Latest / Faster, Please! — The Podcast / 💧⚡🖥️ A deep-dive on data centers: My interview with AI researcher Andy Masley
Transcript
- 0:00SPEAKER_01: Welcome to Faster Please,
- 0:01SPEAKER_01: the podcast, the audio-visual companion to my Faster Please newsletter.
- 0:05SPEAKER_01: I'm your host, Jim Pethokoukis.
- 0:07SPEAKER_01: My guest this episode is Andy Mazely.
- 0:11SPEAKER_01: Andy is an independent writer and researcher,
- 0:13SPEAKER_01: supported by a grant from Coefficient Giving,
- 0:16SPEAKER_01: formerly Open Philanthropy,
- 0:17SPEAKER_01: who does deep dives on AI, among other interesting things.
- 0:21SPEAKER_01: Most recently,
- 0:22SPEAKER_01: Andy has been writing a lot and at length about what's true
- 0:26SPEAKER_01: and what isn't regarding data centers.
- 0:28SPEAKER_01: Writing that has generated considerable attention and I think has been
- 0:31SPEAKER_01: a valuable part of the public debate.
- 0:34SPEAKER_01: We had a really great chat.
- 0:35SPEAKER_01: I think you'll enjoy it.
- 0:36SPEAKER_01: Andy, welcome to the podcast.
- 0:38SPEAKER_00: Yeah, thanks, James.
- 0:39SPEAKER_00: Great to be here.
- 0:40SPEAKER_01: Before we get into all things data center,
- 0:45SPEAKER_01: or if you prefer data center, I'm not sure your preference, we're good with both.
- 0:51SPEAKER_01: I just want to touch on a few of your many other really interesting, excellent writings.
- 0:55SPEAKER_01: One essay really struck me is one he wrote called AI as Art Curation.
- 0:59SPEAKER_01: Asking whether AI art involves real creativity, that mostly misses what's interesting about it.
- 1:06SPEAKER_01: Now I make AI images, my little AI images.
- 1:10SPEAKER_01: I offer every essay using mid-journey.
- 1:13SPEAKER_01: And I think they tend to, you know, I'm biased.
- 1:15SPEAKER_01: I think they tend to be pretty good, pretty interesting.
- 1:19SPEAKER_01: I enjoy making them quite a bit, actually.
- 1:23SPEAKER_01: But is there value in what I'm doing?
- 1:25SPEAKER_01: Because some people say there's no value in what you're doing.
- 1:27SPEAKER_01: And maybe you shouldn't be doing it all because you're taking jobs away from artists.
- 1:31SPEAKER_01: So I probably would not be, you know, hiring artists to make images for my essay.
- 1:36SPEAKER_01: What is the value in my effort?
- 1:39SPEAKER_00: Yeah.
- 1:39SPEAKER_00: I mean, the first value is just, you're having fun.
- 1:41SPEAKER_00: I think it's really underrated that a lot of these tools are incredibly pleasant to use.
- 1:45SPEAKER_00: Like I'm pretty addicted to mid journey personally.
- 1:47SPEAKER_00: And I think part of the reason why I like it so much that often gets very confused
- 1:51SPEAKER_00: and is the reason I wrote the post actually,
- 1:54SPEAKER_00: is that the reason I like it isn't necessarily
- 1:56SPEAKER_00: that I feel like I'm personally involved in some active artistic creation.
- 2:00SPEAKER_00: I don't feel at all.
- 2:01SPEAKER_00: Like I'm drawing or like writing a song or something when I use these AI tools,
- 2:05SPEAKER_00: but there does feel like something else that's happening that's very similar
- 2:09SPEAKER_00: to other ways that I express my taste more broadly.
- 2:12SPEAKER_00: Like if I'm putting a playlist together or even just like,
- 2:15SPEAKER_00: I don't use Pinterest,
- 2:16SPEAKER_00: but if a lot of people We're putting together like Pinterest boards
- 2:18SPEAKER_00: for like general vibes that they like.
- 2:20SPEAKER_00: I feel like what I'm doing with mid-journey as well is kind of like reaching out
- 2:23SPEAKER_00: into the vector space and finding some vibes that I like and putting them
- 2:27SPEAKER_00: in one place and being like, these are things that I personally really gel with.
- 2:32SPEAKER_00: And I've had a surprising amount of like really great moments with people
- 2:35SPEAKER_00: where I think they perceive me mainly as, oh, you're the data center guy.
- 2:37SPEAKER_00: And then I'll show some mid-journey art that I made.
- 2:39SPEAKER_00: And they're like, oh, you...
- 2:41SPEAKER_00: To be clear, I...
- 3:01SPEAKER_01: While there is some level of creativity,
- 3:04SPEAKER_01: I guess,
- 3:05SPEAKER_01: I would not call myself a creator or an artist,
- 3:09SPEAKER_01: but what struck me is I think I am saying something about who I am
- 3:14SPEAKER_01: and what I find interesting to look at.
- 3:18SPEAKER_01: Many of my images, they're not the typical...
- 3:22SPEAKER_01: Like chat GPT, give me a picture of a robot, like in an animated style.
- 3:26SPEAKER_01: They all kind of look the same.
- 3:28SPEAKER_01: I think mine aren't like that.
- 3:29SPEAKER_01: I think it says like, gee, he's making an effort.
- 3:31SPEAKER_01: These are the kinds of like art styles.
- 3:34SPEAKER_01: I mean, I'm not sure there's, I'm conveying a tremendous amount of information.
- 3:38SPEAKER_01: I think I'm, I'm conveying something.
- 3:39SPEAKER_01: I'm adding something.
- 3:41SPEAKER_00: Yeah.
- 3:42SPEAKER_00: And I think this is another point of confusion that a lot of people have
- 3:45SPEAKER_00: about AI art because they're mostly exposed to honestly kind of like sloppy,
- 3:50SPEAKER_00: like art that's made by people whose taste I just don't personally vibe with.
- 3:54SPEAKER_00: Honestly,
- 3:54SPEAKER_00: I had made this long,
- 3:55SPEAKER_00: the separate long list of Just like,
- 3:57SPEAKER_00: here are some AI images I've made that I've liked.
- 3:59SPEAKER_00: And I think that actually changed a significant number of readers' minds
- 4:02SPEAKER_00: about what AI art can do.
