Are We Really Blowing Up the Planet for AI Slop and Porn?
Substack, Apple, Spotify, YouTube, Amazon
In Episode 82 of Rabbit Hole of Research, Joe, Nick, Georgia, and Mary welcome back three-time returning guest (previously on Episodes 13 and 39) and “Friend of the Pod”, David Pincus, PhD, cell biologist at the University of Chicago, to the Basement Studio to talk data centers.
We dig into the water costs, why your electric bill is going up, whether space-based cooling is science or Handwavium, what cellular stress can teach us about AI resource allocation, should we use human values or cell behavior to guide AI training, and is there anything good that will come out of these Data Centers. And it wouldn’t be an episode with David if we didn’t go down all the unexpected rabbit holes, we talk lobster farms, “data furnace”, olympic swimming pools, and of course, Extremophiles!
Previous Episodes with David:
Episode 13: Blockbusters and Frozen Landscapes: Climate Change Through the Lens of Pop Culture
Listen nowEP 39: Fantastic 4 series: Johnny Storm and Spontaneous Combustion
Georgia and Joe will be hanging out at DragonCon, September 3-7, 2026, if you are in Atlanta come on out and say hello (I’m sure we will be hanging out in the Westin Lobby)! And Joe will be on seven panels.
12 months of all-original art by Jim and Georgia; orders are open now, with mailing expected in September
Check out what the RHR crew is creating:
Joe:
Future Events to Hang with the Crew:
Podcast Cross-Appearances
Events & Conventions:
Joe:
- Dragon Con — Atlanta, GA (September 3-7, 2026) – Joe attending as a Professional on seven panels and Georgia will be around also!!
- Maud’s READINGS at Printer’s Row Wine on Sep 19th 2-4pm
- Art show Ultrastructure: Science is Fun, with Carol Estes at Promise You Art House in Highland IN, October 10th, 10-3pm
- Family Day: Seeing the Invisible World to Build a New One; in conjunction with the fall exhibition: teamLab: Everything Exists in Infinite Continuity, at the Smart Museum of Art In Hyde Park, IL, Nov 14th, 11-3pm
- Art and Science lecture about Zombies: Dec 10thGeorgia:The inaugural 4th Street Print Fest at Chesterton Art Center on Saturday, October 3, from 1–4 PM
It’s Science for Weirdos
Want to support the show? Tell your friends. Follow us on social media, Discord, share the podcast, and let us know what topics you are excited about. Leave a Comment. And for email alerts sign-up for the Substack newsletter and never miss an episode, exciting updates or the bonus images we talk about on the episodes.
We discussed these questions on the episode and now want to Hear From You (leave a comment):
- Would you want a data center in your neighborhood? What would they have to offer to make it worth it?
- Should we train AI with human values or cellular values?
- What, if anything, would you like to see AI used for? Are you optimistic that we will get there?
Drop your thoughts in the comments. We read them all, and your ideas often shape future episodes.
Leave a comment
Future Episodes
- Episode 84: ETGuest: Chris Guzman (@theboxinggloveartist)
- Episode 86: Disability in HorrorGuest: Grace Daly (Bram Stoker nominee)
- Episode 88: CraziesRecorded at Reed’s Local with Melissa and Luc
- Episode 90: Witch PleaseGuest: Bill Halliar (@skeletongue)
Share
Leave a comment
For more stuff (Images, Episode Highlights, events, etc), subscribe to our Substack newsletter!
Show Notes & Fun facts
Stuff crew said they’d put in shownotes
Books:
- The Machine Stops — E.M. Forster (1909) short story
- The Last Question — Isaac Asimov (1956) short story
- The Hitchhiker’s Guide to the Galaxy — Douglas Adams
- So Long and Thanks for All the Fish — Douglas Adams
- The Emperor of All Maladies — Siddhartha Mukherjee
Fun Facts to Impress Your Friends With:
- The human brain runs on about 20 watts — roughly the equivalent of a dim light bulb, or as David put it, a sandwich. A modern AI data center GPU cluster runs on hundreds of megawatts. That’s the efficiency gap that needs to be closed.
- The ISS uses radiator panels the size of a tennis court to cool roughly 70-80 kilowatts of heat load. Dumping just 1 megawatt of waste heat in space requires approximately 1,200 square meters of perfect radiator panels, and a modern GPU cluster generates hundreds of megawatts.
- Ireland’s data centers consume about 21% of the national electricity supply— and data center energy demand globally has grown ~12% per year since 2017, and projected to equal Japan’s entire electricity consumption by 2030.
- Evaporative cooling loses up to 85% of its water to the atmosphere — it doesn’t go back into the water table. The Tempe Town Lake in Arizona evaporates between 390 million and 650 million gallons of water a year that never return to the Colorado River.
- The Jevons Paradox — coined by 1860s steam engine engineer William Stanley Jevons found that making steam engines more efficient actually increased total energy use, and the same pattern may apply to AI efficiency gains today.
Episode Highlights
- 00:00 Basement Roll Call — Full house in the basement studio, almost, David has a legit excuse, “car ownership is a debated thing in families.”01:57 Meet David Pincus — Three-time returning guest and Friend of the Pod, David Pincus, PhD, UChicago cell biologist and expert in cellular adaptation04:19 What The Cloud Is — Joe delivers the monologue: the cloud isn’t wispy and floating, it’s “a giant space heater that computes.”08:28 Energy Demand Surge — Data centers have grown 12% per year since 2017, Ireland uses 21% of national electricity on data centers, and one proposed AI complex would use as much power as the city of Houston.14:52 AI Hype And Jobs — David calls out the circular investment logic: “If they’re really as good as they say they are, they should prove it by not ruining the Earth.”19:11 Curtailment And Duck Curve — Joe explains how AI could modulate its own power use to absorb wasted renewable energy during off-peak hours, turning a climate problem into a climate solution.22:02 Jevons Paradox Explained — David drops the Jevons Paradox: the 1860s discovery that making steam engines more efficient actually increased total energy use — and AI may be no different.23:41 Space Data Centers Reality — Space is cold, but that doesn’t mean cooling is free; in a vacuum, heat can only leave as radiation, requiring radiator panels the size of a small town.25:01 Does Deleting Data Help — Georgia wonders if mass email deletion would make a dent; David suggests she “Marie Kondo her inbox.”30:41 Water Use Headlines — David breaks down the scary headline number: about 80% of reported data center water use is actually from generating the electricity, not the cooling itself.32:41 Cooling Methods Breakdown — Joe walks through the three approaches: evaporative cooling (water-heavy), air conditioning (energy-heavy), and liquid immersion cooling (the newer, more efficient option).38:57 Local Politics And Promises — Mary reports from the front lines on CoreWeave’s Hammond, Indiana proposal: “They’re promising birthday parties, ice cream, and pony rides.”43:20 Promises Meet Reality — The crew debates whether data center companies will ever actually deliver on community promises, or just take the tax breaks and run.43:58 Turning Heat Into Benefits — Joe’s fun list: Microsoft’s 2011 “data furnace” concept, Deep Green’s server racks under swimming pools, the Norwegian lobster farm, and the Paris Olympic Aquatic Centre heated by Equinix’s data center.46:25 Cooling Hacks Underground — From Cold War bunkers in Stockholm to SubTropolis in Kansas City to a functioning data center at the South Pole, weird locations already solve the cooling problem.47:37 Taxes And Lax Rules — Data centers go where regulations are lax; Mary asks why they can’t just be taxed like everyone else.50:09 Military AI And Nuclear Power — David raises the military’s 50-year data center leases; Joe brings up Three Mile Island — now rebranded the “Crane Clean Energy Center” — being restarted to feed AI.53:46 Petty AI Complaints — New segment: Mary does not need AI to summarize her coloring sheets, and Georgia’s voice-to-text still can’t get it right.55:34 Sci-Fi Warnings And Jokes — Joe highlights The Machine Stops (1909) and The Last Question (1956); David adds The Hitchhiker’s Guide to the Galaxy — the answer to everything is 42.57:23 Cells, Budgets, And AI Hunger — David draws the key parallel: stressed cells always slow down and budget their resources — AI systems have no such constraint, and the cost is being passed onto us and the planet.01:00:48 Raising AI With Values — David shares Siddhartha Mukherjee’s NYT essay: training AI is like raising a child — you can only hope you sent it into the world with good values.01:07:46 Regulation And Real Benefits — The crew revisits the Industrial Revolution: you can have regulation and growth at the same time, and we already know what happens when you don’t.01:10:33 Freemium Traps And Libraries — The freemium model is coming for AI just like it came for streaming; Mary and Georgia remind everyone that Kanopy and Hoopla are free with your library card.01:14:12 Wrap Up And Hopeful Outlook — David closes on a hopeful note: “If we can let the science lead, that’s going to help humanity”
“Stay curious, stay safe… Love Y’all!”
