Pełna transkrypcja odcinka Is ChatGPT actually draining Arizona dry?

come to become become the data center water guy? 00:02:09.199 --> 00:02:09.680 Yeah, yeah. 00:02:09.759 --> 00:02:12.560 It's um that that's my like universal reputation now. 00:02:12.639 --> 00:02:13.039 It's funny. 00:02:13.199 --> 00:02:19.919 Yeah, so basically in January of last year, I had already been following AI for a while and was using like AI models pretty consistently. 00:02:20.000 --> 00:02:22.080 I was using like ChatGPT and stuff like that. 00:02:22.240 --> 00:02:31.599 And I'd also been following the background hardware of the models and had a pretty good idea of like roughly how much energy and water they were using based on like public expert estimates and stuff like that. 00:02:31.759 --> 00:02:36.479 And I was finding that when I went to parties, a lot of people would be like, oh, like you use ChatGPT. 00:02:36.560 --> 00:02:38.800 Don't you know that's like terrible for the environment? 00:02:39.039 --> 00:02:45.439 And you know, I had a background both in physics and explaining to people exactly how much energy different things use. 00:02:45.520 --> 00:02:50.400 Um I've also been pretty fired up about climate and the environment for over 15 years now. 00:02:50.479 --> 00:02:53.919 And I had a lot of very strong background opinions about both of those. 00:02:54.080 --> 00:03:06.400 One general school of thought that I'm coming from is the basic idea that for dealing with the climate, it's just incredibly not promising and basically completely ineffective to focus at all on extremely tiny cuts to your personal lifestyle. 00:03:06.560 --> 00:03:15.520 Like if you're trying to change um things for the better for the environment and the climate, like the name of the game is basically the green energy transition and helping out with that however you can. 00:03:15.680 --> 00:03:19.520 And if you do want to make cuts in your personal life, there are a few like really big things you can do. 00:03:19.680 --> 00:03:22.800 You can like drive less, you can reduce your meat consumption, you can fly less. 00:03:22.960 --> 00:03:31.520 But I was coming into this debate very skeptical about the idea that like cutting something as small as a computer program, even something like ChatGPT, could have a meaningful impact. 00:03:31.599 --> 00:03:41.919 And so when people at these parties would tell me that like ChatGPT uses 10 times as much energy as a Google search, I wanted to stop them and say, like, like 10 Google searches, that's what we're worried about here. 00:03:42.080 --> 00:03:49.840 Like, if I had told you that I had done like 500 Google searches today, would your first reaction be like, oh, this guy doesn't care about the environment, he's crazy? 00:03:50.080 --> 00:03:50.879 Probably not, right? 00:03:51.039 --> 00:03:53.120 And so I was coming in with a lot of skepticism. 00:03:53.280 --> 00:03:57.280 Um, I'd written a general like overview blog post on my beliefs about this. 00:03:57.439 --> 00:03:58.719 I didn't really expect it

to go far. 00:03:58.879 --> 00:04:07.199 I was kind of expecting it to just share with a few friends and like basically make my case because I was finding I was having trouble making it at these like very loud parties that I was at. 00:04:08.719 --> 00:04:10.240 And it kind of blew up overnight. 00:04:10.479 --> 00:04:13.039 It turned out I was the first person to write about it from this perspective. 00:04:13.120 --> 00:04:19.120 Um within a few weeks it was being talked about in a lot of different places and convinced me that I should probably do a few more deep dives here. 00:04:19.199 --> 00:04:28.560 And so one thing led to another, um, read a lot more about the general state of how much energy and water individual AI prompts use, which led me to learn a lot more about data centers. 00:04:28.639 --> 00:04:41.839 And there I was also finding that my general takeaways, which like I'm not really producing much of my own knowledge and mostly relying on like expert opinion and like writing about this that I feel like just isn't really making it into mainstream conversations about it. 00:04:42.000 --> 00:04:49.360 There's a lot of sensational headlines that I think kind of misrepresent what's actually going on, and was basically finding that I was disagreeing with most of the way this was being covered. 00:04:49.439 --> 00:04:55.759 So I blogged and blogged a lot about it and eventually ended up becoming a pretty significant voice in the broader data center conversation. 00:04:55.839 --> 00:04:58.000 And so, yeah, had a lot of success last year. 00:04:58.079 --> 00:04:59.600 And it's very exciting to talk about. 00:04:59.680 --> 00:05:02.240 It's a really cool intersection of a lot of stuff that I personally care about. 00:05:02.560 --> 00:05:03.439 Well, and it matters. 00:05:03.519 --> 00:05:11.759 I've got teenagers in this house who refuse to use AI because they believe doing so is somehow morally wrong. 00:05:12.079 --> 00:05:18.800 And so they look at me as being extremely morally wrong because I really like using AI. 00:05:19.040 --> 00:05:22.800 But besides just saying it's wrong, how do you know? 00:05:23.120 --> 00:05:27.839 How do you know how much water or the environmental impact of AI? 00:05:28.079 --> 00:05:28.319 Yeah. 00:05:28.399 --> 00:05:30.560 Um, so there's a ton of stuff to say about this. 00:05:30.800 --> 00:05:43.920 In talking about this, I need to be careful that every single time you send a prompt, it's gonna have a different environmental impact because like the length of the prompt and whether you choose like a reasoning model or whether you're making an image or a video, like these all have different costs. 00:05:44.079 --> 00:05:50.879 But the theme of my writing is that all these costs still fall in like a very predictable range, and we can make some general claims about them. 00:05:51.040 --> 00:05:55.360 Like a lot of the conversation will be like, oh, there are all these mysterious costs that we don't know. 00:05:55.519 --> 00:06:01.600 And like we do know enough about them to say some basic things about roughly how big the median cost of an AI prompt is. 00:06:01.839 --> 00:06:04.800 And so we have a lot of different data on this from a lot of different places. 00:06:04.959 --> 00:06:08.720 We just understand the like hardware that AI runs on quite well. 00:06:08.879 --> 00:06:11.360 Labs aren't always super great about sharing information. 00:06:11.439 --> 00:06:13.839 Like that's kind of a separate problem in this conversation. 00:06:14.000 --> 00:06:20.000 I think there's actually some national legislation that might force them to share a lot more information about the energy and water that they're using. 00:06:20.079 --> 00:06:25.279 So I'm personally very excited for that because it'll make my like writing much easier, honestly, once we have a lot more information there. 00:06:25.439 --> 00:06:33.199 But we have a lot of general information about how much each individual prompt seems to use on average, um, with like big error bars depending on what you're doing. 00:06:33.360 --> 00:06:40.160 And so the median chatbot prompt, as far as I can tell right now, probably uses about three watt hours worth of energy. 00:06:40.319 --> 00:06:48.160 And in comparing this to other things that we do, I find that a lot of writing on this will often compare that to other incredibly tiny ways that we use energy. 00:06:48.240 --> 00:06:52.079 That'll be like, oh, it's like 10 Google searches or 100 Google searches. 