- 4:05SPEAKER_00: Because as soon as you see someone who's using mid-journey to create images
- 4:09SPEAKER_00: that you are actually much more interested in,
- 4:12SPEAKER_00: and you find maybe more subtle or just pleasant than the kind of slop
- 4:16SPEAKER_00: that we're all exposed to on our timelines every now and then,
- 4:18SPEAKER_00: I think that can really open people up and make them realize
- 4:20SPEAKER_00: that these tools Are actually much more powerful than they might expect.
- 4:23SPEAKER_00: So yeah,
- 4:24SPEAKER_00: I think that for people who haven't played around with MidJourney especially,
- 4:26SPEAKER_00: I'd very strongly recommend jumping on.
- 4:29SPEAKER_01: There is more than one artistic style than like the default chat GPT image,
- 4:36SPEAKER_01: like animated style, which is repulsive, frankly.
- 4:40SPEAKER_01: Yeah, I hate it.
- 4:41SPEAKER_01: Yeah, no, it's terrible, yeah.
- 4:43SPEAKER_01: I'm just going to quickly skip through a couple here.
- 4:44SPEAKER_01: You wrote a great piece, and obviously I'm biased.
- 4:48SPEAKER_01: I'm here in Washington, D.C., About Waymo, I think a lot of us would love to have Waymo here.
- 4:54SPEAKER_01: They've tested the car here.
- 4:56SPEAKER_01: They're not actually, it's not legal.
- 4:58SPEAKER_01: You still have to have a human driver in the car, kind of defeats the purpose.
- 5:02SPEAKER_01: Is that,
- 5:03SPEAKER_01: when you look at the problems getting that approved here,
- 5:07SPEAKER_01: do you view it as purely kind of a,
- 5:09SPEAKER_01: you know, one-off Washington, the city bureaucracy?
- 5:13SPEAKER_01: I mean, I'm sure that's a factor,
- 5:15SPEAKER_01: but I also view it as kind of like, A small front in the war against AI.
- 5:19SPEAKER_01: Yeah.
- 5:21SPEAKER_00: And this is definitely how it's talked about in a lot of places where DC,
- 5:26SPEAKER_00: a lot of times,
- 5:27SPEAKER_00: at least by my reading,
- 5:28SPEAKER_00: was kind of fitting the general pattern of other very blue cities
- 5:33SPEAKER_00: where labor is very powerful.
- 5:34SPEAKER_00: Basically,
- 5:35SPEAKER_00: there are a lot of constituencies that are very anti-Waymo,
- 5:38SPEAKER_00: not always in the best interests of the city either,
- 5:40SPEAKER_00: you know,
- 5:41SPEAKER_00: like There are a lot of people who will correctly see that,
- 5:43SPEAKER_00: oh, this is going to massively disrupt, say, the rideshare economy.
- 5:47SPEAKER_00: I think that's just generally true.
- 5:49SPEAKER_00: And even people in public transit services who maybe correctly see Waymo as the tip
- 5:56SPEAKER_00: of a spear that is eventually coming for even automated buses or trains
- 6:00SPEAKER_00: or things like that.
- 6:00SPEAKER_00: And I think that might be correct, honestly, but I also think that in considering...
- 6:05SPEAKER_00: Like the questions of whether to legalize autonomous vehicles,
- 6:08SPEAKER_00: we need to mainly consider the interests of the people who are trying to get around
- 6:12SPEAKER_00: the city like that's those are the main beneficiaries of public transit
- 6:15SPEAKER_00: and obviously like the jobs that public transit creates or that ride Yeah.
- 6:39SPEAKER_00: Or similar things.
- 6:40SPEAKER_00: Because, like, you know, making better public transportation also lowers, like, rideshare usage.
- 6:45SPEAKER_00: And so, like,
- 6:46SPEAKER_00: I think a lot of the general debate about this was pretty confused for a while,
- 6:49SPEAKER_00: but the situation's looking much better now in D.C. specifically.
- 6:52SPEAKER_00: I think, like,
- 6:52SPEAKER_00: Charles Allen specifically and city government has just made a lot of good decisions
- 6:55SPEAKER_00: in the last month or two that have been pleasantly surprised by.
- 6:59SPEAKER_01: This is an incredibly parochial question to ask you.
- 7:03SPEAKER_01: Like, A year from now,
- 7:05SPEAKER_01: am I going to be able to take a Waymo around the city or from a near suburb into
- 7:10SPEAKER_01: the city?
- 7:11SPEAKER_00: I wish I knew.
- 7:12SPEAKER_00: I think it's going to be a while.
- 7:13SPEAKER_00: I would expect at least a year, but I'm not sure of the specifics, honestly.
- 7:17SPEAKER_01: And should I view this as just another concerning sign in our ability of a society
- 7:25SPEAKER_01: to deal with what could be a very powerful...
- 7:29SPEAKER_01: General purpose technology and AI and our ability to think like rationally
- 7:34SPEAKER_01: about it rather than kind of have this knee jerk.
- 7:38SPEAKER_01: No job can be lost.
- 7:40SPEAKER_01: We can have this incredibly powerful technology.
- 7:42SPEAKER_01: Great.
- 7:43SPEAKER_01: It's going to do wonderful things.
- 7:44SPEAKER_01: Fantastic.
- 7:45SPEAKER_01: Speed up economic growth.
- 7:46SPEAKER_01: Great.
- 7:46SPEAKER_01: But God forbid it costs us one single.
- 7:49SPEAKER_01: That is the red line.
- 7:51SPEAKER_01: And I worry that,
- 7:52SPEAKER_01: again, that there's more of that in our society than I had hoped for or I feared.
- 7:59SPEAKER_00: Yeah, 100%.
- 8:00SPEAKER_00: And I've definitely seen,
- 8:01SPEAKER_00: you know,
- 8:01SPEAKER_00: like there were a lot of hearings in Boston about Waymo that I was also following
- 8:05SPEAKER_00: where this would come up a lot,
- 8:06SPEAKER_00: where it's like, oh,
- 8:06SPEAKER_00: like any job lost at all is a complete travesty and we can't think about the upsides
- 8:11SPEAKER_00: if this displaces even one rideshare worker.
- 8:14SPEAKER_00: I think that's very strange.
- 8:15SPEAKER_00: Like, obviously I also want to give a lot of sympathy to people because,
- 8:18SPEAKER_00: you know, I really don't understand the long-term future of AI.
- 8:22SPEAKER_00: I haven't really been fully convinced by any one economic story of it.
- 8:25SPEAKER_00: Like one week I'll nod along to, oh, you know, it'll create...
