The Truth About Data Centers w/ Paris Marx — Better Offline

[00:00:00] Speaker 1: You there, boy! What podcast is it? What podcast is it? Sir, it's Better Offline! That's right, folks, we're back. I'm Ed Zitron, this is Better Offline. Today we're talking about a subject near and dear to my heart. That's, of course, those goddamn data centers. And I brought on an expert who has a wonderful book coming out, Paris Marks, host of Tech Won't Save Us, and the author of the newly released Hyperscale, The Ambition and Excess of Big Tech's Data Empires. Paris, thank you for joining me. [00:00:36] Paris Marks: Absolutely. Great to be back on the show and to chat with you about all this stuff. [00:00:40] Speaker 1: All right, so we'll be through the book, because there have been a few people covering this, but what approach did you take to these things? Because I know they kind of, we've had data centers for a while, they haven't been quite as offensive, but where do you start the story? [00:00:54] Paris Marks: Absolutely. So, you know, I really start by looking at one of the stories that really brought me into it. You know, people might remember a few years ago, there was this case in the Dalles in Oregon, where, you know, there was basically, you know, Google, this was where Google built its first company-owned data centers, right? You know, as it was building them out, you know, as its search engine was growing, as it was launching new services, it realized that it needed more computation, more than just it could lease from other people. And so it was like, oh, we need to start building data centers, right? You know, this is like 15, 20 years ago. And the Dalles is where they end up going. And so, you know, by the time that, you know, it comes to like 2021, 2022, Google is looking to expand, but they need a new water deal with the city, right? Right. And so... [00:01:41] Speaker 1: Just water deal, what do you mean by that? [00:01:43] Paris Marks: Yeah. So the deal on how much water that they can actually consume, right? That they can take out of like the public water system. And so people in the city are like, you know, okay, Google's been here a while. Like, you know, they're fine as like a, you know, corporate tenant or, you know, a corporation that's in the city. You know, they're not like amazing, like the industries that used to be there, the kind of industrial industries that left over time. But they're like, how much water is Google actually using? Like, you know, we would, we think this is an important, you know, kind of figure for us to understand in order to talk about giving it more water. And surprisingly, Google will not even not share those figures. But when a journalist at the Oregonian, you know, kind of statewide media... In local paper. Yeah, exactly. Goes to ask about it. Google not only will not provide it, but actually gets the city council to sue the journalists to make sure that the information won't come out. And in this case, kind of proceeds for a whole year before the public relations backlash is essentially so bad that Google relents and eventually shares that information and basically, you know, has to let the city know and the residents of the city that they are using 29% of the city's water, which, you know, for many people is staggering. Exactly right. [00:03:03] Speaker 1: And this is the thing, though. After, I know that, I don't want to go into it because these people are weird, but there was the whole pushback on Karen Howe because she made a mathematical error over water. But people are just pretending that water isn't involved in data centers now, [00:03:16] Paris Marks: which is bullshit. Oh, it's wild, right? And so there's a few different tacks that they take on this, right? And I feel like to me, when I see what they're doing, it's like the data center lobby, the data center supporters, whatever, they see that there are real challenges to public acceptance with, you know, these massive data centers that they're trying to build out across the United States and the world, right? And so they can see that they can't really make the argument on, you know, the kind of plane of electricity use and climate impacts because they're just using so many fossil fuels. It's like ridiculous. But then they see water and they're like, oh, I think that there's an opportunity here to muddy the public perception on what is going on, right? To say like, oh, look, golf courses and agriculture use so much more water. So we shouldn't be worried about the water that data centers are using. You know, when we're looking at kind of like abstract national level figures in the United States. But then of course, when you zoom in on the local level, you can see that this can have some real major impacts, especially because there was some Bloomberg reporting, I think it was last year now, that there are a lot of these data centers, you know, basically being built in areas that are high water stress, you know, they don't have a lot of water to provide to these other industries. And so I feel like, you know, they kind of try to muddy the waters, try to confuse people in that way. But then the thing that we have also seen shift in, I would say the past like six to eight months, is the industry recognizing that this is a problem. And so instead of, you know, basically planning data centers with evaporative cooling, which was the way that they traditionally did it in order to save money. So that's kind of like, you have a constant draw of water into the data center, that water kind of flows through the data center and kind of takes the heat away from the computer hardware, the chips, the graphics cards, things like that, until it gets, you know, warmed to such a degree, that that water, for the most part ends up evaporating away, right? And there's a bit of like non potable water that is ejected after that. But, you know, pretty much you need a lot of water coming into the data center to make that, that cooling system work. What they're doing now is they're switching to what's called closed loop cooling, right? Which I'm sure you've talked about on the show before. But it's essentially like, okay, we're going to take in a certain amount of water, we're going to keep that amount of water in the data center, it's going to flow through, and we're going to keep cooling it down as it, you know, kind of goes through the data center. So in some ways, right, you're obviously using less water by doing that. But you're using more electricity, you know, because you need to cool the water down. And so, you know, then there's the increased cost of the electricity, you're probably burning more fossil fuels in order to do that, given how we know these data centers are often powered. There's also the water that goes into generating those fossil fuels. But the other piece that they don't talk about, and that we're seeing a growing focus on right now, but I feel like it's not kind of fully broken out into the mainstream discussion on data centers, is that when you're doing this form of water-based cooling, where you kind of need to keep cooling this water back down, there are also a lot of coolants, PFAS, chemicals involved in that process, right? So it's not just water that's going through the data center. [00:06:18] Speaker 1: So you're trading one sheety thing for another. [00:06:20] Paris Marks: Exactly. And we're already seeing concerns about those chemicals kind of leaking into water