Join Rabbit Hole of Research on Discord: https://discord.gg/2nnmKgguFV
Subscribe and Share our Substack newsletter to get email updates, never miss an episode, and spread the word!! Don’t forget to give us 5 stars or a like!
Transcript
[00:00:00] joe: Hey, welcome back to the Rabbit Hole of Research down here in the basement studio. We are all crewed up
[00:00:10] Geo: Full house.
[00:00:11] joe: house … full house. Well, almost everyone’s in the basement
[00:00:14] Nick: someone was supposed to be in the- basement studio,
[00:00:16] joe: yeah, but the wife stole the car- … or
[00:00:18] David: car ownership is a debated thing in families
[00:00:23] joe: So yeah, you got me, Joe.
[00:00:24] Nick: got Nick.
[00:00:25] joe: got Nick.
Georgia. We got Georgia.
[00:00:27] Mary: You got Mary.
[00:00:28] joe: got Mary, and we have
[00:00:30] Nick: Three-time returning
[00:00:33] joe: returning. Choo,
[00:00:33] David: Returning guest, longtime listener, third time guest
[00:00:38] Geo: We’ve
[00:00:38] joe: got
[00:00:39] David: David
Pincus Yeah. It’s great to be back
[00:00:42] Geo: And we w- we wanted to know, Is this the first, third time guest?
[00:00:47] Nick: Is he the three-timer club? ‘
[00:00:49] David: always your first third time
[00:00:51] Mary: three- …the
[00:00:53] joe: club, but I don’t know if you’re the first
[00:00:55] Geo: I can’t-
remember
if there’s been anyone else.
[00:00:58] joe: It’s like Bill Hallier’s [00:01:00] been
on
[00:01:00] Geo: I think he’s only been on twice.
[00:01:03] joe: Oh, I think he’s only been on
[00:01:04] Geo: Anyway, we’ll
[00:01:04] David: I’m
gonna, I’m gonna claim it whether or not it’s
[00:01:07] Geo: We’ll, we’re gonna do-
We’re gonna do some research.
[00:01:10] joe: gonna
do some research. Go down a rabbit hole of research. We’ll do
[00:01:13] David: you say this makes me a friend of the pod?
[00:01:15] Geo: wondering-
[00:01:16] joe: that’s right. That’s right. And, and for those wondering- Sure … that wanna go
back, you were on episode 13, so our first season. So you were here before we were here. Like, you’re
our-
[00:01:27] Nick: was Georgia on that episode?
[00:01:28] Geo: No.
[00:01:29] joe: No.
[00:01:29] Nick: Oh. that was Even before Georgia. jumped back on.
[00:01:32] joe: a day one-er,
[00:01:32] Nick: a day one.
[00:01:34] joe: down
with, the crew.
[00:01:35] Geo: that was a very enjoyable. episode, I remember. And
[00:01:38] joe: then
we had episode 39, Johnny Storm, so that was from the Fantastic Four series.
[00:01:43] Nick: that Last year … that was last Last season
[00:01:45] joe: two, so yeah.
[00:01:46] Nick: Damn, all three seasons. yeah,
[00:01:48] joe: been on, you’ve episode every season,
so yeah.
[00:01:52] David: alive
[00:01:53] joe: yeah. I mean, that’s that’s it, so well, cool. Well, we’re here. Do you wanna give your credentials [00:02:00] just in case someone hasn’t listened to one of the other two
[00:02:02] David: Yeah, of course. Sure, yeah. So I’m a scientist. I’m a researcher at the University of Chicago. I probably shouldn’t say where I’m from, but I spilled the beans
again.
[00:02:11] joe: do that every episode you do that.
So, it’s all right. You can, I think they can Google your name and it’ll
[00:02:16] David: yeah, It’s, true. It’s, It’s, not much of a secret. But I’m a colleague of Joe’s in the and we’re both biologists, cell biologists. I’m an expert in cellular adaptation, how cells cope with stress. Stress to a cell is any change in the environment change in temperature, pH, nutrients. You name it, if it changes, the cell has to adjust to it and roll up their sleeves and adapt. So we study the… You know, cells don’t have brains, but they still have feedback networks and metabolic pathways that ensure that they’re doing the best with what they have at any given moment, and we try to figure out the rules of the game for that.
[00:02:56] joe: Yeah. Well, cool. I think it’s gonna be a fun one. You were [00:03:00] on that first season talking about climate change
[00:03:03] Geo: and And didn’t David coin one of our favorite… Not
coin, but-
he-
[00:03:09] joe: Go ahead. Just please
[00:03:10] Geo: What are you- Extremophiles
[00:03:12] Nick: Oh.
Jesus.
[00:03:15] David: Extreme.
[00:03:17] joe: Yes.
I- I don’t know
[00:03:19] Geo: if I’m attributing- that
[00:03:20] joe: was on a, yeah, that was on a, Johnny Storm episode, I believe it was, extremophiles. But in particular we were talking about this episode and we were talking about you returning this season, and data centers came up. So this episode is about climate change and data centers.
So we’ll get into that. But Mary-
[00:03:37] Mary: Well,
I wanna go a little bit in reverse. So you gave your credentials,
but I don’t think
you gave your name. Oh.
[00:03:43] David: Oh.
David Pincus.
[00:03:44] joe: Yeah.
[00:03:44] Mary: there you go. Okay. All
[00:03:45] joe: right. I think he did say his name.
[00:03:47] Mary: Really?
[00:03:47] Nick: I don’t remember. He said that from the top.
[00:03:48] Mary: Yeah,
No, Okay.
Mary
wasn’t
[00:03:50] joe: Mary wasn’t listening.
[00:03:51] Geo: Oh. That’s
[00:03:51] David: Well, just in case you wanna Google me twice.
[00:03:54] joe: Oh. That’s
right.
[00:03:55] Mary: That’s
[00:03:55] joe: They really make sure
[00:03:56] Mary: I
thought… No, I
[00:03:58] joe: of Chicago.
