The AI data center overbuild is starting to worry people — Paris Marx

[00:00:00] Isaac: people working in the tech industry and the AI industry can be very disconnected from how regular people are thinking about, you know, their everyday lives and how they kind of interact with the world. Yeah, there's an impact on climate change today. We're increasing emissions, but don't worry, we'll build the AI so good that it will figure out how we solve climate change. Well, you know what the real concern is? That it's going to kill us all. But obviously, I think that we are seeing a massive overbuilding of data centers right now, that the market feels like it's starting to reassess. Things are not going so great, right? And they're losing just money hand over fist. Hi, I'm Isaac. And joining me on the tech report today is author of the new book, [00:00:44] Paris Marks: Hyperscale, Paris Marks. Thanks for coming on. Absolutely. Great to join you in person. Yeah, good to be in person rather than just over a webcam. Yeah. I want to get into a survey I was sent, which goes into sort of the majority of people using AI are feeling like it's increasing their workload rather than reducing it. But you're not in town just to see me. So like I said, you've got a new book coming out. And I think it's just about to come out in the UK. It's already released in the [00:01:10] Isaac: US. Yeah, just came out. Just came out everywhere. It was originally supposed to come out November 5th in the UK. But we pulled it ahead to October 20th. Nice. Fair enough. So I was there for all to read. [00:01:18] Paris Marks: Yeah. Brief synopsis. It's about data centers and the physical presence sort of the impact of cloud and cloud computing. Yeah. Can you just take us through that? And then what the implications of that would be obviously for the AI data center build out? Absolutely. You know, like we have been [00:01:37] Isaac: seeing, you know, the kind of the growing concern about data centers for the past couple of years, you know, as these tech companies have been planning to build out more and more in large part to power generative AI, but not entirely. Right. And so this book really tries to go into what are these data centers? You know, why are these companies building so many right now? What are the demands that they place on the communities and the regions where they are built? And why are people pushing back against them? Why are people not happy about this? Right. Because obviously in the UK, but many other parts of the world, we're seeing people, you know, not aligned with what the tech industry's vision is here. And so, you know, I go into like the power demands, the water demands, what it means for climate change, the supply chain of all the chips that, you know, go into all these data centers. But then I also try to broaden it out. Right. And ask like, what is the economic case here? You know, what is the justification for doing this? What is the ideological case? Right. So there's the business side of it. But what do these tech billionaires think about? And what is the future that they're trying to realize by doing this? Because I think it's somewhat distinct from the economic case. And then, you know, more broadly than that, I try to say like, look, this is the future that the tech industry is trying to build. Right. This is how they see the future that they want us all to accept. But is this a future that works for us? And do these technologies need to be built in this way? Or can they be built differently? Can they be built in such a way that there are more benefits for the public, but also we don't require nearly as much computation to get those benefits? And so we don't require as many of these hyperscale data centers as the industry is planning. And so it's really trying to say, this is what they're doing. This is the implications of that. But we could do something else if we really wanted to, but we would need to change some things [00:03:21] Paris Marks: in order to realize that. One headline that has really caught my eye recently, I've struggled to sort of turn it into a whole, a whole video, which I'd love to, but I think it's great for exactly what your book kind of covers. And it's that not all, but a few quite enough to write some stories about AI researchers have picked up smoking again, because they expect, I did see this. Yeah, they expect AI to cure cancer. And I was just wondering if you think that that same kind of attitude is being used to justify stuff like that, the 20 odd percent increase in emissions that have happened, I think with Google, Amazon, Meta, and Microsoft, it's about 20%. Some of them are over a little bit under. [00:03:59] Isaac: Yeah. I think, I think Google is something like 18, Microsoft something like 24, like it's in that range. Right. And that's just a year over year increase in emissions. This is not over a long period of time, which is, which is just wild. And, and I do think so. Right. And you actually see this in some of the arguments that they make that, okay, yeah, there's an impact on climate change today. We're increasing emissions, but don't worry, we'll build the AI so good that it will figure out how we solve climate change. Right. And, and so it's the same thing where it's like, okay, they're picking up smoking and they're like, don't worry that the AI will figure it out. So I can just smoke my cigarettes or whatnot. Yeah. To me, this is just like a wild way to live or, or, you know, to, to consider how to run a business, you know, the, the impacts on society, the impacts on your life from the smoking and all this kind of stuff. Um, I, I do think it, it's another illustration of how these tech billionaires