00:06:52.319 --> 00:07:00.160 I think this is mostly not helpful at all because an incredibly tiny cost can be multiplied by 10 or 100 and still be very small. 00:07:00.319 --> 00:07:05.920 Like, for example, a digital clock probably uses something like one million times as much power as an analog watch. 00:07:06.079 --> 00:07:13.360 But I don't think it's evil to use digital clocks because they're still just so incredibly small that they round to zero in your daily emissions. 00:07:13.439 --> 00:07:30.720 And so just to put the Chat GPT number in context in context, using something that emits or uses three watt hours worth of energy on a standard American grid, you would probably have to send about, say, 150 to 200 prompts, give or take, to raise your daily emissions by 1%. 00:07:31.920 --> 00:07:38.639 And if you think about it, like sitting around sending like 200 prompts throughout the day, like that's gonna take up most of your time. 00:07:38.720 --> 00:07:46.399 And if you're spending all your time sending those and reading all the responses, it's very likely that you're not doing a lot of other things that also emit quite a bit. 00:07:46.560 --> 00:07:49.680 And almost everything that we do has a carbon cost, if that makes sense. 00:07:49.839 --> 00:07:54.160 So like if instead of like sitting around using ChatGPT, you like go for a run. 00:07:54.319 --> 00:08:00.720 Like even just running wears out your shoes a little bit, and your shoes

eventually need to be replaced by like new like running shoes. 00:08:00.800 --> 00:08:11.759 And if you divide the cost of like making the shoes by the amount of time that you'll spend running in them, like their carbon cost actually has a higher cost than like sending a Chat GPT prompt every minute or so. 00:08:11.920 --> 00:08:25.360 And so, like, I'm coming from a perspective where um I and a lot of other people who have thought a lot about how to actually reduce emissions are very skeptical of these small cuts because it's very likely that you'll instead just do something else that also emits quite a bit. 00:08:25.519 --> 00:08:33.759 And even if you're only replacing it with super optimized stuff, it's very unlikely that this is ever going to add up to be like a meaningful amount of your emissions. 00:08:33.840 --> 00:08:37.360 It's kind of like pausing a microwave a few seconds early to like save the climate. 00:08:37.440 --> 00:08:40.960 You know, like if I had a friend who was doing that, I'd be like, oh, there's just so much else you could do. 00:08:41.039 --> 00:08:44.480 This just seems like this kind of needless stress that you're adding to your life. 00:08:44.639 --> 00:08:50.720 And I think a lot of people who talk about this just aren't really aware of just how many other things in our life emit and just how much they emit specifically. 00:08:50.879 --> 00:08:52.320 So yeah, a lot more to say about that though. 00:08:52.639 --> 00:08:56.080 Tell me about the story of Empire of AI. 00:08:56.240 --> 00:09:04.080 This is a major book, a credentialed author, a major publisher, and you kind of debunked a theory. 00:09:04.240 --> 00:09:05.679 So talk to me about how that happened. 00:09:05.919 --> 00:09:06.159 Yeah. 00:09:06.240 --> 00:09:12.240 So Empire of AI by Karen Howe is focused on basically like bad behavior of AI companies in different ways. 00:09:12.320 --> 00:09:17.679 And I want to be clear that I'm actually like quite worried about AI in a lot of ways that don't have to do with the environment. 00:09:17.840 --> 00:09:20.480 Like I think AI is potentially going to be very powerful. 00:09:20.559 --> 00:09:22.720 There are a lot of ways that could be dangerous in the future. 00:09:22.879 --> 00:09:34.000 But and I I think Karen Howe has done a lot of like really great reporting on, say, like um, there was a lot of like drama on the open AI board a while ago that like mattered quite a bit for AI policy, and she was pretty pivotal there. 00:09:34.320 --> 00:09:37.840 But basically, my story is I was writing a lot about AI and water. 00:09:38.080 --> 00:09:45.600 We can get into like the water stuff more specifically later because there's the whole separate thing about how like AI uses a bottle of water per prompt, which I think is pretty mistaken. 00:09:45.759 --> 00:09:54.799 But I basically established myself as having a reputation for knowing a lot about water within the US specifically, but I didn't know how AI was interacting with other countries as much. 00:09:54.960 --> 00:10:09.039 And I was having a lot of people recommend this book to me and say, like, oh, she has this whole chapter on how AI is really harmful in like poorer countries that don't have the ability to regulate these AI companies in the same way, and they're like draining their water supplies specifically. 00:10:09.200 --> 00:10:13.120 And I was pretty open to that, you know, like I don't expect these companies to always behave super well. 00:10:13.279 --> 00:10:20.879 So I'd open the book specifically just to that chapter where she talks about the environmental and water impacts of AI in other countries. 00:10:20.960 --> 00:10:23.840 And right away there were a lot of alarm bells going off in my head. 00:10:23.919 --> 00:10:31.039 Like the first time she mentioned water, there was this point where she had described AI as eventually using half as much water as the United Kingdom. 00:10:31.200 --> 00:10:33.120 And I knew the study that she was drawing from. 00:10:33.279 --> 00:10:44.799 There was this really weird way that she had misrepresented what the authors had said said there, where they had specifically reported the total amount of water that will be involved in any way at any part of the process with AI. 00:10:44.960 --> 00:10:50.879 And she was comparing this to water consumed, which is specifically water that isn't given back to the local environment. 00:10:51.039 --> 00:10:57.360 Whereas like almost all water involved with AI is used to generate energy and is eventually given back to whatever local environment it's in. 00:10:57.440 --> 00:10:58.960 Like that's a whole separate conversation. 00:10:59.039 --> 00:11:01.279 But I had flagged that and I was like, oh, that's a little bit strange. 00:11:01.360 --> 00:11:03.679 Like I think that was pretty misleading, but I'll keep going. 00:11:03.840 --> 00:11:16.399 And the big error that I had found that kind of like shot me to like fame on this was finding this one point where she had said that a single AI data center in Chile was using something like a thousand times as much water as a city of 80,000 people. 00:11:16.559 --> 00:11:22.879 And these people had to, you know, like fight tooth and nail to make sure this data center didn't pop up and drain their local water supplies. 00:11:23.039 --> 00:11:34.960 And basically reading about this, I you know, like sirens were going off in my head because I just read a lot about data centers in America, and I just knew there was no data center anywhere that was using as much water as like 80 million people. 00:11:35.120 --> 00:11:37.120 Like that's just way too much, specifically. 00:11:37.360 --> 00:11:52.240 So after looking into it a little bit, I realized that she had done, made a pretty basic mistake with the units of measurement, specifically, where she had emailed like the local government about how much water they used, and they had given her a number without units, and she'd