systems and things like that. So this is, I think, a story that's still developing and that we're going to see become part of the data center conversation a lot more in kind of the coming six to 12 months is my [00:06:34] Speaker 1: thinking. This is the thing, this is what pisses me off with these data center centrists, I call them. The masses of the world. The guys who are like every day like, I imagine they're once again repeating the current how claims as if data centers, they fixed all the problems with them, as if every, by the way, there are still plenty of evaporative cooled data centers out there. Oh, 100%. Tons of them, like they've not gone back. [00:06:57] Paris Marks: Because they're only really like starting to switch over to closed loop now, right? Like tons of the ones that are using it are still on evaporative. So with the book, what do you go through? Like, do you go through the history [00:07:09] Speaker 1: of data centers? Do you, like, where do you start the story though? Is it before the AI bubble, middle of it? Like where do you, and where do you take the story? [00:07:17] Paris Marks: Yeah, it's kind of like right at the beginning of it, right? Because you know, I was telling you about that story in Oregon, which is right around 2022, right? And of course, ChatGPT is November of 2022, it comes out. And so this whole story is kind of, you know, playing out right before generative AI, and kind of finalizes as generative AI is just starting to take off. And for me, it was an indication that like, okay, you know, data centers were already an issue in certain communities, in areas where they were already building a lot, in areas that maybe didn't have the water or the electricity to provide to these massive infrastructures. But with generative AI, especially as we move into 23 and 24, we start to see that this is going to require a lot more data centers, which means a lot more communities are going to basically be facing up against what comes with having one of these hyperscale data centers located within your community or a cluster of data centers there. And that means that more communities are going to be facing, say, what the Dalles, Oregon faced, or what Santiago, Chile faced, or, you know, what Dublin and Ireland faced, you know, just to pull out some of the examples. What were they faced? What did they face? Well, that's it, right? Like the Dalles, it was the concerns about water. In Santiago, Chile, it was also concerns about water, you know, say Google was planning this data center in Serios, in like a working class neighborhood. And people there were really concerned about their continued access to running water because Google wanted to essentially take so much water from the system, right? It wouldn't provide assurances that it wouldn't actually reduce their access to water. And then over in Ireland, it's something else where it's not so much about water, though there has been a bit of a concern over there. It's been much more about electricity, right? To the point now that more than 20% of the electricity in Ireland goes to data centers. And that has impacted grid reliability. It has pushed up electricity prices for people. You know, stories that we have been seeing play out in the United States in the past couple of years, but Ireland was experiencing earlier, right? And so for me, these were like, indications of what can happen when there are a lot of data centers proposed for your region or area or community. And as generative AI was taking off, I was like, wait, so I can already see what's happening in these places. That's probably going to be coming to a lot more places as these companies are trying to build out so many more of these infrastructures, right? - Did you go and visit any data centers if you kind of checked in on any of them? - Absolutely. So, you know, I was outside of data centers in more places that I was inside, you know? [00:09:43] Speaker 1: - Yeah, I mean, I didn't expect you to get inside. - Exactly. - I wouldn't let you inside either. Like, I mean, that would love. [00:09:49] Paris Marks: - Yeah, the major tech companies certainly didn't want me inside. I did ask a number of them. You know, Microsoft was the only one that directly rejected me. The rest just didn't bother responding. But yeah, so I went to like Northern Virginia, which is actually the site in the world that has the most data centers, right? You know, basically stuffed into one region. And it was just wild, you know, kind of driving through this part of the state that is, has been essentially dubbed data center alley and just seeing like the degree of high voltage transmission lines and just the number of data centers. And also just like the thing that really stood out to me was how close they often are to residential communities. Like, you know, you could be kind of driving along like a freeway or something or even just like, you know, kind of a more built up road with, you know, with a number of lanes or whatnot. And on one side is this massive data center. And on the other side is this kind of, you know, kind of suburban community. And what you often notice is that you drive into these communities and they have these kind of like walls of trowal trees surrounding them so that they're not, you know, kind of seeing the data center that's right next door. Though they're probably kind of experiencing some of the impacts that comes with that, you know, hearing the noise pollution, that kind of stuff. Yeah. It was just wild to see how, how close they were and also just how, you know, how large, how large these data centers were as well. Right. And, you know, so I did that in, in Northern Virginia and Ireland and places like that. I actually went down to New Zealand because I knew some people down there who helped me get inside of a data center. Right. This certainly wasn't, I visited a couple down there. They certainly weren't kind of the size of what a data center was like. Yeah. Were they, were they even GPU compute? Were they just regular? There were some GPUs in them, but it was like a mix, right? Yeah. So, so there were some GPU and water-based cooling, but then most of it was kind of, you know, your traditional hardware and kind of air conditioning and that sort of a thing. Right. And so, you know, I, I'm not even sure I would dub it like a hyperscale data center. Like it was large, but it wasn't like on the scale of what an Amazon, Microsoft or Google, or, you know, one of these major data center companies is building. But even then, you know, you know, the thing that struck me was, you know, obviously I expected to go into the server room, uh, you know, and, and, you know, to see what it was like to have all these servers set up, you know, maybe people would be surprised to learn or, or maybe not that it's quite warm inside of those rooms. You know, obviously the computers are throwing off a lot of heat, but they don't cool the whole room, right? Because that would be really costly. They try to, you know, kind of isolate the cooling to the server racks themselves so that you're not wasting all of this, all this cooling and paying all this additional money that you don't need to pay. Um, but the thing that really stood out to me was the level of security. You know, I needed to like send my passport in advance so that they knew I was coming. And when I got there, they needed to check my passport to make sure I was the same person. And then as we