or, or even just people working in the tech industry and the AI industry can be very disconnected from how regular people are thinking about, you know, their everyday lives and how they kind of interact with the world. Um, yeah, it's, it's just wild. [00:05:04] Paris Marks: I mean, I want to move on to some of the AI increasing workloads, that kind of stuff in a bit, but kind of a question that bridges the two, given the sort of physical impact of AI infrastructure, given the impact it's having rather, I want to understand what you think AI is being, who AI is being built for, should I say? Because the obvious answer is those that are building it, but it's kind of growing at such a unsustainable and sort of unstable rate. I would kind of say, it's hard to see who's going to be a winner out of that, but then kind of get back to your point about the ideological reasons that they're building this stuff. Is that kind of the, the, why there's so many hopes pinned on AGI or super intelligence or whatever else? [00:05:45] Isaac: I think that there are a number of answers that, that you can make to that question, right? I think fundamentally, I would say, who is it being built for? I would say it's the people behind it, right? And, and you know, my argument for that would be, you go back to the founding of open AI in 2015 and Sam Altman and the people behind that, you know, then a nonprofit, now it's increasingly becoming for profit. Um, you know, we're very explicit that they wanted to build the AGI, right? The artificial general intelligence, the human brain and the computer, like this was their vision. And so as they are pursuing generative AI, their kind of ultimate goal. And you see this in the way that they talk is to build the AI super intelligence, the artificial general intelligence, you know, the term kind of shifts depending on who you're listening to. Um, but you know, it's quite clear that they are trying to realize something in particular, and we are all along for the ride and having to make sacrifices in order to realize their vision of the future. But I would also say that there are other people who are benefiting from what is going on right now, right? Yeah. So you can look at the other major tech companies, you know, the dominant ones, the Amazon's Microsoft's and Google's of the world. Um, and while they are competing on the AI front, they also have major cloud businesses, right? You know, they build a lot of data centers, they control a lot of computation that they then sell to individuals, but really to companies and to governments. Um, and for them, this is, this is great because people are using a lot more computation in order to power these generative AI tools. So this is really good for their businesses, right? So that's one, you know, group that would benefit from this. I would also say the fossil fuel industry is benefiting immensely. Um, and we have seen the kind of outward ties between the fossil fuel industry and the tech industry in a way that we have never really seen in the past, right? You know, just executives from both, uh, you know, being on stage together, appearing at conferences, you know, being very clear that their interests are aligned in the build out of data centers, but also in many cases, the, the gas fired power plants, um, in order to power those data centers. So that's another group that is benefiting from this. Another one I would say is the military. Um, you know, and if you look at, uh, governments and, you know, the question of who is this AI being built for, we're seeing the tech companies and the AI companies go hard for military contracts, um, you know, to try to make sure that the Pentagon and other, uh, you know, militaries around the world are using and relying on these technologies. Um, and we have seen it very clearly with say anthropics, uh, you know, uh, technologies being used by, uh, the U S military, uh, in the war in Iran, uh, you know, to help determine where it is going to drop bombs there. Um, I would not be surprised if technologies from other companies are being used in similar ways. Um, so yeah, so I would say even though there's a big question of like the economics of AI, who is going to profit at the end of the day, those sorts of things, I think there are still very clearly people and groups and individuals who are benefiting from, you know, whatever ends up coming of this. Yeah. So yeah, moving on to the, [00:08:49] Paris Marks: the survey that I was sent it by web summit, as I'm planning to attend in a couple of weeks, they spoke to about just shy of 1500 founders, investors, engineers, and creators about how AI is, is changing work. And there was a few different findings, but the one that obviously stood out to me was that 60% of them said that AI was, felt like it was increasing their workloads, not reducing it, which obviously goes against one of the fundamental principles of its value proposition. There's lots of other studies saying this as well. Like I've obviously gone back through in some as early as 2023 and obviously more recent ones as well. Why do you think at the very least people are feeling like AI is increasing workloads rather than reducing it? Yeah. I'm not particularly surprised to [00:09:34] Isaac: hear it. And as you say, like, we have seen this for a long time. I remember there was an Upwork survey that I believe was from 2023. And it essentially found 70 odd percent of employees found that the introduction of these generative AI tools was creating more work for them. Whereas the management and the bosses thought that it was making everything very productive and so much more efficient and stuff like that. You know, you can see a real kind of disconnect