assumed the unit was what she had asked for, which is understandable. 00:11:52.480 --> 00:12:01.519 But maybe she had like double check that because it implied this really crazy basically this error that made the data center seem a thousand times as bad as it actually was. 00:12:01.840 --> 00:12:14.799 And so, like the actual number, um, there were some other strange like things she had done where she had looked at the maximum per second water draw of the data center, um, and then assumed that it would draw that amount over the entire year. 00:12:14.879 --> 00:12:18.159 Like she multiplied that by the number of seconds in a year specifically. 00:12:18.320 --> 00:12:20.559 And most permits don't actually work like that. 00:12:20.720 --> 00:12:26.080 Like those permits are made for like absolute worst-case scenarios rather than this is how much it's actually going to draw. 00:12:26.240 --> 00:12:31.440 So my guess was that she was about like 5,000 times off the actual amount that the data center was drawing. 00:12:31.600 --> 00:12:48.960 And so it still uses like a significant amount of water, where you know it's a big part of the local community's water ecosystem, but it basically shrunk to the size of other large industries in the area, like a car factory or a pharmaceutical plant, um, also use quite a bit of water on the same orders of magnitude as data centers specifically. 00:12:49.120 --> 00:12:54.480 And all these things really pale in comparison to the main way that all countries use water, which is agriculture, actually. 00:12:54.639 --> 00:13:02.720 Like obviously agriculture is important, like we need food to eat, but there are also a lot of ways that our agriculture can be really wasteful or suboptimal with water. 00:13:02.799 --> 00:13:08.000 Like for me, my big hobby horse in America is like ethanol for you know, like growing corn for fuel. 00:13:08.159 --> 00:13:10.399 Um, there are a lot of ways that that is very wasteful with water. 00:13:10.559 --> 00:13:12.720 Animal agriculture is another that I talk about a lot. 00:13:12.879 --> 00:13:18.159 Um, but yeah, that's basically what brought me to like some mild like internet fame specifically. 00:13:18.240 --> 00:13:19.279 So I brought that up. 00:13:19.600 --> 00:13:21.759 Karen, the author, was super generous. 00:13:21.919 --> 00:13:29.279 She had like responded in detail to my post, was always very nice, and eventually actually corrected the book based on like my reading of it. 00:13:29.360 --> 00:13:31.519 So like it was very cordial interaction all around. 00:13:31.600 --> 00:13:32.559 So I was very grateful for that. 00:13:32.799 --> 00:13:33.519 That's wonderful. 00:13:33.919 --> 00:13:41.120 Let's talk about the viral claim that every Chat GPT prompt uses a bottle of water. 00:13:41.200 --> 00:13:42.960 I'd like to know where that came from. 00:13:43.120 --> 00:13:45.600 And I would love to hear about your pizza story. 00:13:45.919 --> 00:13:47.679 Oh the p yeah, yeah, definitely. 00:13:47.759 --> 00:13:47.919 Yeah. 00:13:48.080 --> 00:13:49.440 Um yeah, I love the pizza story. 00:13:49.519 --> 00:13:50.559 It's cla it's a classic. 00:13:50.799 --> 00:14:00.559 So yeah, um, basically, in 2024, the Washington Post ran this story called Writing an Email uses a whole bottle of water, like the scary rise of AI or something like that. 00:14:00.639 --> 00:14:02.159 I forgot the exact title. 00:14:02.480 --> 00:14:09.840 Reading that, um, you realize that what they're doing there is they're assuming that every email that you write requires like 10 or 20 chatbot prompts. 00:14:10.000 --> 00:14:12.879 And so it's not actually a bottle of water per prompt. 00:14:12.960 --> 00:14:14.480 It's a bottle of water per email. 00:14:14.639 --> 00:14:22.720 And email is supposed to take like 20 prompts or something, which, you know, I'm an AI power user, but I don't really take 20 prompts to write 100 words personally. 00:14:22.879 --> 00:14:24.720 Like I don't know about you, but that's a little bit much. 00:14:24.879 --> 00:14:26.639 And it was making some other strange assumptions. 00:14:26.799 --> 00:14:34.879 So it was working off of, from what I can tell, this data from this water researcher who is pretty good, but his research was something like five years old at that point. 00:14:35.039 --> 00:14:40.960 And it was assuming that the AI hardware that these chatbots run on hadn't been optimized at all since then. 00:14:41.120 --> 00:14:45.279 And in reality, it seems like chatbot hardware has been optimized quite a bit. 00:14:45.440 --> 00:14:51.679 Like every five years or so, um, the average prompt only costs about 1% the energy that it did five years previously. 00:14:51.759 --> 00:15:00.799 It's just radically dropping off because there's just so much money and interest in reducing AI companies and like energy bills as much as possible, basically, because they spend a lot on that. 00:15:01.200 --> 00:15:03.200 That makes perfect sense. 00:15:03.519 --> 00:15:10.320 It's like I've been talking about, I'm sure people will come up with a better way to do whatever, become more efficient. 00:15:10.399 --> 00:15:14.240 There's too much money incentive not to become more efficient. 00:15:14.399 --> 00:15:16.080 Why is nobody talking about that? 00:15:16.399 --> 00:15:26.240 Yeah, it's it's very it's I think a lot of people just struggle a lot with this because they'll see these correct statistics that AI is going to use more and more energy over over time in total. 00:15:26.399 --> 00:15:33.840 And I think what they're missing there is that what that means, what that's actually talking about, is that we're getting radically new models that do radically different things. 00:15:34.000 --> 00:15:36.159 And those come with a new energy cost, definitely. 00:15:36.320 --> 00:15:41.120 Like coding agents and things like that use way more energy than like simple text prompts. 00:15:41.279 --> 00:15:48.159 But the main reason AI is

growing so much is just that it's being used so much more, not that each individual prompt is getting more expensive over time. 00:15:48.240 --> 00:15:57.519 Like a lot of the smartest, most educated people in the world have all this huge financial incentive to go and work on optimizing these already incredibly optimized AI chips specifically. 00:15:57.679 --> 00:16:10.559 And what I think a lot of people miss in the water conversation is they look at this data center, they don't really know what's going on inside, and they'll just see it consuming huge amounts of water and assume that whatever's happening in there must be very wasteful because like altogether it's using a lot of water. 00:16:10.720 --> 00:16:19.440 And what's actually happening is that the data center is moving the water in the vicinity of these computer chips that have been heated up by being used a lot, specifically. 00:16:19.600 --> 00:16:27.840 And so the water just absorbs some of the heat, it flows to these cooling towers, and it slowly evaporates to reduce some of the heat, and then it circulates again. 00:16:28.000 --> 00:16:49.519 And like, if you think about it, just like flowing water near the vicinity of hot computers is definitely like a way to use water, but I don't really expect it to be a really gigantic part of our total water cost as a country, because other ways we use water are like dumping huge amounts of it into the ground to grow food and stuff like that, or like using it in all kinds of like manufacturing processes and things like that. 00:16:49.679 --> 00:16:57.039 So anyway, if you'd like, I could just circle back around and explain like the actual picture with like individual prompts and like why it's not a full bottle of water, if that would be useful. 00:16:57.360 --> 00:16:58.639 Yes, I would like to understand that. 00:16:58.799 --> 00:16:59.039 Thank you. 00:16:59.279 --> 00:17:00.080 Yeah, great, definitely. 