the article. - Yeah. - coup de grace, not coup d'état. - Yeah, I mean, I know, yeah, you know. I was wondering when we were gonna have a spoonerism, you know? I would, I would, I had a provided one yet. (laughing) (upbeat music) - Look, in classic team K fashion, we just did this in the last episode. We started out with saying, we got two articles to talk about. And then an hour, almost an hour and a half later, I went to article number two. But look, we spent the vast majority of our time talking about the article that I think actually needed our time because I do want to shout out this one in the New Yorker. I was really surprised at how much I enjoyed it because I was not expecting the New Yorker to come out in a really strong, well, I think well reported, well argued, well researched piece that is like fundamentally in favor of opposing AI in schools, right? It's a piece about opposition to AI in schools, but it's written by Jessica Winters. She's a New Yorker columnist who writes about family and K through 12 education. You know, it's, it's someone who this is their bag, right? This is what they do, they write long reported investigative pieces about education. And it's a piece called, what will it take to get AI out of schools? and the deck. here is the tech world assumes that AI-AIDed education is necessary and inevitable, a growing number of parents, educators, and cognitive scientists say the opposite. And again, I think it is a really good antidote for what we see as a technologically deterministic narrative that AI in schools, especially, I mean, here, they're talking about primary schools, K through 5th grade schools, elementary schools. This article is a good antidote against a prevailing narrative that AI in elementary schools is inevitable, unstoppable, and a progressive force that you should not oppose. You cannot oppose it. The starting argument is one that assumes opposition is not possible. It's a great piece that really deconstructs that in a very reported way, in a way that I think actually provides a ton of evidence and a ton of reporting in a way that the Jacobin essay just fundamentally did not. It comes to an opposite conclusion of the Jacobin essay. And so, you know, Comrade New Yorker, I welcome you into the fold of the Marxist-Luddism movement. And there is, yeah, mainly wanted to highlight this piece because there is such a good quote here that I just wanted to get on record for Team K. It's from a parent that was quoted in the piece here, Craig Garrett, who was interviewed here and is a concerned parent and has part of a group called Families for Human Learning in Brooklyn and is also part of the New York Citywide Coalition for an AI moratorium. So, again, a moratorium, which is a New York Citywide group of educators, parents and students petitioning Mamdani, the New York City Mayor, and Kamara Samuels, the school's chancellors for a two-year pause on AI and K through 12 classrooms. And Garrett had a good, Craig Garrett, this member of the coalition for an AI moratorium said quote, "If you ask tobacco companies to help write your school's policies on cigarettes, you're going to end up with guidance on how to smoke responsibly in school." And I thought, that's it. That's because what we're seeing here is a ton of partnerships between Google, OpenAI and Thropic, as well as venture capital firms like GSV Ventures, these edtech focus venture capital firms who are partnering with public schools and private schools around AI usage. Around technological technology, the diffusion and deployment in schools, around locking into software suites. And so what you end up with is a bunch of schools that are writing their technology usage policies in partnership with the companies who are selling and providing the technology. And I thought this analogy here of like, would you go and ask Philip Morris for a good policy about tobacco usage in schools? Of course you wouldn't. You see there, why? Of course, because that's what you're going to get. You're going to get policies for responsible tobacco usage in schools. This is how the smoke cigarettes with appropriate safeguards, right? You have to use one of those long cigarette holders from the 1920s to smoke your cigarette if you're a 12-year-old. You know? They're going to find their way into that market though. That's the funny thing about it. Even if there's a moratorium, they're still going to find a way to get their product in that market. Just the same way Philip Morris got those cigarettes in my 12-year-old hands. I mean, they're already doing it, right? The kids bringing their smartphones to school and shit. It's formalizing it through these partnerships. The New Yorker piece I thought was really good, but it was really that quip that really stood out to me. I just want to read the last paragraph of this piece as well, because I think it really sets up like how the argument, how the positions here are often positioned in a false dichotomy, just as we saw with the jacket in peace. You can either be against this or you can be for it, but in responsible ways. You can't be both, right? And one does not lead to the other. And you see this with EdTech and AI in schools as well. So they go on to say, Brody, and here she's quote, Brody, Kate Brody is a mother of a first grader in the LA school district. And as part of a group that wants to establish clear consent guidelines around the use of digital platforms and to adopt what they're calling a student tech bill of rights. And so someone that is, again, organizing around like clearer consent guidelines around the use of these technologies, you know, responsible guidance bill, right? And the New Yorker piece ends by saying Brody told me that anti-AI advocacy and education is tricky because screens have become virtually synonymous with schools. And AI is increasingly synonymous with screens. quote, you have to be more surgical about it than with a lot of other problems, Brody said, unless you're going to like take the computers and chuck them into the C, quote, but why not? I thought back to what Sinha, who's a Google education representative, had asked me, quote, what do you want from this? And then the piece ends, what if the answer is nothing? And I thought that was exactly it, right? Where it's like the advocates and the activists foreign against this technology are coming at it very calm, very regularly from a position that you can't get rid of it. You can only talk about the right way to do it, but you're going to have to do it, right? Unless you hey, unless you just want to pick up all the computers and chuck them in the C. And I thought the New Yorker piece was so dead on where it was saying, what if I don't want anything from this technology? What if the choice is not what do you want? What if my answer is nothing? Why is it not a legitimate position to say, get rid of it? I don't want to have endless debates about the right way to do it. The right way to do it is to not do it, right? And I thought the New Yorker piece was so powerful in that it provided, again, a well-argued, well-evident, investigated piece that gave voice to the legitimacy of a politics of refusal. Whereas the Jacobin piece bent over backwards to deny the politics of refusal as a possibility of legitimate or credible possibility at all. One of those positions is based in evidence and one of those positions is based in rhetoric, right? And I think we've seen good examples of how things fall on those sides. Of course, find us at patreon.com/thismachinekills for additional premium episodes. And until next time, later. Adios! [Music] [Music] [BLANK_AUDIO]