between how these different groups are feeling. And so now, you know, you kind of go forward. We have seen the technologies advance. We have seen them try to be rolled out in other places, but you still see that a lot of people are not finding that they are benefiting from it when, you know, it comes to using it at work. I would say that part of that comes from the fact that one, I think that the technologies have been rolled out in places that they're not actually good for, you know, because there has been so much hype and so much feeling that like, if your company is not rolling out generative AI, then you're going to get left behind. And so you just need to adopt it and expect your employees to use it. And you'll figure out what the real, you know, beneficial use cases are later. And of course we have seen a bunch of companies pull back on their AI investments and how they have been using AI tools over the past year. You know, Accenture, you know, kind of like a consulting work hire kind of company, you know, really stands out because they were spending a lot on AI. Earlier this year, they said that they were pulling back because they found that employees were like using it to convert PDFs to the PowerPoint slides and stuff like that, right? Paying a lot of money for not very useful, you know, way of using it. But there's other tools that do that as well. Exactly. Exactly. Like that you don't need to use generative AI for, right? That's much less computationally intensive, much less expensive. And there was just a story recently that I think they capped cloud budgets per employee at like $29 a month or something, right? So a real kind of pullback there as they are seeing what actually works. And so I would say that there's a bunch of other companies that are maybe a bit, you know, kind of slower to do that kind of investigation probably because they're smaller and that sort of a thing. But the other piece of that that really stands out is how, you know, these generative AI tools, if you're using the chatbots, they churn out a lot of information. It doesn't mean that that information is accurate, right? Yeah. And so then it's like, okay, wow, ChatGPT wrote me an essay, wrote me an email, but maybe now I need to check that over and make sure it's all accurate and double check the facts and, you know, make sure it's saying the types of things that I want it to say. And that is also work, right? And that and that does take time. So yeah, I think that there are plenty of reasons for that, but I'm not surprised. But maybe the one other thing I would add is there are also people who use these tools who feel that it makes them a lot more efficient. But in practice, when you actually look at it, it does the exact opposite. But there's that kind of bias of like, I'm using the tool, I think it's doing this for me, you know, and I don't really want to know the real impact. I think you see that a lot with like programmers in particular, [00:12:31] Paris Marks: to be honest. Yeah, I've got written down here three, the three theories that came to my head, which was it being shoved into places that it just bloats workflows, characteristic of the tech and the kind of that coming up against, I mean, even if it's producing good work, coming up against the bottleneck of the human brain, not being able to just work at the same speed. So you always just have more work to do. Absolutely. And then we're very aligned on this. Yeah. And then the expectation of productivity gains of bosses and even people, individuals that might be doing this themselves, they think they should be doing more or be able to do more because they've got this tool that is sold as this magical ability to make you do more. But how do you expect with all of those things that the thing that it comes down to is essentially burnout or I've seen it called brain fry in relation relation to AI. How how do you expect the impact of sort of LLMs that kind of the wider tool base as well to impact? Yeah, our ability to continue doing work, especially if it's something that is being outpaced by the tool you're supposed to be using, even if it's a good use case. [00:13:39] Isaac: Yeah, it's it's one of my big concerns. Right. And we have seen a growing number of studies and research looking into this, you know, kind of the questions of the cognitive impact of using these generative AI tools of kind of outsourcing the kind of mental work that that people would typically be doing to a chat bot. And so you're not retaining the same amount of information. You're not processing information in the same way. You know, you're not actually learning and developing those skills. And because you're not using that mental muscle, you're actually losing the ability to do some of these things as well. You know, I've heard about this with programmers, for example, where they're not actually coding very much anymore. They're generating a lot of code. And when it comes time to look things over or to actually have to do the coding themselves, you know, they're rusty, right? Or maybe they're starting to forget things because it's not something that they typically do. And I think that you can apply that to a lot of different sectors or types of work where if you're not actually engaging in the type of thing that you're supposed to be doing for your job regularly anymore because you're having a chat bot do it. And then maybe the chat bot goes away or, you know, again, you know, your employer looks into it and find the chat bot isn't actually very efficient. And then you need to go back to doing the task. It's like, I can't do this as good as I used to. You know, I need to rebuild these skills because I haven't been doing it in