- 8:28SPEAKER_00: An additional 0.1% GDP growth and disrupt a few industries versus, oh man,
- 8:32SPEAKER_00: maybe this will just actually be a huge problem for labor's relative influence
- 8:37SPEAKER_00: over society or just people's ability to get jobs at all in the longterm.
- 8:41SPEAKER_00: I'm kind of a massive agnostic there.
- 8:43SPEAKER_00: So I do want to give a lot of,
- 8:44SPEAKER_00: you know,
- 8:45SPEAKER_00: deference to people's broader concerns here,
- 8:47SPEAKER_00: but I think when we're making individual trade-offs like this.
- 8:50SPEAKER_00: We also just have to be pretty clear about where the benefits and drawbacks are.
- 8:54SPEAKER_00: And while there are lots of industries that I would maybe worry
- 8:56SPEAKER_00: about completely disrupting,
- 8:58SPEAKER_00: like I have to say the number of deaths involved in driving
- 9:01SPEAKER_00: and also just like people's ability to get around very dense cities are both just
- 9:05SPEAKER_00: so important for me as general problems that like,
- 9:08SPEAKER_00: I want to hand the robots this one before we like march forward.
- 9:11SPEAKER_00: So yeah, there I'm just entirely gung ho on like let the Waymos have it.
- 9:15SPEAKER_01: And this, this may help us kind of segue into data centers.
- 9:20SPEAKER_01: If you're going to make an argument that whatever your concerns are about land
- 9:26SPEAKER_01: or water or electricity prices, there are these upsides to data centers.
- 9:31SPEAKER_01: We don't want to miss out on those potential upsides and not just the upsides
- 9:36SPEAKER_01: of like tax revenue in your town or whatever jobs they do create,
- 9:41SPEAKER_01: but one problem is one,
- 9:42SPEAKER_01: if you believe in those upsides, we haven't seen them yet, right?
- 9:46SPEAKER_01: You can,
- 9:47SPEAKER_01: I guess,
- 9:48SPEAKER_01: Try to point to like current real world harms from the data centers,
- 9:52SPEAKER_01: which maybe in some places they are feeding or soon will into electricity prices
- 9:56SPEAKER_01: and they do use water.
- 9:58SPEAKER_01: So not only are those upsides yet to arrive, I'm not sure what those upsides are exactly.
- 10:04SPEAKER_01: I don't know if it's going to add,
- 10:07SPEAKER_01: you know,
- 10:07SPEAKER_01: a tenth of a percentage point to,
- 10:10SPEAKER_01: you know,
- 10:11SPEAKER_01: total factor productivity growth, or it's going to add like 10 percentage points.
- 10:16SPEAKER_01: Like, I don't really know.
- 10:18SPEAKER_01: I can come up with, you know, sort of cases for all those scenarios.
- 10:22SPEAKER_01: So it's like a, you know, a double question.
- 10:25SPEAKER_01: One, we, you know, we think they will happen, but they haven't happened.
- 10:30SPEAKER_01: And we're not really sure what it's going to be.
- 10:34SPEAKER_01: So it does make the argument a little more difficult to make
- 10:38SPEAKER_01: that there are these upsides and whatever the downsides are,
- 10:42SPEAKER_01: they're going to be worth it.
- 10:45SPEAKER_00: Yeah, a ton of different directions to go.
- 10:47SPEAKER_00: Yeah, I do.
- 10:48SPEAKER_00: After, you know,
- 10:49SPEAKER_00: a year of doing a lot of deep dives on data center situations,
- 10:52SPEAKER_00: it's become hard for me not to think actually,
- 10:54SPEAKER_00: which is kind of a contrarian take,
- 10:56SPEAKER_00: but I think is actually just I feel like I have my feet firmly on the ground,
- 10:59SPEAKER_00: that data centers actually mostly do seriously benefit the very local communities
- 11:03SPEAKER_00: that they're built.
- 11:04SPEAKER_00: I think that like,
- 11:05SPEAKER_00: the tax revenue,
- 11:06SPEAKER_00: people have a really warped perception of how much they pay,
- 11:08SPEAKER_00: and it's actually quite a bit.
- 11:10SPEAKER_00: We can get into that later.
- 11:11SPEAKER_00: And I just don't think in a lot of places,
- 11:13SPEAKER_00: citizens would actually make this trade between like,
- 11:15SPEAKER_00: oh, we could use like 3% less land for a mere billion dollars a year.
- 11:20SPEAKER_00: Like that just doesn't really fly in most places.
- 11:22SPEAKER_00: And like, obviously, I'd imagine that for skeptical listeners.
- 11:25SPEAKER_00: And so I think you're making the case then.
- 11:49SPEAKER_01: So if you're making the case that we should generally be building data centers and
- 11:54SPEAKER_01: if companies think there's an economic value to doing so,
- 11:57SPEAKER_01: they should be allowed to do so.
- 12:01SPEAKER_01: I guess the primary argument Really can't be that these things are a doorway to
- 12:07SPEAKER_01: a wondrous future.
- 12:09SPEAKER_01: The argument really has to be, maybe,
- 12:11SPEAKER_01: but for now,
- 12:12SPEAKER_01: they're paying for stuff in your town,
- 12:15SPEAKER_01: and they will create some jobs,
- 12:17SPEAKER_01: and the downsides you worry about, in many cases, just aren't very significant.
- 12:22SPEAKER_00: Yeah, 100%.
- 12:23SPEAKER_00: And I think that like there are downsides that if you somehow like hid
- 12:28SPEAKER_00: the word data center,
- 12:29SPEAKER_00: but included all other information about them where it's like, Hey,
- 12:32SPEAKER_00: would you trade these specific trade-offs in your community using X amount of water
- 12:36SPEAKER_00: and land and stuff for I guess there's...
- 13:02SPEAKER_01: Three criticisms.
- 13:03SPEAKER_01: I usually think of electricity and I think of water, but I guess land's another one.
- 13:07SPEAKER_01: What can you say about each of those that are empirical,
- 13:12SPEAKER_01: fact-based about the impact of data centers on each of those three things?
- 13:17SPEAKER_00: Yeah.
- 13:17SPEAKER_00: Before I say that,
- 13:18SPEAKER_00: I might just quickly flag to listeners that if you are skeptical of me,
- 13:21SPEAKER_00: I'm really not asking you to come in And,
- 13:24SPEAKER_00: you know,
- 13:24SPEAKER_00: trust everything that I'm saying here and instead obviously do your own research.
- 13:27SPEAKER_00: But I do want to flag that I think that the average person's understanding
- 13:30SPEAKER_00: of this is actually based on a very strange game of telephone
- 13:34SPEAKER_00: where there are all these media articles about say like,
- 13:36SPEAKER_00: oh, the construction of a data center polluted very local well water.