went through the data center to go through every single door, it needed to be like, you know, kind of the guy bringing me through needed to scan his card, um, to unlock every single door. You know, like you could not get through a door. [00:12:52] Speaker 1: That's my, and this was my experience of going into one like 15 years ago. I've only been inside one. Yeah, I think it was about 15, maybe more actually years ago. And every few doors you had to, and any, any time you went from a room of service into another room, you had to beep through. Yeah. So it, the fact it must've got even crazier. [00:13:13] Paris Marks: I'm sure. I'm sure. And, and again, you know, this was a data center that was not, you know, built brand new. It was probably 10 to 15 years old, though obviously it had been kind of updated over time and whatnot, uh, because a different company had taken it over from the one that built it initially. Um, but yeah, that was the thing. Like, obviously, you know, I expected it to be securitized, you know, I expected there to be a high degree of security. Um, but the, the fact that like every single door in the place was, you know, kind of locked unless you unlocked it was, uh, yeah, that one was a bit surprising to me. [00:13:59] Speaker 1: So how did things shift upwards? I know that they are bigger, but did you notice anything shift when the real data center party started with regards to like AI and the big boy GPUs? Because just for the listeners, there is a big delineation between a data center for streaming media, which would be very CPU heavy, very storage heavy, a lot of, a lot more cabling to an extent. I've no, actually, I'm sorry. The networking would actually be more in an AI GPU, but nevertheless, they are materially different. Did you notice something different in the ramp up? Because you've been working on this a while. They could be working on this before most people were really obsessing over these things. [00:14:39] Paris Marks: Absolutely. And you know, there were also people who were working on it long before me, right? Just to make sure that they're acknowledged. I wasn't suggesting, wasn't suggesting that was the case, but just, you're the, you're the guest. Yeah, I appreciate it. And I wasn't suggesting that you, that you were saying that, but yeah, to, to get into what you're saying, like, I do think like, and you asked this earlier about, you know, what I get into in the book, like, I do think it's important for people to understand that data centers are not a new thing, right? You know, in some form they have been around since the beginning of computation. You know, obviously there was a particular ramp up as the internet was privatized. We started to have all these digital services in the late 90s and into the early 2000s. And then in the book, I also really talk about, you know, kind of the mid 2000s, you know, especially with the creation of Amazon Web Services and then Microsoft Azure and Google Cloud afterward as, as an important turning point where you have much more of the kind of movement toward the centralization of computation by these major companies, right? And, you know, I think that that really stands out to me as an important moment because without what happened then, right, without the kind of beginning of the cloud, I would argue that you wouldn't really see generative AI in the form that we have it today without those companies beginning to have so much computation under their own roofs, you know, 15, 20 years ago. And so that was kind of an important, you know, an important moment where we had an important shift. But then if we kind of fast forward to the generative AI moment, we already had hyperscale data centers, you know, basically data centers that are a particular size that, you know, have tens of hundreds of thousands of servers within one roof. [00:16:22] Speaker 1: Yeah, I think it's like 50 megawatts, I think is the, I forget exactly what the, what the turning point is to consider something hyperscale, but it's something like that, right? [00:16:33] Paris Marks: Exactly. For a while, there wasn't like an industry accepted definition, and I'm still not sure that there really is, but I feel like the 50 megawatts is one that has been gravitated around, especially in the past year, as like the delineation between what's hyperscale and what isn't. And to be clear, like 50 megawatts today is even on the lower side of what many of these companies are building. It's a pathetic small one compared to the ones they're boasting about. Exactly. You know, you're looking at like 100 and 150 on the lower end for some of these, you know, AI focused ones. And then you're looking at clusters that are like a gigawatt, two gigawatts. You know, Meta says that it eventually wants to build five gigawatts in one cluster in Louisiana. We'll see if that ever happens. It probably won't. Yeah, I don't think so either, right? But you know, this is what they say they want to do. But I feel like the change that we started to see in 2324 was one, an acceleration, right? So they're looking to build a lot more in a shorter period of time. There's also an increase of scale, right? So you already have hyperscale data centers, but they're getting larger and larger. And they're trying to cluster even more of them together in, you know, kind of a single region or, or, you know, a single space, you know, if they can, if they can try to get away with it. And there's also the shift, you know, as you were saying, from the more focus on CPUs and having hard drives in these data centers, to using the graphics processing units, you know, from Nvidia in particular, but obviously, obviously they're trying to use more different ones now, if that's possible. And that's an important shift as well, because they, you know, the, the GPUs do run harder, right? You know, they require more electricity, they throw off more heat, which means they need more cooling as well. And so even if you have a data center, that's, you know, kind of of similar size, and one is a more traditional data center, and one is a more AI focused data center, the AI focused data center is going to have a bigger kind of draw on electricity on the water system, those sorts of things, than, you know, the, the one that would have been built before, or the one that's, that's built for more kind of traditional use cases, right? And so these are all important developments. And I feel like maybe the one other thing that really stands out for me when I think about it, is just how aggressively the tech industry has really gone after fossil fuels in the past few years, as well. You know, if you go back to 2018, 2019, 2020, they were all starting to talk about how they cared about climate change. And, you know, they were committed to net zero and reducing their impact and all this kind of stuff. And then as soon as, you know, generative AI comes along, all that gets thrown out the window. And now, like, just the level of fossil fuels that they are building to power their data centers is just staggering. And, and at a level that makes really addressing climate change, like, seem virtually impossible. [00:19:24] Speaker 1: So, and there has been pushback though, and you've probably seen data center delays, right? Did you happen to see any of the groundswell on the grant? Like, have you seen any of the local opposition? Have you seen how people are pushing back on these things? [00:19:39] Paris Marks: Absolutely. And so I met some people in some of the different countries that I went to, who are doing this work, you know, in France, in Ireland, in places like that. And, you know, spoke to people in, in different parts of the US as well. Right. And I feel like this is one of the really hopeful stories for me. You know, we have really seen the tech industry push hard with its ambitions the past few years, you know, the push generative AI into virtually every facet of our lives, you know, to build out these massive data centers, to bring them to almost any community that, you know, they can possibly think to stick them in. But those communities have been really resilient in pushing back against it, right? We have seen across the United States hundreds of local moratoriums being passed on data center construction. Obviously, we have the first statewide moratorium in New York state right now. You know, people would quibble over, you know, whether it could have gone further, but you know, at least it's kind of like a first step. You have the push on the federal level by people like Bernie Sanders to bring in a kind of federal moratorium on data center construction. And we see it in many other countries as well, right, where data center moratoriums have been passed. You know, just recently in Denmark, they basically put data centers at the bottom of their list for grid connections. So everything else has higher priority than a data center. In Canada, in Quebec, Hydro Quebec just said data centers are going to pay double the rate of other kind of electricity customers for their electricity. So we do see this pushback. And I would argue that, you know, none of the moratoriums, you know, very few of kind of the government restrictions on data centers would be happening if it were not for people turning out in their communities and saying, we do not want this data center here. And I think that there are a number of reasons that they're doing that, right? I think one, they have seen the stories of data centers for the past few years and what it can mean for their electricity prices, what it can mean for, you know, kind of the local environment, what it means to live in the communities where they live. But then I also think that there are some people who are motivated by these bigger concerns, right? You know, the concerns about what the tech industry has been doing, about how their products are affecting us, about the tech industry's alignment with Donald Trump and, you know, the far right. And so I think that there are a lot of reasons that people are getting motivated. But I just think generally, there has been a growing frustration and anger with the tech industry. And, you know, if you have a problem with, say, your social media app or something like that, it's hard to really, you know, in a tangible way, you know, make Mark Zuckerberg feel your anger at his decisions around how, say, the Facebook platform works or something like that, right? Yeah. [00:22:25] Speaker ?: Yeah. [00:22:26] Paris Marks: You know, there have been a lot of kind of moments where we've been angry at different apps or different companies, but there's not, but it's hard to really do something tangible about it. And with the data centers, it feels like, you know, there's finally a way to really stop something that they're doing because it is something tangible that would be coming to your community and you can turn out to your local council and you can get that project shut down in a way that you can't do with an app on your phone or something, you know? [00:22:52] Speaker 1: It's also a physical manifestation, like you said, of all the problems they're creating. And it also feels like a violation because the shit they've done to the internet is disgraceful. Like it, like the, the ways they have twisted the internet up, their allegiance with the right wing, their allegiance with the worst people alive. And also they are now like, no, we need to build the biggest, most power hungry, ugliest fucking things you've ever seen. Just these giant shoebox looking motherfuckers that make endless noise and have tons of lights and all this shite. And they're like, no, we need these. And by the way, in return, we will get a tax break. Yeah. And it's just, it's fucking ugly. I will say I've, I have seen, and if you were listening, there's a guy I think in Wisconsin who in a data center meeting in like a town hall meeting called, like mentioned my name. If you're listening, mate, you're a legend. You're a legend of better of life. But there are people that listen to the show that go to these town hall meetings. You people are way more courageous than I am. You're doing, you're like, whatever I do is nothing compared to the people who are actually going down there. These people are actual fucking heroes and it's working as well. Like it's actually scaring the ever living shit out of these companies. I know because people in finance tell me and people in finance in many cases don't have souls. So if they're saying it, it's because they're actually scared. [00:24:12] Paris Marks: Absolutely. Oh, I completely agree. And you know, I think, and you would certainly be, be much more kind of on top of this than, than I am these days because I've been just been so overwhelmed with, you know, everything that comes, it comes with releasing a book. Um, but you know, the, obviously the recent stories around the growing questions about the financing of these data centers, you know, whether companies like Oracle or CoreWeave, um, you know, are actually going to be viable into the future, you know, whether all the money that they have expended on these data center plays and, you know, these kinds of AI plays are actually going to pay off in the long run. Um, you know, it has always been a question, you know, as you have been writing about for a long time. Um, but you know, it really feels that there's growing concern and questioning about the kind of short-term viability of all of these investments and, you know, these kinds of speculative bets, um, on generative AI and, and on the infrastructure that is supposed to be necessary in order to power generative AI and, you know, kind of the vision of the future that these tech companies have. [00:25:15] Speaker 1: - Within the book, when you were working on it, did you get any feedback from the industry? Did you hear back from any of the companies? I know Microsoft said no, but did you get any feedback from them? Were they trying to stop your line of inquiry anyway? [00:25:30] Paris Marks: - Not really. Um, and I wasn't very interested in really seeking their feedback anyway. - Yeah, I, no, I'm the same way. Like, why the fuck? I don't need to hear from a bunch of liars. - Exactly. And we hear their perspective all the time, right? You know, the media reports on their perspective constantly. Um, you know, if, if they have an opinion that, you know, they want us to know on data centers, they're going to have said it somewhere anyway. Um, and the focus of my book was really on, you know, what they're actually doing and also how that is affecting communities, you know, the wider public, you know, broadly the world that we live in, right? And so if it ever did feel necessary to put their perspective in there, I could grab it from somewhere else where they had said it. Um, but in terms of actually going to them and be like, you know, what's your response to the fact that you're burning all these