00:17:00.320 --> 00:17:11.519 Other um the other weird thing about that chatbot study that the Washington Post was citing was that it was also including the off-site water that was used to generate the electricity that a chatbot prompt uses. 00:17:11.680 --> 00:17:27.039 And if you look into it, a big part of their water cost was lakes dammed by um hydroelectric power plants and looking at the amount of water that evaporates off that lake, like that evaporated water is actually a big part of the quote unquote like bottle of water that a chatbot uses specifically. 00:17:27.200 --> 00:17:33.839 And I personally think it's really good that we use hydropower plants, and I'd much rather we use those than like fossil fuel plants specifically. 00:17:33.920 --> 00:17:40.720 And so this statistic implies that like using a chatbot prompt in Washington state is really bad because it has a really high water cost. 00:17:40.799 --> 00:17:45.920 And like what it ignores is that this is all renewable energy in an area where water is already plentiful. 00:17:46.079 --> 00:17:52.319 So, anyway, like all of this went into this really massive, exaggerated misunderstanding of how much water an AI prompt uses. 00:17:52.559 --> 00:18:03.680 The actual cost, if you look at more recent studies from either the labs themselves or third-party estimates, is it's usually around one milliliter of water, which is about one five hundredth of a bottle of water specifically. 00:18:03.839 --> 00:18:07.359 And that can either shrink or go grow depending on the size of the model. 00:18:07.599 --> 00:18:16.799 But AI prompts never really rise to be a significant part of your water footprint because there's just so much that you do in your everyday life, and all of us do in our everyday lives, that use water. 00:18:16.960 --> 00:18:20.000 Like, even just generating electricity uses a huge amount of water. 00:18:20.160 --> 00:18:30.960 Where, like, if you have a digital clock on your bedside table, generating the electricity for that digital clock probably uses something around like a gallon of water per month or something, it like consumes a full gallon. 00:18:31.440 --> 00:18:34.240 And I think for most people that's very it's invisible, right? 00:18:34.319 --> 00:18:45.119 Because like the only water that we see in our daily lives are like the water that we drink or the like water in our showers and stuff like that, or maybe our lawns, or if you're in Arizona, yeah, maybe you like have like an irrigated lawn or something like that. 00:18:45.200 --> 00:18:52.480 I I don't think most people know that something like 90% of their total water footprint happens outside of their home, especially in like the food that we eat. 00:18:52.640 --> 00:19:05.039 So, like going on to the pizza story specifically, like I've been vegan for 10 years, partly to reduce my own water footprint by quite a bit, because like something like half of the food that we grow in America eventually goes to feed animals that we then eat. 00:19:05.119 --> 00:19:07.440 So there's a huge amount of extra water cost there. 00:19:07.599 --> 00:19:14.319 And I remember um a few months ago, uh, I had a friend who was in like a pizza place specifically, and he overheard these two teenagers talking. 00:19:14.400 --> 00:19:24.640 And one of them was like, our mutual friend claims to be an environmentalist, but it's crazy that she says that because she still uses Chat GPT, and the other person was like, Oh, I know, it's so crazy. 00:19:24.799 --> 00:19:29.039 And then the very next thing that they did was they went over and they ordered a meat lover's pizza specifically. 00:19:29.200 --> 00:19:32.079 And I'm not saying that that was like the worst thing to do. 00:19:32.160 --> 00:19:40.079 I'm not saying that they're not environmentalists right now, but I do think that's a really clear sign that they're just missing where like the really big water baddies are in society. 00:19:40.160 --> 00:19:52.480 And if like you're mad about Chat GPT, like every time you

eat meat, that's something like tens of thousands of prompts worth of water that you're consuming that you didn't have to if you eat it, like if you ate like a plant-based um option instead. 00:19:52.640 --> 00:20:03.680 And again, like we all have to make trade-offs in the way that we want, but like being hyper litigious about these incredibly tiny individual water costs, I think is just kind of like a sad distraction from where the real baddies are. 00:20:04.000 --> 00:20:08.319 So, why is this misinformation so prevalent? 00:20:08.559 --> 00:20:13.039 Who is it that stands to gain from this obvious propaganda? 00:20:13.279 --> 00:20:13.680 Yeah. 00:20:13.920 --> 00:20:19.119 Um, I'm a little bit skeptical that it's motivated by like anyone specifically gaining anything. 00:20:19.200 --> 00:20:26.079 I think that a lot of people are kind of a little bit hungry for the sense that like their personal consumption choices have a lot of power. 00:20:26.319 --> 00:20:33.279 And the idea that like you can identify this bad guy in your life and purge it and be like, anyone else using this, we have to stop that. 00:20:33.359 --> 00:20:37.200 And it makes you feel really good and it makes you feel like you're on a team with other people. 00:20:37.359 --> 00:20:38.799 And that's one aspect of it. 00:20:38.880 --> 00:20:45.680 Um, I think that on another weird aspect of this is that you know, data centers are popping up all over, like they're popping up in Arizona and other places. 00:20:45.839 --> 00:20:58.079 And I think it's totally understandable that people look at these really strange new buildings that are popping up and think, man, like this is this weird new thing that's using a huge amount of electricity and water compared to a lot of other things in my area. 00:20:58.319 --> 00:21:01.599 Whatever's happening in there must be really inefficient and wasteful. 00:21:01.759 --> 00:21:12.240 And I think what they're missing there is that data centers are actually just ridiculously hyper-optimized, and every individual thing that happens inside uses shockingly little energy and water by the standards of like computing. 00:21:12.319 --> 00:21:19.440 Like if you ran an AI prompt on your computer, um, it would be way less efficient and use way more energy to complete, basically. 00:21:19.599 --> 00:21:34.640 But data centers are weird because they just pile so many of these individually very tiny things into one specific building that in total it looks really gigantic and it is really gigantic, but it's not because the individual things happening inside are especially wasteful. 00:21:34.799 --> 00:21:37.599 In fact, they're like really hyper-optimized to not be wasteful. 00:21:37.759 --> 00:21:41.599 So I think a lot of people are getting this misconception from the size of the data centers. 00:21:41.759 --> 00:21:53.279 And then I guess the very last one is that like more as more and more people freak out about this, people writing about this in the media notice that, like, oh, this gets eyeballs, you know, like the media's in kind of bad straits right now. 00:21:53.359 --> 00:21:58.000 And like anything that can get a lot of eyeballs on what they write and read, like that's a big deal. 00:21:58.160 --> 00:22:05.839 And so people will write more kind of Like doom and gloom stories about like the most extreme stories of data centers, which I think are mostly but not always overblown. 00:22:05.920 --> 00:22:08.160 Like there are definitely a few places where data centers have caused some harm. 00:22:08.319 --> 00:22:11.680 And then there's this kind of like vicious cycle of like it hypes more people up. 00:22:11.839 --> 00:22:12.640 They want to read more. 00:22:12.799 --> 00:22:15.680 Some more stuff gets published, it hypes people up, and it just continues. 