- 13:40SPEAKER_00: And then through that game of telephone that gets translated to
- 13:43SPEAKER_00: the normal operational water draw of data centers polluted
- 13:46SPEAKER_00: an entire municipal water system.
- 13:48SPEAKER_00: And that second one just isn't true.
- 13:50SPEAKER_00: And so My ask here is,
- 13:52SPEAKER_00: isn't that you immediately believe everything that I say,
- 13:55SPEAKER_00: but also that you just read these articles and see what are
- 13:58SPEAKER_00: the actual claims being made about data centers and do they actually line up
- 14:01SPEAKER_00: with this perception that you have?
- 14:03SPEAKER_00: That's currently kind of the general agreement among educated people,
- 14:06SPEAKER_00: which I happen to believe is incorrect.
- 14:08SPEAKER_00: So with all that, with that flag, just going down my own perception here.
- 14:12SPEAKER_00: The water use of data centers is significant by the standards of normal industry.
- 14:17SPEAKER_00: Data centers use significant amounts of water.
- 14:19SPEAKER_00: A lot of them are comparable to,
- 14:21SPEAKER_00: say, pharmaceutical plants or car manufacturers, which also use a lot of water.
- 14:26SPEAKER_00: They're often dwarfed by the main way that we use water in society,
- 14:30SPEAKER_00: which is agriculture, just the vast majority.
- 14:32SPEAKER_00: Of American water withdrawals and consumption ultimately goes to agriculture,
- 14:37SPEAKER_00: and ultimately most of America's actual water problems are downstream
- 14:40SPEAKER_00: of how we grow food basically,
- 14:43SPEAKER_00: and obviously we need food,
- 14:44SPEAKER_00: food is very important,
- 14:45SPEAKER_00: but this also means that we should We expect most water problems in America
- 14:49SPEAKER_00: to be involved with farms in one way or another.
- 14:51SPEAKER_00: And so like the question is,
- 14:53SPEAKER_00: how do data centers actually fit into our general water situation in America
- 14:58SPEAKER_00: and should we expect them to cause problems?
- 15:00SPEAKER_00: And so I'm trying not to do like a simple whataboutism thing where I'm not trying
- 15:03SPEAKER_00: to say like, oh, because farms are so much bigger, data centers are worse.
- 15:07SPEAKER_00: Or data centers are fine, I mean.
- 15:09SPEAKER_00: But the general scale of just how big of a difference there is between, say,
- 15:13SPEAKER_00: data centers and farms is just so gigantic that it kind of feels like if you had
- 15:18SPEAKER_00: a budget of $100 a day and you started spending an extra Like how much is that going
- 15:23SPEAKER_00: to impact your total access to money?
- 15:25SPEAKER_00: As an example, like their comparisons do start to matter.
- 15:28SPEAKER_00: And in my investigation so far,
- 15:30SPEAKER_00: I haven't actually found a single place where the normal operation
- 15:33SPEAKER_00: of data centers have actually harmed people's regular access to drinking water.
- 15:38SPEAKER_00: And again,
- 15:38SPEAKER_00: I'd imagine that people might have a lot of like- Like in the Southwest,
- 15:43SPEAKER_00: like, okay, maybe not right in the East
- 15:45SPEAKER_01: or in the,
- 15:45SPEAKER_01: I don't know, Washington State with, you know, wet play.
- 15:48SPEAKER_01: But like,
- 15:48SPEAKER_01: I think like if it's going to have an impact,
- 15:53SPEAKER_01: We should for sure see it right like in Arizona.
- 15:56SPEAKER_00: 100%.
- 15:56SPEAKER_00: Yeah.
- 15:57SPEAKER_00: And like even in Arizona,
- 15:59SPEAKER_00: best I can tell is that data centers are consuming like one 30th of the water used
- 16:04SPEAKER_00: by just golf courses in the state specifically.
- 16:06SPEAKER_00: And the thing is they're also generating way more tax revenue per unit of water used.
- 16:11SPEAKER_00: And so my general claim here is that if you are a city planner and you're trying
- 16:14SPEAKER_00: to use water maximally efficiently for your community,
- 16:17SPEAKER_00: like it actually seems like data centers are at least very preferable
- 16:20SPEAKER_00: to golf courses there.
- 16:22SPEAKER_00: And again, this isn't to say, oh, that means they're fine.
- 16:24SPEAKER_00: There's nothing wrong.
- 16:25SPEAKER_00: But I think they're currently consuming such a small portion of
- 16:28SPEAKER_00: the state's total water that it seems kind of strange to hyper focus on them for
- 16:32SPEAKER_00: the same reason that it would be kind of strange to hyper focus
- 16:35SPEAKER_00: on other really tiny marginal changes to the environment
- 16:39SPEAKER_00: where you have these other really massive bad guys
- 16:41SPEAKER_00: that could really limit what they're doing.
- 16:42SPEAKER_00: And so, yeah,
- 16:43SPEAKER_00: my general claim is
- 16:44SPEAKER_00: that like data centers just aren't really meaningfully contributing
- 16:46SPEAKER_00: to Arizona's water problems.
- 16:49SPEAKER_01: I'm not sure to the extent that you've tracked this,
- 16:51SPEAKER_01: but I keep hearing that,
- 16:53SPEAKER_01: you know,
- 16:53SPEAKER_01: companies that,
- 16:53SPEAKER_01: you know,
- 16:54SPEAKER_01: they're saying like,
- 16:55SPEAKER_01: to the extent that you think we're using too much water, listen,
- 16:58SPEAKER_01: we're recycling the water or coming up with,
- 17:00SPEAKER_01: you know, different ways to cool and the technology is changing.
- 17:03SPEAKER_01: And to the extent you even think it's a problem today,
- 17:06SPEAKER_01: it'll be less of a problem like five years from now.
- 17:09SPEAKER_00: Yeah.
- 17:09SPEAKER_00: So I actually worry about this for one of my least popular takes I think is
- 17:13SPEAKER_00: that I'd actually on average prefer that data centers in a lot of places
- 17:17SPEAKER_00: in America use more water because the trade-offs between using more
- 17:21SPEAKER_00: or less water actually involve using more energy to compensate.
- 17:24SPEAKER_00: Because ultimately both of these things are being used for cooling specifically.
- 17:28SPEAKER_00: And most systems that use less water end up using more energy too.