fossil fuels? Like, it just didn't seem, uh, you know, important to, to, to essentially do that. Right. Um, because we hear it anyway. Uh, and so, you know, [00:26:29] Speaker 1: why do I need to- - We just turn on CNBC, New York Times, Tech, Crunch, Business Insight, pretty much everywhere. It's, I asked the question because I get asked it fairly often. Like, Oh, do these companies ever say anything? I mean, no, because I'm calling them rat fucks and bastards. Cause I'm, I'm like, like also if I needed them to say, I did like, for example, I worked on a story about actually AI GPUs with the Guardian and we got a quote from Microsoft, because I had the numbers of how many GPUs they had. And they literally said, they're completely wrong. You're misreading them only for Bloomberg to report them two months later as completely what we said. It's just like these people fucking lie. So why should we give them the time of day? And people will say, Oh, it's not fair on them. Fuck that. It's not, they don't, they're not fair [00:27:14] Paris Marks: on us. They're not fair on you. They don't want to help. And, and it's also like, you know, if they had let me into the data centers, maybe there would have been, you know, uh, an opportunity for a conversation there. Right. Um, but they were not interested in giving me kind of any access to see what they're doing. So why would I then turn to them and be like, Oh, what's your perspective? And how are you going to try to mislead me on the things that you're very clearly doing? You know, I can quote their sustainability reports. I can quote the, what their executives have been saying, or their comms people. Um, you know, I often feel it's not valuable for me to go to them directly. [00:27:50] Speaker 1: Yeah. And here's another thing that just occurred to me as well. I'm going to guess you didn't find out how many gigawatts there were, because this is the biggest thing I've found with data centers that we quite literally don't know how many gigawatts there are, how many GPUs there are, how many are active. We don't know what it's like inside. We don't really know how many of them are called. It's just the whole thing, despite how prominent it is, is shrouded in secrecy. [00:28:13] Paris Marks: Oh, a hundred percent. And you know, they work very hard to maintain that secrecy as, as I'm sure I don't need to tell you and your listeners, right? You know, the, the, the amount of NDAs that they have people sign, uh, the degree to which they try to even cover, uh, you know, when they're entering a community and whether they're actually involved in a data center project until it's like later down the line. Um, you know, they work really hard to try to keep these things secret, to try to keep them outside of the public conversation. Um, because they know that people are not going to be down with this and okay with this, uh, when they learn what is actually happening. [00:28:48] Speaker 1: Yeah. And it's for me, and this, uh, we talked about it before we went on, but Amazon is business insider revealed that Amazon is actually warehousing GPUs. Fucking exactly what I said a month or two ago, but everyone said I was crazy. Uh, but nevertheless, even it is kind of crazy to me that we don't even know how many are installed that this is what hundreds of billions of dollars financing, the economy resting on it to some extent. We don't know anything. And there's been no government attempts to demand any kind of clarity [00:29:22] Paris Marks: it appears. Yeah. And of course, who, you know, whose pocket is the government in? Unfortunately, it's often the AI companies, right? It's like, you see Donald Trump come out and say that, uh, communities opposing data centers just want to be like, I think it was poor and backward. He, he said, um, Chinese. Exactly. Like it's, it's crazy, right? Especially when you see that so many of his supporters, like so many Republicans are also opposed to these data center projects. You know, so many of them are not buying into it and it's kind of fascinating. Obviously the U S is heading into the midterm elections in like, what, just a couple of weeks now. Um, and you know, you, you have on the one hand the Republican party or, you know, kind of the leadership of it that is very much in line with the tech industry and the AI industry, but so many of its supporters are not buying into that project. So many of kind of the politicians on the lower levels are having to moderate their support for the industry and data centers because they recognize it is an electoral liability for them, right? You know, they might actually lose their seats or their positions if they are as pro data center or pro AI as they have been in the past. And we've already seen this happen in a certain states in the United States. Um, so yeah, it's, I don't know, it's kind of a fascinating moment because on the flip side of that, then you still have the AI industry trying to copy the playbook of the crypto industry two years ago and spending a ton of money in races to try to get the politicians that are more favorable to, you know, their, their kind of regulatory ambitions. And by that, I mean, you know, getting regulations that favor them, um, so that they can say, oh, the industry is regulated, but it's regulated with regulations that they have written to benefit themselves. [00:31:06] Speaker 1: But the thing is what record, this is the thing that kind of pisses me off as well is it's like people say, oh, tech, they just want special regulations. Like what fucking regulation? Yeah, regulate. We don't regulate tech. Well, precisely, right? We don't do it. And people are saying, oh, they want to rig the things and they've already rigged it [00:31:39] Paris Marks: in their favor. Oh, I completely agree. And, and like, to me, if you look at this conversation that's been happening there, that's really kind of gained more prominence the past couple months of like the AI is going to destroy the world and, you know, is going to kill humanity in 10 years or whatever, is very much an attempt to be like, okay, public anger is rising. There is pressure to actually regulate the things that, you know, we are doing and that we don't want to have regulated. And so we need to step into the conversation now and say, hey, there are problems with AI, but they're these future focused fantastical problems, not the things that you're experiencing, whether it is deep fakes or the mental health spirals or the questions about, you know, cognitive capacity and learning from dependence on chatbots or, you know, on and on and on, the impacts of data centers, like whatever, I can give you a very long list. It's like, we do not want regulations on those things that would actually affect how our products are operating right now, how our companies operate. Instead, we want lawmakers to be focused on the fantasies in 10 years or like maybe all the jobs are going or whatever, because, you know, those are things they're just making up anyway. But if they can make sure that, say, Bernie Sanders or someone when they're talking about AI is talking about the made up kind of narratives that they're putting out there about AI that the industry is putting out there, then that benefits them immensely, because then the regulation is based on fantasy and not on reality and kind of, you know, actual harms that people are feeling right now. [00:33:11] Speaker 1: And that's kind of what pissed me off