the way that I used to. And so I think that this is a problem in the workplace. I also think that it's a problem in schools, you know, in universities, you know, in grade school, all that kind of stuff, where if you are outsourcing the work of learning to a chat bot, you are not doing that yourself. And then what does that mean? You know, not just for like, if you were trying to do a training program and you were trying to learn how to do a job and then you get out of it and you can't actually do that job. So how useful is your education? But also just, you know, at a much more fundamental level, you know, if we are trying to kind of teach people, you know, the information that they need to exist in society, you know, the types of skills that we think are important just to be, you know, a human being that engages with other human beings and, you know, that is trying to like live a good life. And then, you know, you're not developing those skills because you've been relying on a chat bot. You know, I think that that can have kind of a long list of harms that we're going to be dealing with for a long period of time. And I'm not surprised to see kind of parents start to get engaged about, you know, whether chat bots should be allowed in schools or not, whether students should be using them. And, you know, I would be on their side. [00:16:05] Paris Marks: What do you imagine will happen? You mentioned Accenture and I had that written down as well. They've gone down to, yeah, $29 per employee per month of AI use from a period of, at the beginning of the year, token maxing. And I mean, there was even Trump removing access to basically Claude across the planet in the space of 90 minutes. And what happens if we sort of look past the point where we've kind of recognized that AI is useful, maybe in this area, not useful in these areas. And you take it away from those areas. And like you say, you have people that have, I mean, for lack of a better way, but grown reliant on the technology to do things that they otherwise would have been practiced at themselves, let alone the people that would have been hired during that period and had essentially zero training on how to do it. And basically been told, you don't even need to learn how to do that because we've got this thing. And then it's, yeah, like I said, it's not hard to imagine a situation where we are a few years on from an hour and we're like, a lot of this technology should not be used in this way. It's inefficient. It doesn't make sense monetarily. We'll take it away from there. And you have people then scrabbling like, I don't remember how to do my job. What [00:17:19] Isaac: happens after that? Absolutely. There are a few things I would say there, right? I want to pick up on what you were saying in just a second, but even going further than that, what about the people who are not hired because of the perception that a chatbot can do the job? You know, for example, we see in the legal profession that, you know, those kinds of early parts of the career where you are kind of, you know, basically, uh, you know, getting your sea legs or whatnot, you know, kind of like doing a lot of the work of like going through a lot of the files and, and, you know, kind of putting together the case and stuff like that, that, you know, the, the kind of top lawyers don't really do, but they get, you know, their, their underlings to do. If you're now feeding that all through the chatbot, hoping that it's all accurate and you're not getting, you know, the kind of like the early stage lawyers to, to do this kind of stuff. And so they are not developing themselves. And, you know, you're not really kind of training the next, the next group of lawyers to be able to come behind the ones who are eventually going to retire. Right. Then I think that that creates the, an even more fundamental threat, right? Because the people just aren't there because you're not hiring for those entry-level jobs anymore. And you're not doing the work of training those people in so that they're ready when, you know, when you need them. So, so I think that's one piece of it. Um, but I would say on your question of like, what if it all goes away? Um, I think that there's the issue we were just talking about, right. It was like, you know, if you haven't been doing the job for a long time, maybe you can't do it as well. But I also feel like that kind of, uh, connects to a larger conversation that I feel like we've been having the past couple of years where it's like, the U S is not a real reliable partner anymore. You know, uh, we have seen the United States already take away access to, uh, its technologies from the international criminal court. And some people who are involved with the UN and the European parliament, for example, uh, you know, those people were sanctioned by the United States and thus lost access to their Microsoft email addresses, to their Amazon accounts, like all this kind of stuff. Right. So that can easily be revoked. And we, and we have very recent examples of that. Um, you know, the, the broader question then is like, you know, one, if the technology, if, if the company fails and the technology isn't available anymore, like that's, that's one scenario. But then the other scenario is just that the United States is mad at the UK or some other country, uh, and says, you're not getting this anymore, you know, and maybe that's the AI tools, but maybe that's like, you know, the, the Microsoft services that so many of our, uh, institutions and governments are built on top of, uh, you know, are expecting to just be there. Uh, but what if that doesn't become a, what if that is not available any longer? Like it wasn't for the international