- 17:33SPEAKER_00: And between like,
- 17:34SPEAKER_00: you know,
- 17:34SPEAKER_00: if a data center is built in a place with,
- 17:36SPEAKER_00: you know, like mild to moderate water stress.
- 17:39SPEAKER_00: And we have the option like, oh,
- 17:41SPEAKER_00: should it use a little more basically like adding another acre of a farm in
- 17:44SPEAKER_00: the area?
- 17:45SPEAKER_00: Or should we emit significantly more?
- 17:48SPEAKER_00: Like I kind of always want to choose water.
- 17:49SPEAKER_00: I think I am like just pretty worried about climate change and emissions
- 17:52SPEAKER_00: and local air pollution is also an underrated problem with them.
- 17:55SPEAKER_01: You don't want them to fire up more gas turbines.
- 17:58SPEAKER_01: Exactly.
- 17:58SPEAKER_01: Yeah.
- 17:59SPEAKER_00: Yeah.
- 17:59SPEAKER_00: They're already such a disaster that like, yeah,
- 18:02SPEAKER_00: gas turbines are kind of one of the main bad,
- 18:04SPEAKER_00: noxious things about data centers that I will totally sign off on as being
- 18:07SPEAKER_00: a problem.
- 18:08SPEAKER_01: If I were to tell you that energy is coming from a small modular reactor,
- 18:11SPEAKER_01: you'd
- 18:12SPEAKER_00: be better with it.
- 18:13SPEAKER_00: The dream one day, you know, but I can't tell how far off those are.
- 18:17SPEAKER_00: So yeah, but we'll see what happens there.
- 18:19SPEAKER_01: All right.
- 18:19SPEAKER_01: Well, I feel somewhat reassured by, now I'm less familiar with the land issue.
- 18:25SPEAKER_01: What, they use up too much land?
- 18:27SPEAKER_01: What is the land issue?
- 18:30SPEAKER_00: This has really blown up in the past few months in a way I really didn't expect, actually.
- 18:34SPEAKER_00: Willie Nelson actually just posted this thing that got 100,000 likes on Twitter
- 18:37SPEAKER_00: about how, don't give them an inch of our farmland.
- 18:40SPEAKER_00: I came in and just tried to estimate, okay,
- 18:44SPEAKER_00: so we know roughly how much energy data centers are projected to use by 2030
- 18:49SPEAKER_00: and their rough size.
- 18:51SPEAKER_00: As a proportion of how much energy they use.
- 18:53SPEAKER_00: And so we can use that to make a general projection of how much land they'll take up.
- 18:56SPEAKER_00: And my basic finding is that data centers themselves,
- 18:59SPEAKER_00: like the physical buildings,
- 19:00SPEAKER_00: will take up something like 1 15th of the land that we currently dedicate
- 19:03SPEAKER_00: to Christmas trees in America,
- 19:05SPEAKER_00: which, you know, isn't nothing, but it's also like incredibly small.
- 19:08SPEAKER_00: And the parcels of land can be much bigger.
- 19:11SPEAKER_00: Data centers need to be kind of set off from where people live because again,
- 19:15SPEAKER_00: living right up against them is annoying.
- 19:16SPEAKER_00: So they do have pretty large parcels of land,
- 19:18SPEAKER_00: but again, like America just has so much land and just so much empty space.
- 19:22SPEAKER_00: This general concern has popped.
- 19:26SPEAKER_00: And I think people just really have no internal map
- 19:30SPEAKER_00: of just how much farmland America has, sometimes even too much.
- 19:33SPEAKER_00: Among other things,
- 19:35SPEAKER_00: the American government will pay farmers a lot to not farm additional units of land.
- 19:39SPEAKER_00: Basically, there's a lot of worry that that will drive prices down too much.
- 19:43SPEAKER_00: Also, something like 30% of I love that.
- 19:47SPEAKER_00: To me, that's the amazing angle, really.
- 19:53SPEAKER_01: Are you worried?
- 19:55SPEAKER_01: Is that your concern?
- 19:56SPEAKER_01: We're going to lose our ethanol
- 19:58SPEAKER_00: generating capacity?
- 20:00SPEAKER_00: Yeah, it's pretty disorienting.
- 20:03SPEAKER_00: And yeah,
- 20:03SPEAKER_00: I think a lot of people don't realize that there's a government mandate
- 20:06SPEAKER_00: that companies buy ethanol, even though they might not necessarily need that.
- 20:10SPEAKER_00: This takes up an amount of corn that's something like the size of Pennsylvania.
- 20:14SPEAKER_00: And my claim here is that we don't Really need that farmland and if
- 20:17SPEAKER_00: a very tiny amount of that is voluntarily sold to a data center by the farmer
- 20:22SPEAKER_00: for something like 10 times as much as what it's worth,
- 20:24SPEAKER_00: which is often happening in cases like this.
- 20:26SPEAKER_00: I don't think this is a unique tragedy and is actually pretty good for the farmer.
- 20:30SPEAKER_00: It also benefits the farms around the data center
- 20:32SPEAKER_00: because data centers generally tend to raise the land value
- 20:35SPEAKER_00: in very rural areas specifically.
- 20:38SPEAKER_00: And so all of these together just make me think that like one American,
- 20:41SPEAKER_00: American farmland is just really,
- 20:43SPEAKER_00: really not threatened by like warehouse size computers like
- 20:46SPEAKER_00: if you think like my genuine ask for people who believe this is to just look out
- 20:51SPEAKER_00: the airplane window the next time.
- 20:52SPEAKER_00: Yeah, there are a lot of other stats I could whip out about this,
- 20:54SPEAKER_00: but I think a lot of people just have...
- 21:07SPEAKER_00: I'm very bad at thinking on the margin about the stuff where like,
- 21:10SPEAKER_00: they'll see a single data center bought some farmland and they'll be like,
- 21:14SPEAKER_00: I should oppose this because I don't want all farmland to be replaced
- 21:17SPEAKER_00: with data centers.
- 21:18SPEAKER_00: And like,
- 21:18SPEAKER_00: no one's suggesting that and like the only way you could believe
- 21:21SPEAKER_00: that American That'd be a fantastic viral tweet.
- 21:24SPEAKER_00: Electricity.
- 21:24SPEAKER_00: They use a lot of electricity, going to use a lot more.
- 21:26SPEAKER_00: Is the case where this is a very, is it like a geographic issue where
- 21:53SPEAKER_01: It's more of a problem in some places, but it's going to eventually be an issue.
- 21:57SPEAKER_01: I'm sort of trying to half answer this question for you.