[00:03:58] Mary: j-
[00:03:59] David: Oh,
[00:03:59] Mary: no, [00:04:00] just wanna make sure that, you know, we know who we’re talking
to. So, yeah.
[00:04:03] David: No name handwavium
[00:04:05] Mary: Yes. Yeah. Okay.
[00:04:07] joe: yes. We gotta have that.
Okay.
Well, cool. Well, I do have a little-
[00:04:11] Nick: was gonna say, w- what’s the list look like?
[00:04:14] joe: Well, I have lists,
but I got my little monologue.
Your
[00:04:15] Nick: Your monologue. I got my monologue
[00:04:16] joe: I got my little monologue. I’m coming back. See if I can do it
when
you hear the cloud, what do you picture? A blue sky with white wispy balls of cotton floating by and morphing into shapes of fantastical animals and familiar faces?
Or
do you imagine data zipping around up there somewhere between your phone and the internet, moving around you like digital air?
Here’s
what the cloud actually is. It’s a warehouse, a very large, loud, and hot warehouse, usually in rural areas by waterways or coming to a neighborhood near you, stuffed floor to ceiling with server racks, calculating away 24/7, consuming natural resources to make the latest viral memes, write emails, [00:05:00] or create a self-aware AI.
And fiction has been warning us about this AI for 70 years, but the real villain isn’t going to say, “I’ll be back,” launch nuclear weapons, or use us as a battery. But it’ll just let us boil ourselves in a pot like frogs, killing us slowly with thermodynamics. You know what? Let me look that up because I think there’s a meme for
[00:05:21] Geo: that
Can you Google it?
[00:05:24] Nick: Don’t
worry, I’ll ask ChatGPT
Yeah.
[00:05:28] joe: It’ll make it for us,
[00:05:29] Geo: Wow.
[00:05:30] joe: Yeah, yep. So that’s it. I mean, but yeah
[00:05:34] David: Cloud’s such a fluffy word, you know?
[00:05:36] joe: It is.
It
sounds-
[00:05:37] Geo: so
[00:05:38] David: a euphemism? I didn’t think of it at the time. You know, it seemed like a nice, just a real metaphor for, like,
it’s up there
But now it, with the sort of dark side of all of this stuff, cloud’s sort of taken on a Orwellian kind of tone to it to me
[00:05:56] joe: Right.
I mean, yeah, it the word cloud [00:06:00]
is just not,
you- it’s not here in front of us, right? It’s something that you cannot obtain or see, and it’s all about scaling, right? So you’re scaling something up, and so how do you scale is just having more. And usually when you have more, you want to put it somewhere that’s not in your face, right?
It’s that’s the idea of it. And it’s really just server racks. Just a place, a physical location
[00:06:23] Nick: Mm-hmm That you don’t own
[00:06:24] joe: you don’t own … yeah, that you don’t own, that you don’t see. well,
[00:06:27] David: of sight, out of mind.
[00:06:28] joe: might see it if, depending where you live. But
[00:06:31] David: Increasingly
[00:06:32] joe: so. Yeah. Just a giant space heater that computes.
[00:06:35] Geo: I think of it like, like when we were talking about electricity and you just Turn the light switch on- you don’t think about
well, what,
you know, what-
went on to turn the light
on It’s the
same kind of- Yeah, and
[00:06:47] joe: what powers that? Is it coal? Is it you know, nuclear? Is it solar? Is it wind? Is it hydro? You turn the lights on and you expect,, you expect light.
Like, you don’t… You flick the switch or you plug your plug in, you, you expect it to work, and you [00:07:00] don’t think about the back end. Same thing with, electric cars. You still have to think about if you plug in and it’s your electricity’s from a coal-fired plant, is it– are you probably coming out a little bit better?
Yeah, but you’ve kinda just displaced your thinking about that,
so. You have to change them. Right. That’s right.
[00:07:15] Nick: Like, it’s not a full step ahead. you’re just moving a different lane
[00:07:19] joe: yeah. You’re
[00:07:20] David: I do think that pretty much everybody thinks electricity is good though,
and Has improved the quality of lives, right? And so- the sheer fact of having to gen- you know, use fossil fuels, have environmental impact, of course, that’s not ideal, but at least with the trade-off of generating electricity and having the lights on at night and you know, being able to cook food over, you know, places.
Anyway, there’s a whole bunch of benefits that are obvious for electricity, and I think everyone
[00:07:55] Geo: Correct.
Correct
[00:07:56] David: different
with data centers is every… To a [00:08:00] first approximation, every one of those servers is running porn, right? Like, or like
that
[00:08:06] joe: probably right. That’s
[00:08:07] Geo: Bored.
[00:08:08] David: like, and
like,
or people asking about their own like health or
mental health issues and that’s like sad, but in- interesting.
But anyway, like
it’s AI,
[00:08:18] joe: I
th- yeah, and
[00:08:19] David: and I just don’t think people can agree universally that it’s a good
improvement over
everyone’s
[00:08:26] Nick: of life isn’t being lifted.
[00:08:28] joe: I wanna say also, I know AI is the one of the drivers now of the extra increase, but probably starting, 10 years ago or so, data centers started to ramp up just the devices that we use to stream everything. You guys listen to this podcast.
That is streaming out of a cloud from a data center somewhere. So
[00:08:50] Nick: Wait, are you saying that we’re part of the problem?
[00:08:52] joe: You’re part of the problem Yes, Nick, you
are- He’s the worst … you are
part of the problem,
[00:08:55] Geo: of the problem, Especially you, Nick. No.
[00:08:57] joe: You’re 100%.
[00:08:57] Mary: 100%.
[00:08:59] Nick: sin.
[00:08:59] joe: Yeah, but I [00:09:00] mean, it’s just kind of …data
centers started to ramp up and be built before even AI.
So now we have this onslaught. Yeah, right, and porn and movie
[00:09:09] Geo: streaming- Dancing cat videos
[00:09:11] joe: was part of it. But AI now
[00:09:13] Geo: But AI now has
[00:09:13] joe: accelerated that, where we were seeing some growth, you know, 5% a year maybe, and now we’re starting to escalate that to 10%, you know, growth of energy need consumption. And that’s what we’re now dealing
[00:09:28] Geo: And so because it really feels very recent.
that you hear about every
single town-
[00:09:35] joe: Yeah,
[00:09:36] Geo: is struggling with,
you know- whether they’re gonna put up a data center. It’s like, how did this happen just in the last-
[00:09:44] Nick: Well, what, we had recently a town meeting. Yeah. And it’s hitting closer to home for everyone.
And- Right.
[00:09:51] Geo: So I just think, wow, I mean, it really- is that because… And I don’t
wanna
like just depressing
and
about [00:10:00] politics, but is it because of…
What is it? It’s really because of the demand
with
[00:10:06] joe: wanna, like, just depressing and about politics, but is it because of… I mean, what, what is it? It’s really because of the demand with- On it, yeah. I mean, AI is, like I said, part of it, and we’ve seen an acceleration of growth. But one of the things is that we’re seeing since 2017 we’ve been having about a 12% growth per year for data center energy.
And I think it’s just hitting the news now because if you keep growing, right, no one’s put a cap on it. That’s the problem. It’s like, oh, we’re gonna double by in five years. , hopefully we can hold that line because it could double again in 10 years. And so that’s the issue. As the consumption of, media, of AI, of using that, and,, that you have to build new data centers out.