00:22:15.839 --> 00:22:17.839 So these are some of my like personal theories. 00:22:17.920 --> 00:22:19.200 Um I don't really know about that. 00:22:19.440 --> 00:22:21.359 Like there's a lot of could be happening here. 00:22:21.599 --> 00:22:23.039 I'm in Gilbert, Arizona. 00:22:23.200 --> 00:22:29.920 And the reason that we are talking about this is because the water in Gilbert, Arizona has gone crazy. 00:22:30.160 --> 00:22:35.519 Water bills have doubled and tripled in the last year or two here in Gilbert. 00:22:35.680 --> 00:22:44.640 And one of the things you've done is looked at the difference between the data center water usage and golf course water usage. 00:22:44.799 --> 00:22:52.720 I'd like to understand if you think there's any relationship between our water bills skyrocketing and data centers first. 00:22:52.880 --> 00:22:53.440 Let's just do that. 00:22:53.680 --> 00:22:53.920 Yeah. 00:22:54.000 --> 00:22:58.079 So I mean, like, I'm definitely not an expert on like Arizona water policy. 00:22:58.160 --> 00:23:00.880 So I don't want to dive in here and be like, oh, this is exactly what's happening. 00:23:01.039 --> 00:23:08.880 I am very skeptical that data centers are playing a significant part of Arizona's water bill rise to the point that I'm not sure they're contributing at all, really. 00:23:08.960 --> 00:23:10.240 And I can go into the reasons why. 00:23:10.400 --> 00:23:17.119 The first one is just if you actually look at how much of Arizona's water data centers specifically are actually using, including new data centers. 00:23:17.279 --> 00:23:18.640 We're I'm making some assumptions here. 00:23:18.720 --> 00:23:23.920 Like if we just assume that water prices are only determined by how much water is being used, which isn't actually correct. 00:23:24.000 --> 00:23:25.680 There are all these other factors that go into play. 00:23:25.839 --> 00:23:42.160 But just starting with that, from what I can tell, like data centers, even all the new ones together, seem to use only like 0.1% of Arizona's water altogether, just like the vast majority is going to irrigated agriculture, including like alfalfa farming, which mostly

goes to feed animals. 00:23:42.319 --> 00:23:49.839 And from what I can tell, Arizona currently spends about 30 times as much water on golf courses as it does on like data centers specifically. 00:23:50.000 --> 00:24:00.720 However, like data centers are bringing in way more tax revenue in total than golf courses are because you know they're basically these like building-sized computers that are associated with one of the most lucrative new industries ever. 00:24:00.880 --> 00:24:08.640 And like if you just kind of passively let that run in your town, it does generate a lot of tax revenue per unit of water used. 00:24:10.799 --> 00:24:11.440 Oh sorry, go ahead. 00:24:17.680 --> 00:24:26.400 Basically, they're just very lucrative, and they are like this is basically entirely a decision of like local, city, and state governments specifically. 00:24:26.559 --> 00:24:29.759 So each individual state will have very different rules around this. 00:24:29.920 --> 00:24:49.599 I think a lot of states both understand that like data centers do draw a lot of like water and power, and so they threaten to be somewhat unpopular, but they're also just generating such massive amounts of revenue, and like AI companies are just really hungry for a lot more data centers that they realize we can get a lot of like mutually beneficial deals out of this, and so we'll set taxes that way. 00:24:49.759 --> 00:25:01.759 This doesn't happen everywhere because a lot of municipalities are starting to compete with each other for data centers, and so there are definitely places where tax revenue is way lower than it could otherwise be because they're trying to draw more data centers to be built there. 00:25:01.839 --> 00:25:05.759 So I have to admit, I don't really know the tax situation with like new Arizona data centers. 00:25:05.839 --> 00:25:12.480 And so my ask for your audience isn't to assume like everything is totally fine, but my ask is basically to like look into this for yourself. 00:25:12.559 --> 00:25:18.160 And instead of thinking about data centers as like, oh, there's this new weird use of water that we should never ever consider. 00:25:18.319 --> 00:25:22.240 I'd basically like people to consider like, how does this compare to other things? 00:25:22.400 --> 00:25:28.160 If the data center were already built, would this be the first thing we would want to shut down for the sake of water conservation? 00:25:28.319 --> 00:25:44.160 And like usually when I look at that, it's very hard to understand why people would want to target data centers first because they're just generating so much local revenue, so much revenue, especially per unit of water used, that they actually seem like quite water efficient if that's your goal as a policymaker to bring in a lot of tax revenue for your area. 00:25:44.319 --> 00:25:45.440 Yeah, I tend to say about this. 00:25:45.519 --> 00:25:54.480 I was gonna add that the the main way that data centers at least threaten to raise water bills is requiring a lot of new infrastructure to be built for delivering the water. 00:25:54.559 --> 00:25:58.160 So it's not actually really about like the net amount of water used. 00:25:58.240 --> 00:26:06.880 It's usually more like passing on the costs of like building all this all this new pipe system and stuff like that and like wastewater treatment, like passing that on to consumers. 00:26:06.960 --> 00:26:09.440 And that's usually a problem for water utilities specifically. 00:26:09.599 --> 00:26:17.839 And a lot of that can be avoided with good policy, like making sure that data centers internalize that cost is not too difficult because, again, they're kind of really desperate to be built. 00:26:18.400 --> 00:26:20.480 All these people are coming in and they really want it to be built. 00:26:20.640 --> 00:26:24.480 And so local communities do have quite a bit of control over what they can get out of them. 00:26:24.559 --> 00:26:27.119 And there's potentially a lot of ways that could be mutually beneficial. 00:26:27.519 --> 00:26:36.480 How do we, as citizens and taxpayers, know if our city is recouping the infrastructure investment from the data centers? 00:26:36.799 --> 00:26:41.519 Um, I think this is a little bit beyond me because it changes so much based on individual places. 00:26:41.759 --> 00:26:43.279 I would recommend reading a lot more. 00:26:43.359 --> 00:26:52.160 So I live um in Washington, D.C., pretty close to Loudoun County, Virginia, um, which has the far and away the largest cluster of data centers anywhere in the world. 00:26:52.400 --> 00:26:58.960 Like a huge amount of the world's data center capacity is like within a like half an hour drive of where I'm at right now. 00:26:59.119 --> 00:27:04.000 And the city government has been dealing with data centers for like 15 years and are mostly quite excited about it. 00:27:04.079 --> 00:27:08.799 And they have all these like policies and recommendations that are working for them in some ways. 00:27:08.960 --> 00:27:10.400 There are still issues with the data centers. 00:27:10.480 --> 00:27:12.480 Like more and more people are drawing attention to noise. 00:27:12.640 --> 00:27:16.000 Different, like data centers are just governed in different ways in different counties. 00:27:16.079 --> 00:27:18.000 And so I can't speak too confidently to that. 00:27:18.079 --> 00:27:24.559 But I would just recommend that people look into this a lot and not automatically assume, like, oh, we have to, you know, destroy the data center. 00:27:24.640 --> 00:27:42.000 Whatever happens, we have to make sure it doesn't get built because you could potentially be missing out on a huge amount of local revenue for the cost of the water use of like any other local industry in the area, as long as you're able to make sure that your politicians are keeping track of at least the things that I worry about the most, which are um like noise. 00:27:42.160 --> 00:27:44.880 Like noise pollution does seem to