about Bernie Sanders. I like Bernie. I like what Bernie is pushed for, but I feel like he's just boosted this entire AI thing ever since he did that weird conversation with Claude. He's like, Claude, are you alive? And it's just, but it is exactly what you said. It's this thing of getting this imaginary creature that doesn't exist and won't exist regulated so that they can keep doing whatever they want right now. Like they could, there are obvious regulations they can make right now. My favorite one to choose is notifications. I think we, if we regulate them like emissions, I think it could do genuinely good to hundreds of millions, if not billions of people. It sounds small, but making it so like the growth hacking things are harder to do, that would be an obvious one. But it's like, no, we need to make sure that when AI makes a profit, the profit share might happen one time if you're this person. And it's just like, fucking useless, worse than useless. [00:34:10] Paris Marks: Yeah, I completely agree. And, you know, I think I, I hate to, you know, kind of make the, the age commenter. I don't know. I don't, I don't really like, I think, I think Bernie Sanders is like a very well-meaning person, but I think he's a bit in over his head on this and really doesn't understand what's going on with the AI industry. And, you know, I, that's not to say that I don't think older people can understand technology. I think they definitely can. And I think part of it is probably like his staffers are probably feeding them, him this stuff as well, right? Which means that he doesn't have the right people around him to help him understand what is actually going on with the technology and what the right approach would be. You know, I was completely behind the effort to, you know, bring forward a nationwide moratorium on data centers that he was advocating end of last year. But then when it came to actually looking at the regulations on AI that he was moving forward, it really felt that he was just, you know, kind of getting caught up in the narratives that came from industry that were designed to be very deceptive. I would say, I think that Alexandria Ocasio-Cortez has been better on this and has been more focused on deep fakes and mental health and these sorts of questions. But I feel like her efforts have not been getting the attention and the traction that Bernie Sanders have because Bernie's reflect the, you know, kind of goals and intentions of the industry and the kind of sensationalized claims that they put out there. Whereas hers have been focused much more on very tangible things that are actually hurting people. And that obviously don't get the same [00:35:38] Speaker 1: degree of attention, right? Yeah. And I mean, the, what the other bloke on the bill, I forget who it was, his PR people came at me. They tried to go through contacts of mine to try and get on. And just to be clear, if you were listening, mate, I don't give a fuck about moderating super intelligence. Stop pretending that that's the problem. No one gives a fuck about, if you're regulating the scary computer, because Jacob Coxon said something, you've already fucking lost. Yeah, exactly. It's a classic capitalism trick as well. It's obfuscating the actual problem so that people can feel good about themselves because you can be like, oh, I, when the scary computer wakes up, this will happen and we will already be ready. We get to feel good about something without doing anything without restraining capital, even slightly. And it's just, it's very frustrating because by the time this all unwinds, we are also going to need regulations around what to do with unbuilt data centers. Yeah. And we don't, we are not going to be prepared for that. We're going to be prepared for the mystical Jack Clark computer machine that wakes up and fucks your granny or whatever. And it's, it's just, it frustrates me as well because you know this as well as anyone, the problems have been obvious for years. Yeah. All of them, like the environmental, the social, the economic, they've been there. They've pretty much been the same, but maybe smaller a few years ago. And we have not tried to stop any of them. Like we've just, we focused entirely on an imaginary [00:37:02] Paris Marks: one. Exactly. Like if you're paying attention, you knew this stuff, right? And, you know, I would say if you're a lawmaker and you have an office and staff, those staff could have been paying attention to it as well. If this is an issue that, you know, you as the lawmaker do want to be, you know, active on right. And, and, and doing things on, you know, the, the other thing that is, that is frustrating, of course, you know, obviously someone like Bernie Sanders being on the left, very you know, opposed to the far right has been very vocal about you know, the power that Trump has, the way that the tech industry has been working with the far right. And then you see him on stage with Steve Bannon because AI is going to kill the world or something or kill humanity or something. It's like, what are you doing? And, and what kind of alliances are you forming? And, and how are you giving, um, you know, basically credibility, uh, to someone who a few years ago, you would have been like, you know, you can never touch that person. You can never work with that person. Um, and, and there's some really dangerous, I think alliances being built in that moment as well. Um, which is quite distinct, I would say, you know, when we're seeing like a Bernie Sanders and Steve Bannon on stage [00:38:10] Speaker 1: together, I think that's a big problem. Yeah, that really, that turned my fucking stomach. Oh my, a hundred percent. Like, but like, especially given the, I would say questionable previous thing would be like, or Bernie bros, but it's like, what the fuck do you say to that when you're on stage with fucking Steve? What do you say to a woman to like, try and defend Bernie Sanders, where you're like, he's standing on the stage with fucking Steve fucking Bannon. And it's because of this wibbly wobbliness of AI. And honestly, I think, I also think it's an attempt to paraphrase hyper-normalization to make a complex situation simple. Oh, if we only prepare for a future that sounds roughly like the ones we've seen in the movies, we'll be fine. Rather than dealing with, rather than dealing with the actual obvious problems, which is we have unrestrained capital that is able to manipulate the debt markets. And we have a captured tech and business media who are actively marketing [00:39:11] Paris Marks: for the capital classes. Exactly. A hundred percent. I completely agree. And, you know, just to finish the point I was making there, it's like, there's a real problem. No, not at all. There's a real problem with Bernie Sanders and Steve Bannon on stage together. I don't see that problem at all when you're looking at, you know, kind of local organizers who are Democrats, who are Republicans, who don't have a fucking clue what their political ideology is, and they're all coming together to kind of face off against these data centers together to take them on and to have like, you know, maybe they come at it from very different political persuasions. You know, they have very different reasons for opposing these data centers. But at the end of the day, you know, they are people who are motivated to protect their community, to push back on the power of the tech industry. And maybe it's actually going