criminal court. Um, and then all of a sudden, all of a sudden kind of the workings of these key institutions are threatened because of that. Like that is a fundamental problem that I feel like we're still trying to grapple with. And we see a bit like on the European continent that they're trying to look at alternatives and what this is going to look like. And, you know, part of that is developing their own AI models and that sort of a thing, but it's also just, you know, replacing Microsoft, replacing some of these other services that, that we depend on. Um, I feel like we've had those discussions quite a bit less in the UK and Canada, where I'm from, uh, in terms of like, what is the alternative approach and has much more been like, how do we get this investment from these companies here? Uh, and, you know, make sure, make sure we're part of the ecosystem. Right. [00:20:40] Paris Marks: Yeah. I have seen some stuff on sovereign AI and, and I think a really good example I saw recently was, uh, France moving almost all of its government, um, operating systems over to Linux. So they just had no more Microsoft dependency anymore. [00:20:54] Isaac: Exactly. And, you know, they, they have Mistral there, you know, they're trying to develop the kind of AI models. They have this kind of group within their government, uh, the digital something directorate that is trying to build out like an alternative suite to the Microsoft services that they rely on right now. Uh, you know, these efforts are being made in places like France, Germany, the Netherlands, Sweden. Um, but I feel like we're not doing as much of that in the English [00:21:18] Paris Marks: speaking countries for some reason. Maybe not. Um, it does raise the question though, if, cause there's, there's always the open source question and pretty much anything you talk about AI could be just say, what about open source? Yeah. And that does not have American control. That is, I mean, arguably the large reason why, um, China has been a big part in, in pushing open source where, uh, and even if you don't, I'm just struggling how to figure out how to say this, because even if you don't have open source in the situation, there'll be a lot of people watching. And I think part of what we've been talking about as well is that the, the use case for AI as yet is not as ubiquitous as perhaps Trump and American AI CEOs would like it to be to make sure that they then have that sort of lever to pull if they need to. So I struggling to figure out what the question in that is, but where, where between open source and, and, and this technology not even necessarily being as useful as it would need to be for those kinds of levers to be pulled where, where does America find itself in the situation that those two things do come to fruition that they don't succeed in, in making it so unequivocally useful that you have to use it and you can't just use open source. You can't just not use it in the first place. Yeah. I don't think that's going to [00:22:45] Isaac: happen, you know, um, for, for a number of reasons. Like I think that, you know, the, the, the Microsoft services and some of the Google services and things like that, that we rely on, those are the real threats. You know, those are the things that like, we are very dependent on and we do not have an alternative to in terms of the AI models. I'm, I'm far more skeptical that that is going to be the place where we're going to get really held up. Right. I think that once the hype sort of dies down and once the investors reconsider how they are going to make money through generative AI and data centers and all that kind of stuff, um, that we're going to have a real deep reassessment of how this whole kind of past few years has gone. Right. I, I don't think that generative AI is going to fully go away, but I think that we're going to have, um, a lot of companies and a lot of governments saying we tried to put generative AI and chatbots into places where it really didn't make sense because, you know, we were hoping that it was going to deliver this like productivity and efficiency miracle that the companies were promising us and that just didn't happen. Right. Um, and so I think what we're going to see is a pullback on these much more expensive services from companies like anthropic and open AI in favor of the open weight models that you're talking about, you know, that can be cheaper to run, but that are also often not kind of, um, the models that can do everything. And, you know, that you're expecting to be able to respond to any query, but can often be much more targeted to like a specific sector or use case or, or something like that. Right. So they don't require nearly that amount of computation. You'll need a fair amount of computation to train the model. Um, but, but after that, you know, uh, it doesn't need nearly as much for the actual use of it. Um, and I think that, you know, those models ultimately will be used in many fewer cases than where they're trying to be used today. Um, and you know, how you make money from that, like what's the economic case. That's a whole, that's a whole question. Um, but obviously I think that we are seeing a massive overbuilding of data centers right now, um, that the market feels like it's starting to reassess, you know, um, we have seen the difficulties with core weave, uh, you know, Oracle has seen pressure on its share price because of the amount of money that it has put into data centers. Um, there, there are a number of companies recently that had to pull their IPOs data center companies, um, because, you know, investors were like, you're asking for a way too much money than, than you're worth. Right. We even