be a pr pretty big issue with data centers. 00:27:44.960 --> 00:27:47.599 So just making sure they're set pretty far back from homes. 00:27:47.759 --> 00:27:52.720 Air pollution might be an issue, although like data centers in Arizona seem to be powered a lot more by solar power, which is great. 00:27:52.880 --> 00:27:54.079 And then water use. 00:27:54.240 --> 00:28:01.039 Most places that I know of, I'm not really aware of a single municipality that's actually seen their water bills rise because of data centers yet. 00:28:01.119 --> 00:28:06.400 That might change, but it does seem like most utilities have thought about this a lot and understand just how to build this stuff out. 00:28:06.480 --> 00:28:15.599 And again, I don't want to speak too confidently to your specific area because I don't know like the policies around every single thing, but we definitely encourage people to just look at like how this compares to where this has been done well in the country. 00:28:15.839 --> 00:28:17.839 Let's talk about water more generally. 00:28:18.079 --> 00:28:33.519 There's there's general anxiety over whether or not there's going to be enough water, especially in places like Arizona when everything's getting hotter and we're kind of far down on the list of Colorado River recipients. 00:28:33.759 --> 00:28:38.640 What can you tell us about overall broader water availability issues? 00:28:38.960 --> 00:28:39.279 Yeah. 00:28:39.519 --> 00:28:46.720 My reading of the situation with the Colorado River is that so much of it is just downstream of specific agricultural decisions and regulations. 00:28:46.880 --> 00:29:03.200 And there are a lot of places in the country where it can be quite difficult to regulate the behavior of farmers because they're just so like politically popular in general that it can be the second that you suggest, like, oh, maybe farmers should pay a little bit more for the water they use, they'll like swoop in and be like, oh, do you not want any food? 00:29:03.279 --> 00:29:05.440 Like, where you're your one lifeline to this. 00:29:05.599 --> 00:29:15.759 And I think this is kind of a silly political equilibrium where there are a lot of ways that like farming could use more optimized like practices and like water could be just priced better to represent that. 00:29:15.920 --> 00:29:22.480 Um, thinking about Phoenix and Arizona more generally, with respect to you all, like I think Arizona is like a really amazing place. 00:29:22.720 --> 00:29:31.200 So largely like this large, you know, civilization constructed in the middle of the desert by like pulling in huge amounts of water from like hundreds of miles away. 00:29:31.359 --> 00:29:37.759 And, you know, I think that's both pretty cool, but also like it needs to come with some acknowledgement that this is already a very weird situation. 00:29:37.839 --> 00:29:45.599 And if you're looking around at this gigantic city that's just like grown out of the sand, and you look around and you're like, oh, but that data center, that's going way too far. 00:29:45.680 --> 00:29:47.279 That's exactly where we draw the line. 00:29:47.519 --> 00:29:49.839 I think that's a little bit suspicious, honestly. 00:29:50.079 --> 00:30:00.559 Like if you're thinking about a drought situation and thinking about water getting more expensive over time, like it's an important question to ask: do we just not want any additional industry in the area at all? 00:30:00.720 --> 00:30:05.680 And if we do want industry in Arizona that's new and different, what are the criteria for that? 00:30:05.839 --> 00:30:12.240 Like, how should we think about whether, like how different industries use water and what we should allow and what we should not allow? 00:30:12.400 --> 00:30:24.960 And I think my argument, which is a little bit counterintuitive, is that because data centers generate more tax revenue per unit of water used, in a lot of ways they're actually quite water efficient as industries compared to especially things like golf courses. 00:30:25.119 --> 00:30:35.200 Like if Arizona replaced all of the golf courses it has and all that water instead went to data centers, you'd probably get like 50 times as much tax revenue as you get from all golf courses combined. 00:30:35.359 --> 00:30:39.599 And I think that's kind of a sign that like data centers might not be the main baddie to target. 00:30:39.680 --> 00:30:49.519 But again, you know, if someone's looking around the situation in Arizona and thinking about just how crazy water could be in the future, I can also totally understand wanting to just completely hit the brakes until you all figure it out. 00:30:49.599 --> 00:30:52.480 Like there's a lot that I just don't understand about the local ecosystem. 00:30:52.640 --> 00:31:01.200 I think my main worry is just that like people are drawing these very arbitrary lines around data centers just because they're new and weird rather than they're like objectively the absolute worst thing that we do with water. 00:31:01.440 --> 00:31:03.519 I want to understand this tax revenue thing more. 00:31:03.680 --> 00:31:11.119 If I walk into a resort and I pay to stay at a resort, I'm paying specific taxes that go to the local economy. 00:31:11.759 --> 00:31:20.640 How does a data center with a bunch of chips in it and probably some employees create local city-specific revenue? 00:31:20.960 --> 00:31:21.200 Yeah. 00:31:21.440 --> 00:31:23.839 Um, so let's see, there's a lot to say about this. 00:31:23.920 --> 00:31:30.079 So basically, like the data center is where the main economic activity of a lot of companies is happening. 00:31:30.160 --> 00:31:35.039 Like everything that we do online, almost everything happens within a data center specifically. 00:31:35.200 --> 00:31:42.000 And now AI companies specifically are also extremely they're just deeply in need of these very large data centers for all these different reasons. 00:31:42.160 --> 00:31:45.680 You need