to lead them to have some interesting conversations that they wouldn't have usually had because they wouldn't be talking to one another in this way. Right. And so I think that- [00:40:05] Speaker 1: Yeah. And that is actually bipartisan. That is actually two people- [00:40:07] Speaker ?: Exactly. [00:40:07] Speaker 1: ... of potentially different political creeds who share a common community. Bernie Sanders and Steve Bannon have maybe two or three things in common, like old, white, but not looking great. Like, I don't fucking know. And it's just frustrating as well, because especially, again, I've been on the left myself and like watching Bernie and being like, wow, like this guy believes in things I do. Wow. Like my father worked at the NHS. Like I really believe in these socialist causes. And then this fucking guy gets on stage with Steve Bannon to basically protect the interests of capital by claiming you're regulating them. Yeah. And it just makes me wonder how much the rest of it meant. I don't know. Maybe that's a little bit taking it a little bit far, but it's when your inherent beliefs are so, or even just your knowledge is so weak that you're willing to align with an actual fascist, a guy who has called multiple times for actual real life fascism. Just like, let's be clear about who Steve fucking Bannon is. And you are on stage with him because he also agrees that a thing that isn't happening might happen and it's bad, even though he is obviously doesn't give a shit about anything and just wants attention. Like Jesus, it just, it turns my stomach. It's the same, same criticism I have for the media. It's these people who claim, oh, we're going to, we're here to find the truth and report on the truth. And then they just repeat whatever the fuck Altman or Amadei says. It genuinely pisses me off. [00:41:34] Paris Marks: I completely agree with you. And like, you know, I have not been quiet about my frustrations with, with the media and how it reports on any of this either. Right. You know, we're very, we're very aligned on that, on this question because it's like, you know, for me, I see it as a real, you know, if you think about what journalism is supposed to do, right, it's really supposed to help to have an informed public, right? You know, they know what's going on, but also you would hope, you know, if you think about the more kind of idealistic visions of what journalism is, it's also to hold power to account. Yeah. And unfortunately what we see is, you know, a journalism that all too often defends power, but also repeats power's narratives such that the public is misinformed about what is actually happening, right? You know, the public is led to believe that these chatbots are actually thinking and developing, you know, independent consciousness and all this kind of stuff when we know very tangibly that that is not occurring, right? You know, repeating the other narratives that come from these executives and these companies that are designed to deceive and mislead. And so then the public, when they have their conversations about AI or, you know, about any of this stuff that's going on with the tech industry, it is kind of shaped and influenced by this deception that has been echoed and proliferated by the media instead of them kind of stepping in and saying, like, this is what the CEOs are saying, this is what the companies are saying, you know, this is what experts are saying though, and that does not really align with what is going on there. And certainly there's some of that sometimes. But, you know, as me and you and many of your listeners will know, the overwhelming, you know, kind of amount of writing and reporting that happens on tech is just echoing what the companies themselves say, which is not helping. [00:43:21] Speaker 1: And what's crazy is kind of like Bernie. I think a lot of these writers think they're doing a good job. Yeah. I think they think that they are because they have a, in a story that largely parrots what the rich say, they have a single paragraph that says, there are questions about the economic viability of this. [00:43:37] Paris Marks: And I just want to say paragraph 38 in the story or something. [00:43:40] Speaker 1: Exactly. And if you're a journalist listening to this, I am talking about you. Just if you're wondering if I'm talking about you, yes, sorry, it's just, I'm just going to skip right to it. If you think I'm critiquing your work, I probably am. If you have any concern about this, I am talking about you. Fucking do your fucking job. I'm fucking sick of reading your bullshit. I'm fucking sick. Like I'm actually this Amazon data center GPU thing, the GPUs in the book. It has in, it has really pissed me off because like Jensen Huang basically made everything more expensive to sell these GPUs in advance. They build these data centers and for years, the journalist just, just gone, wow, they're building the data centers. Wow. They're so big. Wow. These things are going to print money. Wow. Anthropic has five gigawatts of capacity, even though they don't, there's no proof of that. Let's just keep repeating it. And it's why these companies keep being able to do what they do. It's not just a regulatory problem. It's an accountability problem led by journalism. Absolutely. And, and the other piece, and one of [00:44:37] Paris Marks: the things I really try to get to in the book is like, okay, we're building out all this computation, or at least supposedly building it out, planning to build it out, whatever the companies are saying that this is all required and that they're deploying all this capital to build out these data centers. But what is actually the consequence of that, right? Okay. So we know that there's all these resources that we have that are being deployed now for data centers, whether those are capital resources, but also, you know, energy, fossil fuels, water, mineral resources. If you're thinking about the production of the chips and everything that goes into that, right, we're, we're kind of dedicating so much capacity to building out these data centers. And we're also accepting that there are a a lot of harms or, or at least the industry is hoping that we accept that, you know, we're just going to burn a ton more fossil fuels and, and contribute a bunch of emissions to the climate. And we're going to make a ton of communities, you know, be more worried about whether they're actually going to have access to water or even threaten that access to water. You know, not to mention all the things in the supply chain and, you know, the, the impacts on where the production of the chips happen and where the mining of the minerals happens. Okay. So there are all these impacts that we're just supposed to kind of go along with because generative AI is the next big thing. Yeah. But then what are we actually getting from generative AI, right? You know, a few years ago, the companies were saying it was going to be this like next industrial revolution. There was going to be so much efficiency and productivity that was going to come from it. You know, that it was going to improve access to like key services like healthcare and education. As far as I'm concerned, and as far as I've seen, we've not seen much of that actually come to be. You know, we've not seen many, if any of these kind of positive benefits be realized, but the amount of kind of social community environmental harms that have proliferated because of these