saw, uh, open AI reduced the amount that it is asking or like kind of the valuation that it's expecting from, from investors. Um, and of course, uh, it looks like Anthropik is trying to rush out an IPO right now, uh, recognizing that things are not going so great. Right. And, and they're losing just money hand over fist. And so, yeah, I think that there's a reassessment coming. I think, you know, parts of that reassessment are already here. Um, but when it does come, we will find that, you know, the places where generative AI is actually useful are places where we don't need these massive models that the American companies are trying to sell us and many companies and, you know, other organizations will make do, um, with less powerful and in some cases less capable, um, you know, open weight models because, you know, it's perfectly good for the task that they, that they needed for. [00:25:54] Paris Marks: After the last few months of sort of AI industry alarmism, Ipsos did a survey with a couple of other publications. I forget the name off the top of my head. Um, they surveyed, uh, I think it was two and a half thousand Americans and found that 57% of them now think that AI is, is bad for society as a whole that rises with younger age groups. But I mean, uh, from my personal experience, I have no doubt that a significant portion of that is people worrying about sort of the Skynet of it all. I'm constantly asked by people, is AI going to kill us? And I've, I've kind of shortened my response to, I don't think AI will, but the company's making it might. Yeah. Yeah. That's more or less where I've got to on it. If I'm just not trying to engage too much in the conversation, but what's your response to those kinds of questions? Cause on the one hand, you've got the fact that there are some very real concerns here. And then on the other hand, the thing that most people seem to be concerned about not really being real or at least being [00:26:56] Isaac: very theoretical. Yeah. I, I think, I think if I were to like to quickly summarize it, I would say that I don't think that AI is going to kill humanity, but if one thing is going to threaten us, it's the fact that, uh, we're basically going to, uh, you know, kind of find ourselves in a situation of like runaway climate change with all the fossil fuels that are being added right now in order to power these data centers. Right. Um, I think that if there's a real threat, uh, you know, in, on that kind of a scale, uh, from AI, it's from the emissions and the way that that's going to drive, uh, our climate in, you know, the natural disasters that we'll all be experiencing that kind of stuff. Um, because these people want to try to build the human brain in the computer or something like that, because Sam Altman feels that in the future, uh, he wants to upload his, his consciousness to the cloud or whatnot and, you know, live in a simulation or like, this is the, this is the way that these people think. Right. Um, so in terms of like bigger threat, that would be how I think about it. But I think the problem for me is that these companies and these executives have very effectively weaponized this narrative to say you have concerns about AI. Well, you know what the real concern is that it's going to kill us all. And then we shouldn't think about the ways that data centers are, you know, driving climate change and other local impacts, but also the way that, again, we were talking about chatbots are affecting how people learn and their cognitive abilities. Uh, you know, we have people who are developing dependencies on chatbots and having mental health breakdowns or even getting advice on how to harm themselves or harm others. Uh, you know, we have the deepfake crisis that we were talking about a lot earlier this year of people, um, you know, creating explicit images of women and children and the effects that that can have on the people who are victimized by it. But even beyond that, we can see the creation of deepfakes or just false information that, you know, are all across social media and trying to influence how people see the world, uh, how they see politics or, you know, just trying to scan them or, or make some ad revenue off of really engaging content or something like that. Right. There are many different motivations there. Um, and you know, the list really goes on. You can also see how this is being rolled out in public services by governments. Um, you know, in Canada, the auditor general in Ontario did a survey, um, earlier this year of the AI scribes, you know, kind of like the chatbots that are used to transcribe conversations between patients and doctors. Um, and found that in over half of the transcripts, there was incorrect information, there was made up, uh, hallucinated information, if we want to use that term. Uh, the drugs that were being recorded were wrong half of the time and they were not even saving the audio backup. And so all you had was this falsified transcript that then patients kind of, you know, medical decisions were being based on because that was their new medical history. Right. And so this is just one example of how it causes problems in service delivery in the public sector as well. And so for me, it's like these companies have really taken advantage of the narrative, you know, have shifted the narrative to let's be concerned about the sci-fi scenario that might happen in 10 years. If you believe that, you know, our computer is going to be as powerful as we believe it's going to be. And let's not talk about this whole array of problems, this growing list of problems that our technologies are actually creating and that regulation should be targeting, but that we do not want anything