But there’s a lot of use. I mean, it was already building, so it’s not like last year or a couple years ago when ChatGPT really hit it. That’s when… And the memes and things like that, it was like, “Oh, we need new data centers.” Like I said, probably, , 2015, ’16, ’17, there was a growth already starting to happen, and people [00:11:00] were preparing for that and saying, “How do we get ahead of this?
How do we do something?” So there are people thinking about it, and it… But it’s hit the news now because now it’s coming to a town near you. That’s what’s
changing-
[00:11:11] Geo: town near you.
[00:11:12] joe: right. You’re just hearing it about it more. And, politics have, is very partisan, and so it’s a flashpoint and you can use it,
um-
[00:11:21] Nick: I don’t believe-
that the politicians understand.
it.
[00:11:24] joe: mm-hmm.
[00:11:25] Nick: Like, it’s new technology still to them.
[00:11:27] Geo: Yeah.
[00:11:28] Nick: And they still don’t understand half
The technology that is common in every household nowadays
[00:11:34] joe: And doing just a little bit of research on this, the, one of the extreme cases is like Ireland. Data centers use about 21% of the national, electrical
generation 21% is
[00:11:46] Mary: that’s huge 21%.
[00:11:47] David: Virginia, I know in the US, Virginia has a
huge fraction of its-
total power going to data centers, maybe one in,
one in four or something.
[00:11:56] joe: So
[00:11:56] Nick: yeah, and They also don’t tell you the full amount that they- [00:12:00] are actually gonna use. They say- a lowball number and then- exceed those numbers significantly. Yeah. Yeah,
[00:12:08] joe: I think, it’s gonna be, by 2030 we’ll be using, like, the, what the country of Japan uses, worldwide.
I mean, , we’ll be equaling, that, that level you know, globally as we start to scale up so,
[00:12:23] David: And
[00:12:23] joe: but yeah,
[00:12:23] David: in
terms of
[00:12:24] joe: is… Go ahead
[00:12:25] Geo: raw energy is
[00:12:28] David: in– I think it’s something like you know, this year there are these… The AI companies are all… There’s this– We can probably talk about this a little bit more, but there’s this circular investment
[00:12:41] joe: Mm-hmm.
[00:12:41] David: thing where the NVIDIA chip company that
makes the chips is investing in the AI companies that buy the chips to buy the chips and build the data centers. And then meanwhile, these IPOs they’re raising are gonna raise a bunch of money to build more of these data [00:13:00] centers, and it’s just, you know, it’s gonna be like half a trillion dollars potentially spent in, per year in the next few years just on, on building. And it’s just a Ponzi scheme. I don’t
know what the name is, but it just seems like economically false.
Like, like it’s
built on some kind of, I, I don’t under- it’s scary. But like, one of these super complexes that they’re talking about building would use as much power per year as the city of Houston you know? It’s
like enormous amounts of energy that they’re saying they’re gonna use in these things.
Anyway, sorry
[00:13:41] joe: and yeah, and the problem is that, you know, as you forecast this out, you know, it’s not a new problem. I mean, you can grow the power grid to compensate. The problem is that we’re lagging behind there, so these communities are suffering not because it’s, “Oh, wow, you know, we gotta add… They’re gonna consume this much [00:14:00] power.”
You know, there’s been other times in our history where you had to have c- power consumption, like you’re building ships and things like that, and you got your blast furnaces. That’s a consumption of power. It’s different. But w- we’ve ramped up, and now over the last decades, we’ve ramped down. We’re not doing heavy industrial here in the States, and so our power grid now is at a, some low baseline level.
And now the thing is ramp it up. Let’s go. And so you plug these data centers in now that have great demand with now public demand, and that’s where you’re starting to see the strain out on it.
[00:14:33] Geo: Places have failed to keep up their in- Right
infrastructure.
Infrastructure, yeah. Mm-hmm. yep. And so- Right … it’s-
already failing, and then you’re adding all this other-
[00:14:42] Mary: adding all this other.
[00:14:42] joe: Right.
Yep. And so that’s where the strain then comes in, and communities are like, “Whoa, hold on. We’re having brownouts and blackouts,” and
[00:14:48] David: And people pay more in their electricity bills.
[00:14:51] Geo: A lot more.
[00:14:52] joe: And I wanted to circle around, too, to the point you were talking about, David, about the AI and the hype and the super hype, because some [00:15:00] of it, that is what it is.
Like, there’s our trillion, billionaire, now one’s a trillionaire, hyping up this technology, and consumers are consuming it. But at some point in time, and we’re starting to see this with, ChatGPT, Claude, it’s like, oh, I’m running out of, prompts that I can put in.
They’re hitting limits of that. And so that’s because this costs money, and so at some point in time, they’re gonna go… These are like the tasters. Everyone’s getting a little taste and getting hooked. And then after you’re hooked, then they’re gonna go, “Oh, you know what? Oh, my bad.
That’s gonna cost you 20 bucks. Oh, my bad. If you wanna keep this prompting and impress your friends with your
memes and stuff,” you It’s gonna be… It- it’s kind of on the back end, they’re gonna have to pay for it. Right now, they’re not paying for it. And the investors,
at some point,
the stockholders are gonna go the share- the board is gonna go, “We need to make some money on this, so you know what?
Let’s let’s crank this up.”
[00:15:54] Nick: Aren’t they already
losing money?
[00:15:56] joe: they
already making money? Well, tons of money. Yeah.
It’s, you know, it’s not-
[00:15:59] Mary: With the [00:16:00] promise
that
later-
there’s,
The jewels
Will
rain down. they’re,
[00:16:04] Geo: they’re-
they’re- And then, yeah, and then, oh, this will bring so many jobs.
It’s not gonna bring so many jobs. It’s bringing, like, a couple people to check on the machines. I don’t understand. I don’t understand. that. Like, how
[00:16:16] Mary: like, how many jobs are there in a AI data center after the initial construction?
[00:16:23] Nick: I
wouldn’t say too
[00:16:24] Mary: Are, are- yeah.
I
mean-
[00:16:26] Nick: have a security
[00:16:27] Mary: I was
[00:16:27] joe: was
actually- That’s right.
[00:16:28] Mary: Like, I, I actually don’t know. I
d- I
[00:16:30] Nick: mean, Well- It’s
[00:16:31] Mary: It’s like,
[00:16:31] Geo: Homer
Simpson sitting there watching
[00:16:33] joe: I’m sure there’s there’s gonna be folks
who need to repair things. I’m sure there’s gonna be folks who need to go in and maintenance things.
But it- But
[00:16:40] Nick: not gonna
be an
[00:16:41] Geo: it off and back on aga-
[00:16:43] Nick: Showing up, walking around, making sure- that all the lights are flickering correctly.
[00:16:47] joe: Right.
[00:16:47] Nick: And then moving on …
[00:16:49] joe: If they can’t get a machine to do that, like a robot, that’s like a…
[00:16:52] David: here’s the thing
that, it like if th- there’s some like inherent bullshit in, sorry for the [00:17:00]
PG-13
[00:17:01] Nick: good. that’s still part of it
[00:17:02] Geo: No,
[00:17:02] Mary: You, you said that correctly.
Okay. Yeah.
[00:17:07] David: You didn’t mishear me.
Because if they’re claiming the, if the entrepreneurs, the Sam Altmans, the Elon Musks Daria Aamodei if the, if they’re claiming they’re gonna really make super intelligent models, couldn’t those models make themselves work more efficiently so they
wouldn’t need these freaking data centers? Like, so like,
if they’re really as good as they say they are, they should prove it by not ruining the Earth.