technologies are, you know, the list is so long. And I mentioned some of them earlier, I could go on, but I'm sure that your listeners know what they are. And so it's like, okay, we're deploying all these resources for data centers. We're, you know, kind of pissing off all these communities in order to build them. But what are we actually even getting for all that? And I just don't see how any of it is justified. And, and to add to that, to pick up on our conversation that we were just having, how is it that our lawmakers and governments are just going along with all this or not questioning whether any of it makes sense, just because there's so much money wrapped up in it. It's like, it just makes you kind of disillusioned even further with the whole system. [00:47:05] Speaker 1: Yeah, it's just, I mean, it's because we're in the scourge of neoliberalism. We are still within a world that believes in growth at all costs. And also, I won't be honest, I think a lot of journalists, I think a lot of lawmakers, I think that they want the tech companies to win. I think it is that simple. I think that they believe it's a good thing. I think we still have people in this mind, this kind of cutesy 2011 mindset that these are little startups versus basically the blitzpins of our economy in a very brittle way. And it fucking pisses me off. I just, I don't have an eloquent way of saying it. It's just frustrating. And the, but I think we can like thing, the thing that I'm trying to come around to, because it's been in the back of my mind for a while is that the financial apocalypse to follow will be a good thing because I think there's going to be a division between people based on how they reacted to AI. And just to be clear, the people that use LLMs normally, the people like, oh, this is kind of useful. I don't think they're like, it's fine. Like whatever, like, Oh, a hundred percent. And there's a lot of people who use generative [00:48:12] Paris Marks: AI tools who also don't like what generative AI is and the impact that it has. Right. You know, [00:48:17] Speaker 1: but there are also the rest of them. And there's so many people who acted profoundly weirdly over this who are also waved away, hand waved the whole economic side, the social problems, socioeconomic problems, the deep fakes. They were just like, oh, you know, but we're getting something in return. I don't know how we are going to talk to those people in the future. I don't know how we are going. It's going to, it's like, we've permanently put on the they live glasses, just these people. I don't know how we are meant to have a conversation with someone who was super pro AI, who was super aggressively defensive of AI and they're everywhere. [00:48:57] Paris Marks: I, I, I like want to believe and, and like hope that what you're saying is accurate. But then I also think back to like the crypto moment and so many of the people who are boosting crypto just moved into boosting AI and like, you know, kind of kept the grift going and, you know, like obviously, uh, you know, we can, we can think about many people, but like someone like Kevin Roos comes to mind, obviously, uh, out of the New York times, formerly of the New York times. I can't remember exactly. He's got, he's got to NPR I think now. Okay. Interesting. Interesting. But you know, he was, he was telling us how crypto and NFTs were the next big thing and then it all collapsed. And he didn't really seem to have to apologize for that and just kind of pivoted into telling us that AI was the next big thing. And we were all going to have to use that [00:49:45] Speaker 1: and blah, blah, blah. Right. You know, there is one big difference though. And that's with crypto, it never actually got much social presence. Like crypto never, people weren't really, and aren't really transacting with crypto. There wasn't a massive build out of, there was a build out. Don't get me wrong. It's where CoreWeave came from many others. Uh, I think it's cipher mining as well, like came from crypto, but we didn't have a massive one. And we also, while we had some media pressure, we certainly didn't have everyone screaming. This is the next big thing in the way we do today. And on top of it, people didn't change their entire lives. Like regular folk, like people, not even talking Bruce who gives a shit. Uh, he's, he'll have fun in, in a year watching this burn, but he is, I'm talking more, there are reg, people who would read relatively normal who something about AI and defending the AI industry just uncorked them. And there are a lot of them and they're everywhere. There are journalists, policy makers, regular people we know. And these people have alienated themselves in their friend bases because they just became weird AI freaks. Like they, I don't know how else to put it. And the people who are defending the financials who were like, oh, I'm taking detailed notes on those who say open AI would go bankrupt. Could be talking about anyone. And it's, it's just, I don't think that it's going to be easy to put that toothpaste back in the tube because of the level of crazy they went. Like the things that people have been saying about AI economics alone are fucking crazy. [00:51:26] Paris Marks: No, I, I'm definitely on board with you. Right. And, and I definitely hope that's the way it plays out. It just seems to feel like, you know, anytime one of these things happen, if, if you're on board with what the industry says, the next big thing was supposed to be, they will like make room for you to, to have a resurrection on the other side of it. Um, but hopefully because of the scale of what we're, what we see with AI and data centers and the scale of what the collapse will ultimately have to be, you know, when that comes, uh, you know, and hopefully we're starting to see the beginning of it and of it now, um, is yeah, hopefully that means that, that it is different this time and that there will have to be a proper reckoning. [00:52:03] Speaker 1: All right. As we wrap up here, I just want to be clear. There will be a reckoning. I've been taking detailed notes and every journalist has massively backed this. I, I will not be kind. I will not be sweet and I will not let up until everyone has apologized or left the industry. You have been listening to better offline Paris. Thank you so much for joining me. I'll have a link to your book and everything, the podcast as well in the notes. Thanks for joining, man. [00:52:26] Paris Marks: Absolutely. Thanks so much. And looking forward to having you, uh, back on tech won't save us soon. Looking forward to it. Thanks for listening to everyone. [00:52:33] Speaker 1: Thank you for listening to better offline. The editor and composer of the better offline theme song is Matt Ossowski. You can check out more of his music and audio projects at Matt Ossowski.com M A T T O S O W S K I.com. You can email me at easy at betteroffline.com or visit betteroffline.com to find more podcast links. And of course, my newsletter. I also really recommend you go to chat dot where's your ed dot at to visit the discord and go to our slash better offline to check out our Reddit. Thank you so much for listening. [00:53:13] Speaker 3: Better offline is a production of cool zone media. For more from cool zone media, visit our website, cool zone media.com or check us out on the I heart radio app, Apple podcasts, or wherever you get your podcasts. [00:53:28] Speaker ?: And I'll see you next time.