especially large data centers to be able to train the newest models. 00:31:45.759 --> 00:31:48.079 Like we can get into training, that's a whole separate conversation. 00:31:48.240 --> 00:31:49.680 Um, but also just running them. 00:31:49.920 --> 00:31:59.440 And like these companies are completely blowing up, you know, like they're all like if you look at their revenue streams over time, it's balanced out by just how much they're spending on new data centers. 00:31:59.519 --> 00:32:02.400 And they're really desperate to build new data centers specifically. 00:32:02.480 --> 00:32:05.359 So it's mainly like there are different types of data centers too. 00:32:05.440 --> 00:32:12.240 They're like hyperscaler data centers that are mainly owned by these individual giant companies like Amazon or Meta and things like that. 00:32:12.319 --> 00:32:19.200 And they're also smaller co-location um data centers that um can host like individual activity from a bunch of different internet companies. 00:32:19.359 --> 00:32:26.799 But basically, you have these giant companies that really want to build and use a community's resources and use their land and things like that. 00:32:26.880 --> 00:32:30.640 And they come in and like most communities can say, like, hey, we want something for this. 00:32:30.799 --> 00:32:38.880 We want some exchange for all the power and water you're using, in addition to just the utility bills that you're like paying. 00:32:39.039 --> 00:32:41.519 And so like that's where the tax revenue comes from. 00:32:41.599 --> 00:32:52.400 And in terms of like at least physical surface area, like the average data center will generate something like 10 to 15 times as much tax revenue as, say, like a Walmart supercenter or something, which is like a comparable size. 00:32:52.559 --> 00:33:04.640 And so the main reason communities are able to like extract this money from these big AI companies is just that they're just so lucrative and they're very desperate for more data centers everywhere because there's just so much demand for their product. 00:33:04.799 --> 00:33:11.920 And even now, like there are definitely limits in how much AI companies can provide because like they have a limited amount of compute. 00:33:12.079 --> 00:33:26.160 Like there are times when you know I use Claude a lot specifically, and there are times when like Claude is just much more slow, or like it specifically has to limit how much I can do at a specific time because too many people are using it, which is basically another way of saying there's not enough data center capacity for it. 00:33:26.640 --> 00:33:27.519 Does that answer your question? 00:33:27.839 --> 00:33:28.319 Sort of. 00:33:28.480 --> 00:33:31.119 So like let's say there's a Google data center. 00:33:31.279 --> 00:33:37.279 That money is going to Google or Anthropic, and Anthropic is not in Gilbert, Arizona. 00:33:37.440 --> 00:33:47.599 So does the Gilbert, Arizona revenue depend on the negotiated deal that

our city does with the specific data center built in the city? 00:33:48.000 --> 00:33:53.519 So the data center, like the data centers that I'm talking about, like hyperscalers specifically, are owned by these specific companies. 00:33:53.599 --> 00:33:58.880 So it would involve like a specific negotiation with like Google or Meta or Anthropic or something like that. 00:33:59.039 --> 00:34:04.960 So yeah, it's specifically with like the company as a whole, rather than as I understand it, the specific data center. 00:34:05.119 --> 00:34:05.839 I might be wrong about that. 00:34:05.920 --> 00:34:07.599 I would have to circle back actually. 00:34:07.759 --> 00:34:12.320 But yeah, this is a situation where like I start to scrape up against the edge of what I know. 00:34:12.400 --> 00:34:21.280 I know a lot of basic statistics about like, oh, you know, like this is how much data centers pay in taxes per unit in different areas, but the specific tax structures I'm a little bit less familiar with. 00:34:21.599 --> 00:34:22.000 Got it. 00:34:22.159 --> 00:34:22.480 Okay. 00:34:22.960 --> 00:34:26.960 I want to look into more, you know, the hyperlocal issue. 00:34:27.119 --> 00:34:30.719 Like what is what is what is this data center doing for Gilbert? 00:34:30.800 --> 00:34:32.800 Because I think people would be very interested to know. 00:34:32.960 --> 00:34:37.599 And they might be a whole lot more excited about a potential data center coming in. 00:34:37.760 --> 00:34:44.800 Because right now you've got people showing up at city council meetings all over the state saying, don't put a data center in. 00:34:45.119 --> 00:34:45.519 Yeah. 00:34:45.679 --> 00:34:46.079 Yeah. 00:34:46.239 --> 00:34:56.559 And, you know, I want to like be very sympathetic to those people because I think most of what the average person reads about these things are like these are these things that come in and destroy local communities. 00:34:56.639 --> 00:34:57.840 They sap your water. 00:34:58.000 --> 00:35:02.960 They're associated with these like far-off tech billionaires who don't care about you, and it's crazy. 00:35:03.119 --> 00:35:05.039 And I think the truth is actually quite far from that. 00:35:05.199 --> 00:35:09.440 I'm not entirely sure why this media narrative has popped up in the way that it has. 00:35:09.599 --> 00:35:12.159 Would be happy to share maybe after for the show notes or something. 00:35:12.239 --> 00:35:15.280 There are a bunch of like articles that I'm a big fan of that are still quite critical. 00:35:15.360 --> 00:35:22.960 You know, they're not saying data centers are always great, but they're just a lot of examples where, you know, like local town governments have just gotten a lot of like really great resources from them. 00:35:23.119 --> 00:35:30.960 Like again, Loudoun, near me, um, both has the most data centers of any county and is also the wealthiest or one of the wealthiest counties in the country. 00:35:31.039 --> 00:35:32.639 It's at least top five per capita. 00:35:32.800 --> 00:35:35.280 And the locals there get quite a bit from the data centers. 00:35:35.360 --> 00:35:39.599 Like they get thousands of dollars every year in like property tax abatements alone. 00:35:39.760 --> 00:35:42.880 Like a huge amount of the county's revenue comes from the data centers specifically. 00:35:43.039 --> 00:35:49.440 And if you start to think about the data centers as like, this is the main industry in the area, you know, like it's generating revenue. 00:35:49.599 --> 00:36:00.480 They're basically these huge computers that run and do cause some problems, especially with noise, but they're also kind of preferable to other industries that maybe use like a lot of chemicals or emit a lot. 00:36:00.639 --> 00:36:04.239 Like, I would probably rather have a data center next to me than a car factory or something. 00:36:04.320 --> 00:36:08.639 And like they can still generate a huge amount of revenue just by kind of passively running in the background. 00:36:08.880 --> 00:36:16.880 So like I do worry that like if the average person listening to this is just incredibly skeptical of what I'm saying, like I'm asking you, like, don't trust me. 00:36:16.960 --> 00:36:29.760 Like, I'm not trying to paint data centers as this perfect thing that is always great, but just look into instead of like the scary headlines, just try to find information about like how do these actually interact in like communities that are taxing them and governing them well. 00:36:29.920 --> 00:36:50.559 And like in most of them, it seems like people are actually quite excited, at least, about the benefits that they bring to the tax base specifically, especially in places that maybe couldn't handle a big influx of new people for a new industry that does require a lot of jobs, but could just use like a large computer running somewhere kind of far separated from everybody that is associated with this really large business that's very easy to tax. 00:36:50.880 --> 00:36:53.840 We don't have a ton of time, but I would like to ask you about something. 00:36:54.159 --> 00:37:00.480 We had a recent election for our local energy company, SRP, the Salt River Project. 00:37:00.559 --> 00:37:11.599 And there was this huge political campaign between turning point USA backed kind of fossil fuel board members versus the clean energy candidates. 00:37:11.760 --> 00:37:16.400 And the clean energy candidates won a majority of the seats. 00:37:16.559 --> 00:37:30.960 But I think a lot of us who know that are struggling to understand what is this fight between what what is happening between the clean energy folks and why are why would turning point back people to be on a board? 00:37:31.280 --> 00:37:31.519 Yeah. 00:37:31.599 --> 00:37:37.199 So is the question like what why are people still like so eager for like new fossil fuel stuff specifically? 00:37:37.599 --> 00:37:39.039 I just want to make sure I'm understanding it right. 00:37:39.280 -->