[00:17:39] joe: You know, these large language models and how they’re predict and their nodes, the billions of predictive choices, that’s what takes the computing power that they need. And so can they make themselves more efficient? , that means that they would have to then build their own chipsets [00:18:00] and,
[00:18:00] David: the AIs design their own
chips, right? And so you have a kind of like hardware recursive
[00:18:05] joe: We’ve seen that movie. We’ve seen that movie, and
[00:18:08] David: that movie.
[00:18:09] joe: It doesn’t turn out well for humans.
[00:18:11] Nick: Turning out good
for us
[00:18:12] Mary: let
[00:18:13] David: Yeah, well,
I just think, but so therefore I think the other explanation is their models aren’t that good.
They’re never going to
solve the universe and therefore
be able to make themselves work efficiently.
Like
[00:18:30] joe: or
you have the other end that potentially they need X amount of energy to run, and that’s just their need, right? They’re not gonna be as efficient as the human mind, right? So the human brain,
[00:18:41] David: there, that’s a great, that’s exactly where I wanted to go.
[00:18:44] joe: it runs on, like, what, 20 watts? I mean, that’s nothing.
That is, a dim light bulb, right?
[00:18:48] David: That’s a
sandwich
[00:18:49] Nick: why did you look at me when you said that? Are you calling me a dim
[00:18:52] joe: at,
looking across
the room
here.
I was looking for confirmation, like, yo, yeah, that’s it. I- it’s been a while since I’ve… Now with LEDs, like, [00:19:00] 20 watts is, like, when I was growing up, that’s, like, 100 watts, so.
But yeah, no, you have this power need, and so for training, things like that, it’s just, it’s not efficient. And I don’t, I don’t- You’re right. I don’t know if it’ll get there. But the other… What you can do, and I was reading a little about this curtailment. Have you heard of that? Where the idea is that
with-
Green power,
solar wind farms, at times they generate more power than can be used, consumed.
Mm-hmm.
[00:19:31] Geo: Mm-hmm.
[00:19:31] joe: And so, and then you get this like a duck curve. It’s kind- it’s called a duck curve, and it’s cool. So you have these things where they generate all this power, and so that’s during the daytime. Let’s say during the day, sun’s out, solar farm, it is just banking energy. But the problem is everyone’s at work, or the majority of people are at work not doing anything.
Or, you know, they’re not d- they’re at home using their devices, turning the lights on. And then at 5:00 PM or so they get home. Now the sun’s kinda [00:20:00] going down. The power generation is now going down, and then you get a spike in electrical use and you need to actually now use more. But now you’re at a low point of capacity, right?
‘Cause now you’re not generating as much and banking as much as you were earlier in the day. And so one of the things that’s interesting and one of the ideas that are pitched and AI is starting maybe helping to feed itself, is to figure those trends out
and then
modulate its power usage in relation to this duck curve so that when energy is
plentiful
and its consumer use is low, that it will now, that’s when all the jobs should come in and it will, go and, you have full bore data center processing.
And then when people need power- … you then down regulate, and you then modulate and AI can figure that out. It can start to [00:21:00] find the trends, it can start to self-regulate. It can begin to get to this, , negative power consumption if you want to say.
It’ll be totally a green thing. So yeah, it was a really interesting idea and folks are starting to try
to apply that to these data centers in these areas where you have green technology kind of- … banking up this energy and then trying to really efficiently use it. And that’s just lost anyway, right?
So it’s even if a data center’s not there, this is just lost
[00:21:29] Nick: W- it doesn’t get stored anywhere?
[00:21:31] joe: Can it get stored anywhere? Only so much can get stored, right? That’s , that’s a different problem, right? That’s battery technology. Batteries aren’t efficient, so they don’t convert 100%. So i- there’s a lot of things going on there because then you have more efficient batteries, then you can store more and you can have AI modulating.
So you can have AI figuring out these trend lines and, just kind of throwing it the opposite there that we don’t– it doesn’t have to build itself better. It can manage the resources better that are provided to it
[00:21:58] David: But there’s also this other [00:22:00] paradox that it, you know, it, what is it? The Jevons paradox. Do you know this guy, Stanley, William Stanley Jevons? So this
is a steam engine thermodynamics guy in the 1860s, and
he basic-
[00:22:15] joe: then? That guy was a
[00:22:16] David: No, but he was Making steam eng- He was trying, he was making steam engines more efficient. right?
And what he found was that the more efficient steam engines, rather than decreasing total usage of energy,
increased total usage of energy, ’cause now more people
wanted to use steam engines for more things. So there’s this para- this efficiency paradox that
Every
little
bit of
efficiency you
gain in, you actually offset it a little more by more people- using it. And I think that’s clearly in AI, the more powerful they get and the more useful they are, the more people will wanna use them. So the more
efficient even they Get that
may be [00:23:00] more than offset by then
How many people- actually use them.
[00:23:02] Mary: Mm-hmm.
[00:23:03] Nick: yeah.
[00:23:03] joe: I think that’s a really… That is a good point, and then you have that. But that’s also the problem then becomes you gate that with money or some resource, and then that just the technology becomes only affordable for people at the top end, right? So I, I think that’s the crux of it all, that you have this more…
Right? ‘Cause it’s the more consumers you have, the better, but at some point you’re gonna have to… Or you figure out, or you solve these issues some other way,, as curtailment or build data centers in space. Which is,
[00:23:37] Geo: On the cl-
on the clouds really.
[00:23:40] joe: The real clouds, yeah.
And there’s,
[00:23:41] Geo: Which
[00:23:41] joe: It’s, that’s a funny thing, ’cause that gets thrown out there a lot and it’s a mis- it’s somewhat a misunderstanding ’cause it’s like, oh, space is, really cold,
and
you need to keep the data centers cold, cool, so you put it out in space then. But that’s not because you’re in a vacuum, [00:24:00] you don’t have. Heat transference,
and heat is moved from one media through another. So there’s no air or water or anything, so how do you move heat?
And so
essentially what you do is radiant heat. So you would just, you build, big panels that would then dissipate the heat almost passively through.
And so the bigger your data center is the larger the panels are. So it’s like radiant heat, like so from a radiator, it’s just radiating heat out, so you gotta have this larger panel. So it’s really it can work, but it’s not like, oh,
[00:24:35] Nick: it’s, not as easy of effects as people think
right.
[00:24:38] joe: just not
like, “
Let’s put it in space. Like, it’s gonna be great. Like, it’s gonna be kind of one of these things.” And the cost of launching rockets has come down through the reusable rockets, so you can get materials up into space. You can have things like that. I know-
[00:24:51] David: say What you will about SpaceX, but that is really cool.
I mean,
[00:24:56] joe: Yes
[00:24:57] David: ugh. Anyway, that’s really cool
[00:24:59] joe: [00:25:00] Yeah
[00:25:00] Geo: So I have this kind of a
stupid
question, but I have like so many emails in my Gmail that like, It’s ridiculous. I don’t even wanna say
how many.
[00:25:11] David: You’ve never deleted an email
[00:25:12] Geo: I m- I try,
[00:25:14] Nick: What do you
mean- Wait, What you try?
[00:25:16] Geo: I try.
[00:25:16] joe: mean-
[00:25:16] David: spam from TJ Maxx
[00:25:18] Geo: I mean-
[00:25:19] joe: Exactly. It’s
[00:25:21] Geo: That’s right. And I’m wondering, like, is All these people similar to myself have all this data that’s just hanging around.