- 21:59SPEAKER_01: So to what extent is this a problem right now and today?
- 22:04SPEAKER_01: And how about like, again, five to seven years down the road?
- 22:07SPEAKER_00: Yeah.
- 22:08SPEAKER_00: For electricity prices,
- 22:10SPEAKER_00: data centers do use gigantic amounts of electricity compared to most other things
- 22:13SPEAKER_00: that we do.
- 22:14SPEAKER_00: Kevin O'Leary's data center in Utah might have something like 2% of the power draw
- 22:20SPEAKER_00: of the entire country if it's fully built out.
- 22:22SPEAKER_00: That's crazy.
- 22:23SPEAKER_00: Orders of magnitude more significant than their water draw.
- 22:27SPEAKER_00: At the same time,
- 22:27SPEAKER_00: electricity markets are just very complicated and don't always work in
- 22:30SPEAKER_00: a very simple,
- 22:31SPEAKER_00: oh, you increase the demand and the price just immediately goes up for everyone.
- 22:35SPEAKER_00: That's not really how it works.
- 22:38SPEAKER_00: We've seen so far is one, there are a few utilities that have reported.
- 22:43SPEAKER_00: We have had to slightly raise household prices for electricity because of the data center.
- 22:48SPEAKER_00: They've reported the data centers as being one of several causes,
- 22:51SPEAKER_00: but that's usually not a significant percentage.
- 22:53SPEAKER_00: It's like a meaningful amount to the point that people can get annoyed.
- 22:56SPEAKER_00: Then it kind of becomes a question of tax revenue and how much you're willing
- 22:59SPEAKER_00: to trade off like additional utility costs for way more tax revenue.
- 23:03SPEAKER_00: But I think the general pattern is way More confused than most people seem to understand.
- 23:08SPEAKER_00: Like in general,
- 23:09SPEAKER_00: the Lawrence Berkeley National Laboratory had done the study a few months ago
- 23:12SPEAKER_00: where they observed a general pattern where The places where
- 23:15SPEAKER_00: the most new data center capacity exists are also the places
- 23:19SPEAKER_00: where electricity prices have risen the least.
- 23:21SPEAKER_00: And there's this weird kind of inverse relationship there.
- 23:24SPEAKER_00: And there are a lot of different ways to understand that,
- 23:26SPEAKER_00: you know,
- 23:26SPEAKER_00: like a very constant new load and demand can sometimes be really good for prices for
- 23:32SPEAKER_00: the same reason.
- 23:33SPEAKER_00: You know,
- 23:33SPEAKER_00: if a lot of people start going to a restaurant consistently,
- 23:36SPEAKER_00: the restaurant can expand and benefit from economies of scale
- 23:39SPEAKER_00: and maybe lower prices, et cetera.
- 23:40SPEAKER_00: And there's a lot of intense confusion about this.
- 23:42SPEAKER_00: I think the,
- 23:42SPEAKER_00: maybe the single most confused stat that I've seen about this is there was
- 23:46SPEAKER_00: a Bloomberg piece last September that claimed that there is at least one place in
- 23:51SPEAKER_00: the country where a specific node in the electric grid,
- 23:55SPEAKER_00: its wholesale prices went up by 267%,
- 23:58SPEAKER_00: and it happens to be pretty close to a data center.
- 24:01SPEAKER_00: And a lot of people misinterpreted that to mean that the general household price
- 24:05SPEAKER_00: of electricity had gone up by 267%.
- 24:07SPEAKER_00: Elizabeth Warren had actually put out a video where she said,
- 24:11SPEAKER_00: if you live near a data center, your prices have gone up by up to 267%.
- 24:15SPEAKER_00: And that's just completely confused.
- 24:17SPEAKER_00: Wholesale prices at very specific nodes just tell you almost nothing
- 24:20SPEAKER_00: about what residents are actually paying.
- 24:23SPEAKER_00: It's kind of really complex and in the weeds.
- 24:25SPEAKER_00: But that's an example of this general game of telephone that's happening
- 24:28SPEAKER_00: where All these facts are coming out about data centers.
- 24:30SPEAKER_00: People kind of mangle and just slightly misinterpret them.
- 24:33SPEAKER_00: And suddenly you get from like, oh,
- 24:35SPEAKER_00: there's this one node where wholesale prices went up a lot
- 24:38SPEAKER_00: to residents are now paying 267%.
- 24:41SPEAKER_00: And like the whole debate is kind of just filled with these like weird results
- 24:44SPEAKER_00: of this game of telephone that's happening with articles about it.
- 24:48SPEAKER_01: If we're going to keep building more,
- 24:50SPEAKER_01: and I'm assuming we're going to keep building more,
- 24:52SPEAKER_01: they're going to draw a lot of power.
- 24:53SPEAKER_01: Again, might there be a more noticeable aggregate impact in the future?
- 24:59SPEAKER_01: And someone's going to have to pay for that.
- 25:03SPEAKER_00: Yeah.
- 25:04SPEAKER_00: This one,
- 25:05SPEAKER_00: good data center policy can force them to internalize the costs of the grid upgrades
- 25:11SPEAKER_00: that they require.
- 25:12SPEAKER_00: And this can sometimes be positive some,
- 25:13SPEAKER_00: because the phrase I hear the most in this debate is America's aging energy
- 25:18SPEAKER_00: And like,
- 25:19SPEAKER_00: it is kind of a miracle that the main new industry
- 25:22SPEAKER_00: in America is already fully electrified.
- 25:24SPEAKER_00: You know, like I've always been pretty interested in climate change.
- 25:27SPEAKER_00: If you went back to college me and told him like,
- 25:29SPEAKER_00: hey, the main new industry in America in 2026 is already fully electric.
- 25:33SPEAKER_00: We don't have to do this complicated engineering problem
- 25:36SPEAKER_00: of converting primary fossil fuel use into electricity.
- 25:39SPEAKER_00: I would have been like, that's pretty cool.
- 25:40SPEAKER_00: That's neat.
- 25:41SPEAKER_00: And also, A lot of data centers are just investing a lot in the grid overall.
- 25:45SPEAKER_00: The top four hyperscalers make up something like 50% of all money spent
- 25:50SPEAKER_00: on private power purchasing agreements for renewable energy and stuff.
- 25:54SPEAKER_00: But even though they're investing all this
- 25:56SPEAKER_00: and there's all these potentially positive sum trades,
- 25:59SPEAKER_00: they're still just consuming so much electricity that I could see bigger
- 26:02SPEAKER_00: and bigger problems coming down the line for American consumers where a lot
- 26:07SPEAKER_00: of costs could still be traded off and there could just be a lot of other problems
- 26:10SPEAKER_00: with generating electricity.