00:37:52.880 Well I I honestly don't really understand, I guess, the politics behind why there would be a war between clean energy and turning point, specifically. 00:37:53.199 --> 00:37:53.519 Okay. 00:37:53.679 --> 00:37:59.679 I mean, like, like I guess, you know, I I'm not conservative myself, so I can't speak to turning point too much. 00:37:59.840 --> 00:38:08.159 My my main read of this is that like the conservative criticism of like a lot of renewable energy stuff is that it does in a lot of places raise electricity prices. 00:38:08.320 --> 00:38:10.320 Like I think in the long term it's going to be good. 00:38:10.400 --> 00:38:22.000 I'm very excited about the green energy build-out, but I do need to be kind of honest with a lot of people talking about this, being like, yeah, like green renewable energy in a lot of places actually still isn't totally cost competitive with fossil fuels. 00:38:22.159 --> 00:38:29.360 And if I didn't believe in climate change at all, I would probably be much more excited about continuing to use fossil fuels because they are just objectively cheaper. 00:38:29.519 --> 00:38:36.000 If anything, like the thing that makes climate change so difficult is that fossil fuels are basically magic. 00:38:36.159 --> 00:38:41.679 Like you can put this tiny amount of liquid into your giant car and propel it for like 500 miles. 00:38:41.760 --> 00:38:52.159 And like, you know, obviously like the gas has been like refined and there's all this stuff that's happened to it, but it's still kind of this amazing miracle of nature that we have all this stuff on the ground that just contains so much energy. 00:38:52.320 --> 00:39:02.960 And so, like, the big challenge of renewable energy is building out an infrastructure and building out materials that can kind of store and transmit similar amounts of energy that don't also emit basically. 00:39:03.039 --> 00:39:14.719 And so my assumption here is that like Turning Point is like mainly excited about this because like they do want America to use a lot more energy specifically, because energy being cheap is just associated with a lot of good things and economic growth. 00:39:14.880 --> 00:39:29.039 And I think for me as someone who's more worried about the climate, I think that the trade-offs between growth that's as fast as possible and energy bills that are as cheap as possible, and just like investing a significant amount amount of money to make sure that like the planet doesn't warm to a dangerous degree. 00:39:29.199 --> 00:39:39.039 Like, I'd honestly rather like purely for building out the green energy grid, like I'd unfortunately rather Americans' energy bills be a little bit more expensive for a little while to fund that specifically. 00:39:39.199 --> 00:39:41.039 And I think a lot of this is where that's coming from. 00:39:41.199 --> 00:39:44.719 And obviously, like both parties have like money interests behind them. 00:39:44.880 --> 00:39:52.639 I think, you know, being honest, like I have to assume that there's some fossil fuel interest money behind like the push for like maximized fossil fuels. 00:39:52.719 --> 00:39:53.920 But does that answer your question? 00:39:54.000 --> 00:39:54.639 Am I getting to the other? 00:39:54.800 --> 00:39:56.960 Yeah, you did an excellent job explaining that. 00:39:57.039 --> 00:39:57.920 I really appreciate it. 00:39:58.320 --> 00:39:59.119 Last question. 00:39:59.440 --> 00:40:14.480 If you were the king of the world, yeah, and you wanted to do one to three things that would have the greatest positive impact on the world, what would they be? 00:40:14.800 --> 00:40:17.039 The greatest possible impact just on the world, period. 00:40:17.599 --> 00:40:18.880 It's such a broad question. 00:40:19.119 --> 00:40:19.760 Oh my god. 00:40:19.840 --> 00:40:20.960 Okay, yeah, let's see. 00:40:21.119 --> 00:40:31.679 Um, so I'm actually like, I have some very extreme beliefs about animal farming, actually, where I'm pretty worried that the main way that we treat and farm the animals that we consume is really disastrously bad. 00:40:31.840 --> 00:40:39.280 And so I think my first thing would be a global ban on specifically industrial animal agriculture with factory farming, specifically. 00:40:39.440 --> 00:40:46.239 Like if people wanted to raise their own chickens or their own cows and their, you know, pastures, like as long as they're treated well, that's fine. 00:40:46.400 --> 00:40:49.519 But being honest, that would all also really skyrocket the price of meat. 00:40:49.599 --> 00:40:52.639 But I think this is something that I personally worry about quite a bit. 00:40:52.880 --> 00:40:54.960 Um oh man, there are just so many options here. 00:40:55.119 --> 00:40:55.760 It's difficult. 00:40:56.079 --> 00:41:01.199 Probably much stronger governance of AI as it might exist in about five or 10 years or so. 00:41:01.280 --> 00:41:06.719 Um, I'm quite worried about the long-term implications of AI getting more and more intelligent and powerful. 00:41:06.800 --> 00:41:09.199 There are just a lot of downstream potential risks of that. 00:41:09.360 --> 00:41:17.599 And so probably a lot more like oversight of AI labs and just like kind of mutual agreements between countries to like govern this. 00:41:17.760 --> 00:41:19.360 That's kind of like very hand-wavy. 00:41:19.440 --> 00:41:24.000 God, and then probably a lot more global investment into um like renewable energy technology. 00:41:24.159 --> 00:41:38.880 Like I've been, you know, I'm I'm very much not like I'm very skeptical of a lot of what the government of China does specifically, but like I have to tip my hat that they are producing a huge amount of um like electric cars and solar panels and just flooding the markets with them and making them incredibly cheap. 00:41:39.039 --> 00:41:42.639 And I would like more people, even just like private companies, to be doing that. 00:41:42.719 --> 00:41:46.639 And I guess like if I'm the king of the world, we're already in kind of an authoritarian situation. 00:41:46.719 --> 00:41:48.559 So I might just direct more of that specifically. 00:41:48.639 --> 00:41:50.400 So those would be my top three. 00:41:51.360 --> 00:42:02.480 Andy, thank you so much for coming on this show and for sharing all of your knowledge and you know, taking the time to educate us on these things that are kind of new and um a little scary. 00:42:02.559 --> 00:42:03.039 So thank you. 00:42:03.280 --> 00:42:04.239 Yeah, this was a blast, Billy. 00:42:04.320 --> 00:42:04.719 Had a great time. 00:42:04.800 --> 00:42:05.679 Yeah, I appreciated the questions. 00:42:05.840 --> 00:42:06.880 This was this was all fantastic. 00:42:07.199 --> 00:42:07.440 Thank you. 00:42:07.519 --> 00:42:08.239 We'll talk soon. 00:42:08.400 --> 00:42:08.639 Bye. 00:42:09.280 --> 00:42:10.239 Yep, bye. 00:42:12.320 --> 00:42:14.159 Thanks for joining me on Law Labs. 00:42:14.320 --> 00:42:18.719 If today's episode gave you something to think about or something to act on, I'd love to hear from you. 00:42:18.880 --> 00:42:24.079 Don't forget to subscribe, leave a review, and share the show with someone building the future of law. 00:42:24.239 --> 00:42:28.800 Until next time, keep experimenting, keep evolving, and I'll see you back in the lab.