C- And even if you delete it, where the hell does it go? And then, so is that like, that’s pretty much the w- way it is, right? We’re
holding this data that
[00:25:40] Nick: it’s
[00:25:40] Geo: garbage more or less.
[00:25:41] joe: But those emails are- Right … relatively small
and they probably
[00:25:46] Geo: photos, right? Photos. You’re calling the photos
[00:25:48] Nick: calling your photos. garbage?
[00:25:50] Geo: No, but I’m just
saying- Photos are garbage … is it a matter of,
like, if everybody, like,
kinda
got
rid of a bunch of stuff, would that help?
[00:25:57] joe: kinda got rid of a bunch of stuff, would that [00:26:00] help?
[00:26:00] Geo: I don’t know. I,
[00:26:01] joe: Y- yeah, no, you’re right. I mean, it’s, to go back you would have to, a lot of people delete a lot of stuff, but mean maybe. Yeah.
Like, What’s people holding? I don’t know, videos versus photos, and do you limit people? But all that’s also gated by
[00:26:14] Nick: But that’s also like- the billionaire saying- “Hey, if you guys do this- that’ll help the environment.” Well, they’re holding- so much more. Yeah.
[00:26:22] joe: And
It’s gated
by money also, right? So if you want a terabyte of cloud service, that costs you this much money. If you want two… So there’s some of that built in already that, you know, it becomes this have
and have
[00:26:34] Nick: But you can go ahead and buy that- terabyte to put for your photos and videos.
[00:26:41] joe: Like
[00:26:41] Nick: nothing- is stopping you from doing that and owning it yourself.
[00:26:46] Geo: I’m
not
sure if I understand what you’re
[00:26:48] Nick: You can get a terabyte hard
[00:26:50] Geo: I don’t want a terabyte
[00:26:51] Nick: don’t want a terabyte
[00:26:51] Geo: drive. and
[00:26:51] Nick: hold everything that you have.
[00:26:53] Geo: I just wanna purge. I wanna get rid of it.
[00:26:56] Nick: Sure. You know? Go for it. No one’s stopping you. [00:27:00] N-
[00:27:00] David: Marie Kondo your emails.
Does
it
[00:27:02] joe: that’s right. Yeah,
that’s
[00:27:03] Nick: was talking about her photos, not her emails- you know.
[00:27:07] joe: I think the emails are pretty low. I mean, you probably can fit all
those on a floppy disk. Yeah I don’t think that’s the-
On a floppy
disk That’s not the driver. If people don’t know what that is, a three and a half it’s 1.3 megabytes, I think it was.
It’s like, I don’t know, like what’s the solution?
[00:27:19] Geo: think.
But it’s like, I don’t know, like what’s the solution? I don’t-
[00:27:23] Nick: I just gave you a solution.
[00:27:25] Geo: Yeah.
[00:27:26] Nick: Just get- I don’t- Own your own
terabyte. No, what I mean
[00:27:27] Geo: what I mean is it feels like it’s this thing that we’re just gonna keep building and it’ll never be enough.
[00:27:34] joe: that’s
[00:27:34] David: collective digital detritus.
[00:27:37] Geo: And nobody’s
really willing to go back to
not- That’s right …
doing computers and, I mean, even though- that- I think that sounds really nice.
but. Yeah,
[00:27:45] joe: right.
I don’t think… W- like I said, we were already on this trajectory, so I don’t… It’s probably not, the genie’s not going back into the bottle. And so really it’s about how do you…
[00:27:56] Geo: this trajectory, so I don’t… It’s probably not, the genie’s not going back into the bottle. And so really it’s about how do you
[00:27:58] joe: What is, what are the safeguards then put [00:28:00] in? You know, we were making all these climate change goals and moving forward and, I forget what it was.
We need a reduction by 2035 and, all these kind of things. But now it’s more can we hold that line? Like, we were making progress
[00:28:15] Geo: and and That, really does go back to politics
and, money. you know?
[00:28:19] joe: Right.
And I was
gonna go back, I had it in my notes, I was, when I looked, but the space thing and the International Space Station as a reference point, ’cause they use, they, they actually control their heat loads through radiant
cooling.
And so they have radiator panels the size of a tennis court to cool roughly 70, 80 kilowatts of heat load. And so that’s a pretty small fraction of what
a, a, an AI compute cluster would generate,
[00:28:46] Geo: can you give a comparison? Like, what’s that? Like- So, Is that a fraction of it? Like- I
[00:28:51] joe: I mean,
so something here was, like, the dumping. It was, like, one megawatt of waste would require approximately 1,200 square meters of [00:29:00] perfect radiator panels to go. And a modern GPU cluster generates hundreds of megawatts.
So that’s just kind of, the way to think about it. So that, that you would… This, you would need very large radiant panels. You also would have to face the cold side, so you gotta face away from the sun. So that’s the other thing you gotta do, because if you’re facing the sun, then you’re gonna heat up and you’re not gonna do, it’s not gonna be good.
So you gotta, actually gotta position your panels to face the dark side of space so that they’re actually on there. So yeah, it could be kind of cool, and it sounds cool. One of the things I read and I was looking at is lava tubes on the Moon
might be
a place to put your data center.
We had the episode with Ernie Bell on talking about the lava tubes on the Moon and Mars, and that would be a place where you can, potentially put them in this lava tube. You still have the radiation problem, the cooling only as in orbit, but you don’t have the continuous sun, so you can actually have [00:30:00] both sides.
You can have radiant, you can have radiant panels all over the place and not have to worry about that. So, but yeah,
so.
[00:30:06] Mary: But that gets back to your paradox then
If It becomes it becomes easier and easier to do.
it’s like, we just need to have
more moons. But I
wanted to ask
you, I kn- I wanted to ask you a question too about AI data centers,
about,
about water.
We’ve been talking about electricity.
And could you
talk to why
are AI data, what happens to the water in AI data centers? Is it… I have questions maybe maybe like- go ahead and start
and I might chime in, but I, yeah, could you talk about that a little bit?
[00:30:41] David: So definitely water is a scary headline number, right? You hear this huge amounts of water, and part of that… So there’s good news and bad news in the head- that number. Part of it is, as I understand, is that they take into account all
The water that’s [00:31:00] also used to generate the energy in the first place. it- for the data centers, right?
So th- approximately like 75 to 80, 85% of, the total water when people say this much water is being used for data centers is really just making the electricity, whether we used it
For a data center or not. So- really, I think the– there still is a ton, so it’s still… But it’s like 20% of the headline number,
which is still a
Huge amount. So it’s like 20 billion gallons of
water a year,
which sounds like an enormous amount, but it’s not 100 billion gallons a year, which
is what sort of the number that I– you can find out there So,
you know,
Basically the way
a way to think about it is right now
it’s About 0.3% of
all the
water in the public supply that
we have is going to cool data centers.
So 0.3% is not zero. It’s definitely- a lot of water, but it’s not even [00:32:00] 1% of all of our water. So it… T- 20 billion gallons is a ton of water. It’s
20 billion gallons, but it’s only 0.3%. So it… Relative to other general uses like general energy, overall energy generation, that uses way more water than the specific use of data centers.
That’s, that, that would be the way I would talk about it.
Joe, what do you think?
[00:32:24] Mary: Yeah.
And well,
I had another question, too, about that. So when the water is being used, can it be reused?