- 26:11SPEAKER_00: So I don't want to make super confident claims about the future,
- 26:14SPEAKER_00: but at least right now,
- 26:16SPEAKER_00: I think the situation is very,
- 26:17SPEAKER_00: very different from what everyday people understand,
- 26:19SPEAKER_00: where they kind of just assume the more data centers, the more your bill goes up.
- 26:24SPEAKER_01: So we need to have a more modern electrical system,
- 26:27SPEAKER_01: hopefully create more sources of clean energy.
- 26:31SPEAKER_01: We need a government that makes it easier to build all that stuff on budget, on time.
- 26:40SPEAKER_01: All that stuff needs to happen going forward.
- 26:43SPEAKER_01: Then I think about this moratorium in New York, which seems to be Weird.
- 26:50SPEAKER_01: Because they have a nuclear power plant there they don't want to use.
- 26:55SPEAKER_01: If you're worried about electricity prices in that particular area,
- 26:58SPEAKER_01: there are policy tools making them internalize that cost.
- 27:03SPEAKER_01: It's hard for me to look at that moratorium and view that as a data-driven,
- 27:08SPEAKER_01: policy-driven response to a problem.
- 27:11SPEAKER_01: And you're not a political analyst, but I view that as a purely sort of political...
- 27:18SPEAKER_01: Yeah, 100%.
- 27:18SPEAKER_01: It's very hard for me not to read that as basically like throwing a bone to
- 27:21SPEAKER_01: the majority of voters at this point who are quite angry about data centers.
- 27:25SPEAKER_01: Like data centers poll so terribly compared to most things.
- 27:27SPEAKER_01: It's a really unifying issue despite, you know, I
- 27:31SPEAKER_00: feel like I'm going...
- 27:46SPEAKER_00: I'm taking crazy pills a little bit by being so far out of the norm here.
- 27:52SPEAKER_00: The stated reason is that this is a temporary pause while we try
- 27:55SPEAKER_00: to figure out good data center policy.
- 27:57SPEAKER_00: And I have to say, we just know how to interact with these things I think that...
- 28:26SPEAKER_00: The cost is very real.
- 28:27SPEAKER_00: I come from a post-industrial forgotten area
- 28:32SPEAKER_00: in central Massachusetts that's just not doing super well.
- 28:35SPEAKER_00: I know that my community and communities around it could actually use a source
- 28:40SPEAKER_00: of new tax And I think it's depressing that a lot of places this are going
- 28:45SPEAKER_00: to miss out on that because I think of kind of confused overreaction to
- 28:49SPEAKER_00: the general build out.
- 28:50SPEAKER_01: But do you think that, I mean,
- 28:52SPEAKER_01: you think like you have the facts on your side,
- 28:54SPEAKER_01: you have the facts about water and land and electricity and those facts,
- 29:01SPEAKER_01: as well as like, again, the revenue generated.
- 29:04SPEAKER_01: Do you think that is sufficient?
- 29:07SPEAKER_01: Listen, people still may have a problem with AI because they're worried about
- 29:10SPEAKER_01: the robots killing them or taking all their jobs, not necessarily in that order.
- 29:13SPEAKER_01: But do you think that even though people may still have AI concerns that
- 29:16SPEAKER_01: the data center sort of vector for those concerns will turn out a year from now,
- 29:23SPEAKER_01: two years from now to have been a very kind of ephemeral issue?
- 29:29SPEAKER_00: Yeah, I want to be clear that I share those broader concerns and I'm very sympathetic.
- 29:32SPEAKER_00: I think there's two things to say about this.
- 29:34SPEAKER_00: One,
- 29:34SPEAKER_00: I don't think that the data center backlash is actually a very effective response
- 29:39SPEAKER_00: to AI's effects on society.
- 29:41SPEAKER_00: There's just so much capital behind these that it's very That's not going to be
- 29:44SPEAKER_00: a big win for the environment for me.
- 29:47SPEAKER_00: But yeah, so basically, like,
- 29:59SPEAKER_00: I don't think that local data center nimbyism is actually going to push
- 30:02SPEAKER_00: the needle very much on the progress of AI.
- 30:04SPEAKER_00: If that's your concern, pushing for good state and federal policy just makes a lot more sense.
- 30:09SPEAKER_00: And also like,
- 30:10SPEAKER_00: you really don't want to push the data centers out of America because
- 30:13SPEAKER_00: that makes them harder to I'm actually really unsure about how much
- 30:19SPEAKER_00: of this is actually based on people's reaction to AI specifically.
- 30:23SPEAKER_00: My impression is that there are actually a significant number
- 30:25SPEAKER_00: of anti-data center activists who are actually using AI tools quite a bit
- 30:29SPEAKER_00: in their activism.
- 30:30SPEAKER_00: And when I talk to people about this, AI rarely comes up.
- 30:34SPEAKER_00: I just find that the average person isn't really thinking about
- 30:37SPEAKER_00: the long-term societal effects of AI.
- 30:39SPEAKER_00: Nearly as much as I would expect.
- 30:41SPEAKER_00: And basically every single smart person in their lives have told them, oh,
- 30:45SPEAKER_00: there are these monstrous new buildings that plop down,
- 30:47SPEAKER_00: don't create any jobs, just radiate harm and sap your community dry.
- 30:52SPEAKER_00: Do you want this in your community?
- 30:54SPEAKER_00: I think regardless of how people feel about AI,
- 30:55SPEAKER_00: they mostly just tend to say, no, that sounds very bad.
- 30:58SPEAKER_00: Why is this happening?
- 30:59SPEAKER_00: Why is this like strange anti-democratic thing rolling over society?
- 31:03SPEAKER_00: And I think they're missing out on the complexities of what data centers are actually like.
- 31:07SPEAKER_00: I don't know.
- 31:07SPEAKER_00: I personally just listen to a lot of town meetings and stuff,
- 31:09SPEAKER_00: and I just don't hear the actual effects of AI on society coming up nearly as much
- 31:13SPEAKER_00: as I expected.
- 31:13SPEAKER_01: I tend to be very skeptical of activists about many issues that they sort of have
- 31:18SPEAKER_01: a belief,
- 31:20SPEAKER_01: and if something comes up that they can kind of throw at the wall and see
- 31:23SPEAKER_01: if it'll stick,
- 31:24SPEAKER_01: and if you kind of don't like AI,
- 31:27SPEAKER_01: or you're a certain kind of environmentalist,
- 31:31SPEAKER_01: or maybe you're not in love with capitalism,
- 31:34SPEAKER_01: AI, the data center seems to be like a great vehicle for the moment.