to be done about. And, you know, just as my final point, I would say this is a place where unfortunately, a lot of our media have really let us down, right? By buying into the kind of sensationalist narrative coming from the tech industry, rather than, you know, doing the harder work of talking about all these kind of lists of smaller, but still very important problems. You know, I would say more important than the fantasy one that they're making up, but it's like a bunch of smaller things rather than one big narrative that can shape everything. And so I'm not surprised that a lot of the public would be concerned about this thing that they, they're constantly hearing about on the news. But unfortunately that doesn't mean that that's the real problem. [00:30:58] Paris Marks: Yeah. I mean, I totally agree. I think there's the way I think of it is if it happens, we will deal with that maybe hopefully before it happens. Yeah. But right now we already have problems that we need to be dealing with and seems to have very little discussion, like AI psychosis, delusions. I mean, the AI management, the, there's so, there's so many things which we're just pretending just kind of aren't really happening, or most people don't seem to really know is happening. [00:31:24] Isaac: And I saw the other day that apparently even Elon Musk tweeted something about how he thinks he has some AI psychosis and it's like, it's good to at least see them admitting it, you know, because I do think that they are all just like kind of losing their minds up there, you know, in, in kind of billionaire world. And, and again, we need to live with the consequences, right? It's, it's wild. [00:31:46] Paris Marks: And then maybe just to end on a, on a potentially a little bit more positive note, we've only got a couple minutes, but what are, yeah, what are the positives you see coming out of, out of maybe not generative AI and the companies making it, but the technology 20 years from now when it is maybe more relied upon in ways that it should be relied upon rather than, yeah, giving people burnout and AI psychosis? [00:32:09] Isaac: You know, I, I think the real opportunity is not in generative AI, right? I think it's like, the good thing will be when all the hype around generative AI falls, and then we can actually kind of have some space for AI researchers who are interested in different approaches to say, hey, we were all focused on this really computationally intensive version of AI that was driven by fantasies of tech billionaires. But there are many other ways that we can approach this technology that can probably be more beneficial to the public, and also require far less computation than, you know, this particular version of it is requiring, right? You know, one of the things that stands out to me is how, you know, the tech industry knows that people are concerned about the climate impact of their technologies, and they'll say, ah, but, you know, we need AI to address climate change, right? But it's, again, a deceptive narrative, because, yeah, AI can help us to address climate change, to understand how emissions are being done, you know, what can be done to reduce those emissions. But those AI technologies are the ones that we've had for a long time that don't require a lot of computation, right? They're not the ones that we need all these massive data centers for. And so it's like a sleight of hand that is going on there to try to mislead us. And if I would say the thing that is really giving me the most hope in this moment is not so much what's going on with, you know, the technology or anything like that, but it's seeing how people across the United States, in communities in the UK, in many other parts of the world, are coming together, despite their political differences, you know, this is often, you know, very much across the political spectrum, to fight back against these data centers, and to say, we don't agree with this being in our communities. In some cases, we don't agree with the technologies that they're supposed to be powering. We're not alone for this vision of the future that the tech industry is trying to foist on us. And this is the thing that we can most directly push back against, right? This is a physical thing that they want to locate in our communities, that is much easier for us to challenge than say, deleting an app on our phone or something like that, right? And we see that, I would say we see, you know, a lot of that energy around data centers, but we also see it expanding, right? To the degree that we have countries now looking to pass legislation against these smart glasses, like the ones Facebook is putting out there with the cameras in them. You know, Norway is moving on that just recently. But we also see in the United States, this growing movement against the flock cameras, you know, these kind of license plate reader cameras with AI integrated in them, that again, people are saying, you know, this is like too much surveillance. There's, you know, this has gone too far. And so there does really feel like there is somewhat of a reckoning now. And I think the question for me is, you know, let's stop some of this stuff, but then also how do we turn this opposition into a broader conversation about where we go next? And, you know, what technology we want to build instead and how we build a better world with better technology in the process. [00:35:04] Paris Marks: Well, Paris Marks, thanks for taking the time. Absolutely. Great to talk to you in person. If you enjoyed today's episode and you want to hear more of The Tech Report, please consider liking and subscribing. Or so you can get episodes of The Tech Report wherever you get your podcasts.