I that’s one of the questions-
I had. Is the water polluted after it’s been run through an AI data
center?
[00:32:38] joe: So I was gonna touch on that unless you…
I
[00:32:40] David: yeah.
go ahead, Joe
[00:32:41] Geo: ahead.
[00:32:41] joe: So I was gonna say, ’cause if there’s just, we talked about space cooling, and then we skipped over water, the way we actually do it, and there’s three
kind of approaches to that. In most of the data centers, what they do is evaporative cooling. So if you live down in Arizona, the Southwest essentially you use water, and then you [00:33:00] evaporate it to cool the surrounding air.
And so swamp coolers in Arizona, what they would call it, and I think there’s other names for other places. And so a lot of them that’s the technology they use. And so the water you’re using, you don’t really recapture it. It is lost to the environment, but also now you’re breaking, people studied the water cycle and how it flows.
And so now you’re using water in a different way, and it doesn’t… Any environment, it’s just loss. And so that’s really water heavy and you- but you use less energy in your cooling cycle. The other way, the other ways people are talking is air cooling, so like an air conditioner.
So
you can do that, but that takes more energy, ’cause now you got the energy for the data center the servers, and then you need energy for the air conditioners. And so then that gets to the point David was saying, how now that water use is on the electric generation side of it. And so when you add that in, you really are- You know, you’re right.
You get an … It doesn’t incl- You can use it for something else, but you’re not. You’re using it for an
[00:33:59] David: [00:34:00] not. Yeah, So, it, it d- yeah
[00:34:01] joe: right. And then the other, the third way and newer that people are talking about is actually liquid immersion cooling, so you would actually submerge the servers into water, which then could be, reused, and it’s not
So it’s not a pollution. It’s not being polluted. This isn’t the idea. It’s just water loss, so that water is taken out of the water table and then moved along. So that’s a problem. And the same thing if you put a artificial lake in, Tempe. So when they put the artificial lake in there it was an artificial lake.
[00:34:32] Nick: in where.
[00:34:32] Geo: Tempe. In a desert.
[00:34:33] joe: Arizona. Oh. And so you’re evaporating something like a million gallons of water, or a couple, few million gallons of water a year. That’s, it was just, you know, it’s not going back into the Colorado River and then, and flowing back into this artificial lake.
[00:34:46] Mary: just- lost.
[00:34:46] Nick: lost.
[00:34:46] Geo: Right.
[00:34:47] joe: So you’ve taken water from, , a source that was being replenished by snow cap melt and into the Colorado River, and then you’re filling a lake that then gets evaporated and it’s just lost. So that’s kinda the problem there. So [00:35:00] we think of these three kind of cooling approaches that, that was it.
And I was gonna throw … I also looked at these numbers and, a little bit. It was something in U- United States, we use something like 322 billion gallons of water per day, so that’s you know, it’s a lot of water. But the three biggest categories, and
I I was curious,
It’s irrigation, about 118 billion gallons a day.
Thermal electric power, 133, and then public supply is about 39 billion gallons a day. And that was a little bit ago, so these numbers, you know, this was- So
[00:35:31] Nick: for irrigation- … a couple years ago …is that for, like, home irrigation too, or is that just-
[00:35:36] Geo: Agricultural … agricultural? Agricultural. Yes. Yep, yep. So
[00:35:38] joe: farming.
[00:35:39] Nick: Okay. Not people just watering their grass for eight hours a day.
[00:35:43] joe: I think that goes into the public supply, ’cause that’s probably how they’re me- m- your meter is measuring that versus, farm use, so-
[00:35:49] Geo: I lived in Arizona- we did have irrigation, and then it would, like, they, it would flood our yard, like, every whatever… Do [00:36:00] you, did you ever
experience irrigation?
in- I did, yeah. No, Isn’t that crazy … Yep, yeah. Also, it would just-
I don’t know. I don’t know. Do they still do that They probably- I don’t know. There was,
[00:36:08] joe: know. There was, well, there was a lot of water restrictions put on when I was leaving, so this
[00:36:12] Geo: It’s Just weird … Yeah …2001 when
[00:36:14] joe: left, but they were starting to restrict, like, you only can water certain days.
[00:36:17] Mary: Unless you’re a data center, and in that case you’re like, “Hey,
[00:36:21] joe: And it’s
[00:36:21] Mary: it’s fine.”
[00:36:21] joe: Always find it funny when they put super water intensive things in, the desert. You don’t have a lot of water.
[00:36:29] David: or golf courses,
[00:36:30] joe: golf
courses,
[00:36:31] Geo: Yeah. number one
[00:36:32] Nick: number one. It’s
[00:36:32] joe: so
yeah, data center fits perfectly. Like, you know, why do people need to drink water?
[00:36:36] Mary: Well, I had ano- I had another question too about the water, and and so I’ve heard that data centers need to, need a ton of water, but they at the same time need they operate best in, in dry environments.
……is that a- Mm-hmm. Am I right on that?
S- okay. [00:37:00] So-
so they, so
[00:37:01] Nick: Yeah
[00:37:02] Geo: Is that because of the equipment?
[00:37:04] Nick: because-
[00:37:05] David: you don’t want the Chips getting wet.
[00:37:07] joe: right.
[00:37:07] Mary: That’s
[00:37:07] Nick: Yeah.
And You want a nice solid temperature- for the computer components.
[00:37:12] Mary: So-
[00:37:13] David: want to use the water to carry the heat away and then evaporate the water. Like s- you know, kind of like sweat
in, in a sense, right? We use water to get the heat out of our bodies by sweating it off and having it evaporate. It’s effectively what we’re the, having the data centers do, but just at a massive scale
[00:37:31] Mary: I
see
So they operate best in like the desert
that
doesn’t have a lot of water. And, but, you know, then, and then we have to pump in the water from somewhere, right, to cool these things down. What a mess.
[00:37:47] Geo: The whole thing just is-
Yeah
ridiculous.
[00:37:51] joe: then- you have the- Okay
point also that, a lot of these I mentioned they’re in rural areas, but there’s also now people are like, “Oh, we should [00:38:00] really put ’em in, densely populated areas.”
[00:38:03] Mary: Not- It’s
not densely populated, but Hammond,
you know-
[00:38:05] joe: Well, I mean, you’re- They’ve never had a bad
[00:38:07] Mary: id- There’s never
been a bad idea
yet- …
that they haven’t gone, “Hey, hold my beer.”
You know?
[00:38:12] joe: I mean, you’re close enough to… You got two things there. You’re close to a large densely populated- Mm-hmm … major metropolis in Chicago, and you have the largest source of fresh water on Earth in
[00:38:26] Mary: lake-
[00:38:26] Nick: great
[00:38:27] joe: lake- Yeah
And
associated watershed reservoir.
[00:38:30] Mary: Yeah. Hammond can’t wait to piss that all
[00:38:32] joe: that is,
that’s, those are two big, Yeah.
[00:38:34] Geo: Yeah.
They’re like-
[00:38:36] joe: Two big
[00:38:36] Geo: But maybe they’ll get the Bears too. yeah.
[00:38:39] joe: Maybe we’ll, maybe improve our power grid. I mean, you know, do something
[00:38:42] Nick: useful, you know?
Ooh, what if The Bear stadium is on top of the data center. yeah. Whoa.
[00:38:47] Mary: Yeah. It ju- But
yeah,
[00:38:49] Nick: I mean-
[00:38:49] joe: I- No.