- 31:38SPEAKER_01: Now, the average person might care about the water and legitimately care about the water.
- 31:43SPEAKER_01: But I wonder again, if more your activist class that, you know, that could be an issue.
- 31:48SPEAKER_00: Yeah, it's hard to say.
- 31:49SPEAKER_00: I know there was an article in the New York Times a few weeks ago about how,
- 31:53SPEAKER_00: oh, data centers are this amazing opportunity for the Democrats.
- 31:57SPEAKER_00: This could be our issue in 2028 specifically.
- 32:00SPEAKER_00: And if you look at the article,
- 32:01SPEAKER_00: I think there's exactly one sentence that actually talks about the harms
- 32:04SPEAKER_00: of data centers,
- 32:05SPEAKER_00: where it just kind of hand waves like,
- 32:07SPEAKER_00: oh, they use a lot of water and the electricity prices.
- 32:09SPEAKER_00: But then it circles back to like, oh,
- 32:11SPEAKER_00: this is this really magical Oh, I can find this deep meaning in this local struggle.
- 32:27SPEAKER_00: I want to finish with is...
- 32:52SPEAKER_01: I think about this issue and I think about the issue of nuclear power plants in
- 32:57SPEAKER_01: the 70s where they got all this bad press and you had people
- 33:01SPEAKER_01: who didn't like their impact on the environment and they didn't like that they were
- 33:05SPEAKER_01: a sign of out of control capitalism and all this stuff.
- 33:08SPEAKER_01: And if you had a different view, I don't think, how were you going to express that view?
- 33:16SPEAKER_01: And to me, like,
- 33:17SPEAKER_01: for all the criticism of, like,
- 33:18SPEAKER_01: social media,
- 33:20SPEAKER_01: this to me is a great example how, like, I think you've had an impact on this issue.
- 33:26SPEAKER_01: And I think what you said has mattered.
- 33:29SPEAKER_01: And, like,
- 33:30SPEAKER_01: pre-social media, I mean,
- 33:31SPEAKER_01: I don't know, maybe you would have written one article in The Atlantic or something.
- 33:35SPEAKER_01: But, like,
- 33:36SPEAKER_01: your ability,
- 33:37SPEAKER_01: but the ability for you to keep, like,
- 33:38SPEAKER_01: get this message out there and talk about it and engage and say,
- 33:41SPEAKER_01: like, no, no, you know, people say, like, you're wrong.
- 33:44SPEAKER_01: No, here's my fact.
- 33:46SPEAKER_01: To me, if you believe overall in the power of reason to create a better future,
- 33:54SPEAKER_01: to me, what you're doing here is very encouraging.
- 33:58SPEAKER_00: Yeah, I appreciate that.
- 33:59SPEAKER_00: I mean, I'm massively grateful to one,
- 34:01SPEAKER_00: just be living at a time when like blogs are kind of back,
- 34:03SPEAKER_00: you know,
- 34:04SPEAKER_00: I feel like I kind of missed the golden age of blogging,
- 34:07SPEAKER_00: but like we're in the silver age now or whatever this is, or maybe it's better.
- 34:10SPEAKER_00: But yeah,
- 34:11SPEAKER_00: I've been consistently really pleasantly surprised because like
- 34:14SPEAKER_00: when I first started this, I was going completely crazy.
- 34:16SPEAKER_00: Honestly, it was just very eerie to be standing across from adults telling me.
- 34:20SPEAKER_00: You can't use AI.
- 34:21SPEAKER_00: It uses 10 times as much energy as a Google search.
- 34:23SPEAKER_00: And just standing there being like, we're worried about 10 Google searches.
- 34:27SPEAKER_00: Now, that's the amount of energy that we're all freaking out about.
- 34:29SPEAKER_00: And so I actually didn't really expect my first blog on this to take off very much,
- 34:33SPEAKER_00: but it almost immediately kind of exploded.
- 34:35SPEAKER_00: And there was just a lot of market for talking about this stuff much more directly.
- 34:39SPEAKER_00: Yes.
- 34:40SPEAKER_00: Being able to just sit down and provide these general overviews of like, Hey,
- 34:43SPEAKER_00: you know,
- 34:44SPEAKER_00: I don't have any like unique expertise,
- 34:46SPEAKER_00: but I can have access to the same facts that everybody else has.
- 34:49SPEAKER_00: And I think we're just misinterpreting these relatively simple, accessible facts.
- 34:53SPEAKER_00: I've been pleasantly surprised.
- 34:54SPEAKER_00: I'm surprised at just how much this has actually moved the needle on stuff.
- 34:57SPEAKER_00: So that's been really great.
- 34:58SPEAKER_00: I do find that in general conversations,
- 35:00SPEAKER_00: if I talk to the average person about this,
- 35:02SPEAKER_00: their beliefs about this aren't super strong.
- 35:04SPEAKER_00: They'll be like, oh, I heard every prompt uses a whole bottle of water.
- 35:07SPEAKER_00: And then I can say like, oh, nobody actually really believes that anymore.
- 35:10SPEAKER_00: That was this one-off thing.
- 35:11SPEAKER_00: That turned out to have been mistaken.
- 35:13SPEAKER_00: And they'll be like, oh, okay, now I feel okay about using AI.
- 35:15SPEAKER_00: And like, it's actually pretty straightforward.
- 35:17SPEAKER_00: A lot of people understandably read a lot of like hostility and anger into the debate.
- 35:22SPEAKER_00: But behind the scenes,
- 35:23SPEAKER_00: the average person is just kind of responding to what they're seeing on the ground.
- 35:25SPEAKER_00: And if you can just change that with just like a decent argument
- 35:28SPEAKER_00: and being like generally pleasant, like I think that's pretty powerful.
- 35:31SPEAKER_00: So yeah, I'm living the dream.
- 35:32SPEAKER_01: It's been great.
- 35:33SPEAKER_01: Being generally pleasant is an underappreciated attribute.
- 35:36SPEAKER_01: Andy, thanks so much.
- 35:37SPEAKER_01: This was fantastic.
- 35:38SPEAKER_01: Thanks for coming on the podcast.
- 35:39SPEAKER_00: Yeah, thanks so much, James.
- 35:40SPEAKER_00: This was great.
- 35:40SPEAKER_01: Thank you.