[00:38:50] David: blown
[00:38:51] joe: But I’ll
[00:38:51] Mary: yeah,
that,
that’s the other thing, too. I th- I think that I mean,
I…
like
it
feels like a bubble to me, you know? Like
all …of [00:39:00] this massive
build out, and we just can’t build things fast enough.
And
like
right now in Hammond, they’re going, I, they’re going back and forth. There’s this company called CoreWeave. I I have, I’ve,
[00:39:14] Geo: I have, I’ve,
[00:39:15] Mary: I would imagine, have you heard of them?
Oh, okay. And so, and they’re promising Hammond like birthday parties and ice cream and pony rides. If you could just only, you know, if we can just come into your
town
and-
[00:39:29] Nick: do Mary?
[00:39:30] Mary: They want… Their CoreWeave is an is an AI data center- place, and they want to they have the mayor’s attention because they’ve floated some
[00:39:40] Geo: like
[00:39:41] Mary: bonus or something. Like they’re gonna, like the city’s gonna get $25 million from this company,
and
I’m like, “
Says who?” You know, I mean,
I don’t know.
Do these AI data centers just
like promise they gonna get a tax break
Oh, I’m,
oh, I’m sure. I’m sure they’re not gonna pay any [00:40:00] taxes forever, you know? But
[00:40:03] joe: I
[00:40:03] Mary: don’t under- you know, like, you know, I guess do these centers come in and promise the moon to the cities- and then when it comes time to pay up, they go, “Oh, you know what?
You know, we, you know, we’re running so hot right now. We’re making a bunch
of promises.”
But when it comes time to…
You know,
Is that,
Have you- Run across that in, in any of your, in any of your research?
[00:40:25] David: I don’t know enough about the kind of community company kind of interface. I mean, it reminds me, the way you describe it, of what I heard about when the fracking was going on and the
f- the, and the companies would, go town to town if they found that there was a good place to do fracking in, in, rural Pennsylvania, and sell the towns on these too good to be true things and come in, and then there’d be reports of earthquakes and, nobody o- i- it
was all… It just smells like a deal with the devil kind of,
you know, for lack of a [00:41:00] non parochial metaphor. But
[00:41:02] joe: Yeah.
[00:41:03] David: yeah.
Anyway
[00:41:04] joe: I think that’s one of the problems with this now is that data centers where they’ve gone, it’s all about corporate profit. And then to Mary, to your point- … I think there has to be a switch because now you’re saying, let me build this in your community. Let me build this in Hammond next to your home.
Right. It’s gonna be great. And I think the companies haven’t really talked about community engagement and how they’re going to take the profits that they make and put that back into… You know, so will the schools get, free data storage? Will infrastructure be improved?
I think
that’s
[00:41:38] David: at least give free Wi-Fi to the
[00:41:41] joe: Exactly. Exactly. And not kickbacks to the mayor. Like, you know, actual,
[00:41:45] Geo: Real, Well-
[00:41:46] joe: real community things
that you’re gonna help anybody.
[00:41:49] Mary: it’s j- it’s just gonna help us.
[00:41:50] Geo: Yeah. right.
[00:41:51] joe: Not gonna build a new body … I mean- it’s just gonna help us. I mean… Yeah. Right. Right. I mean, that’s… But but your community would be the one directly impacted, and so if you’re coming into that community or the watershed community, let’s say that.
[00:41:58] Mary: Mean,
[00:41:58] joe: you know- right, so that would be Lake [00:42:00] County, you know, for instance, because now they’re talking Munster, Hammond, Crown
Point.
So you- Oh, for sure. You’re right. Hobart. So
you would have several data centers coming in. And so what are these communities getting out of that? Are the libraries gonna get some books , and maybe some free Wi-Fi and some updated computers or access to just the server?
So you-
[00:42:18] Geo: you’re talking about ways- I
[00:42:19] joe: think there are ways-
[00:42:20] David: To get back to the beginning thing we were talking about, right? Like, the electric company, when they build a power plant in the 1800s, everybody in the town’s lights turned on, right. Like, you got something very clearly beneficial at the community level when these companies came in to exploit your natural resources.
So it only seems like the least they could do is give us Wi-Fi
[00:42:45] joe: Yeah.
[00:42:46] Geo: And I
[00:42:46] Nick: And I, honestly, I don’t think the trade-off is worth it for just
[00:42:49] David: No. At least they’re not selling it in
any
[00:42:52] joe: right
[00:42:52] Nick: It’s
not gonna be a good-
Wi-Fi … Wi-Fi It’s gonna be a- Like- It’s
gonna be subpar at best
[00:42:56] joe: yeah
[00:42:57] David: be dial-up.
[00:42:58] Geo: And
they’re
giving you noise [00:43:00] pollution. They’re giving you, you know, all
[00:43:03] Nick: Yeah.
no the benefits
don’t outweigh the cons.
Like- I know
[00:43:08] David: Or at least they’re not making them obvious
[00:43:09] joe: if they’re making, $100 billion then they can patch up the roads, they can invest in schools, they can invest in libraries. So you can sweeten the… I mean, I think if your community- They- But they won’t
sweeten the pot. Well- If, if- I think at some point they’re gonna have to because- They have to look at the impact … I just don’t think… I think they’re gonna run into a wall. Then don’t get out. I get that they’re promising things now. But for other businesses- For now, right.
That’s right. Well … are not gonna happen. Right. They’re not, they’re go- Yeah … they’re promising the moon right now if they can- Yeah … if only they can come and pollute our water and- Oh, we’re just a little AI company … And drive up our energy- Just a small one … sky high. Small one. We’re gonna need a billion gallons of water,
patch up the roads, they can invest in schools, they can invest in libraries. So you can sweeten the… I
[00:43:18] Nick: But they
[00:43:18] Mary: But they
have no-
[00:43:20] Geo: But they
[00:43:20] Nick: But they’re not going to
wants something- And it, if they promise it- Right … it’s not gonna be an actual thing that they do …
because
[00:43:25] joe: think… I think they’re gonna run into a wall.
[00:43:28] Geo: n- n-
[00:43:29] Nick: don’t get me wrong You know? I guarantee they’ll promise it eventually.
But fulfilling these promises Well… Well … are not gonna happen. right
[00:43:35] Mary: They’re not, they’re g- they’re promising …the moon right now. If they
could, if only
they could come in and- pollute our
[00:43:41] Nick: water and-
“Oh, I’m just a wittle AI
center. Don’t look at me” sky drive up our stock. A small one. “It’s just a small one. they need. worry.” A few
gallons of water,
[00:43:49] Mary: Yeah, fingers crossed.
But they’re making so many promises right now, how could they ever fulfill these financial
[00:43:55] Nick: Especially with them not making the money that
they need back. Well- Right …
[00:43:58] joe: have a, do have a [00:44:00] fun a fun list. I don’t know. I have a few kind of lists about this. I figured but where data centers have gone in, in heat use and to the community, and it, a lot of these are really interesting. Most are in Europe, but in Microsoft… So we, the, we we’re hearing a lot about data centers now, but Microsoft in 2011 in a white paper coined a term data furnace, and their idea was that they would take their server racks and replace home heating systems.
So homeowners would get free heat.
[00:44:32] David: That’s awesome and then you would have this whole, and so-
[00:44:36] Geo: how do they- … one of those companies-
How do they get it from the data center to the person’s house? I,
[00:44:40] joe: Yeah. Yeah, yeah, yeah. And get your
[00:44:41] Geo: You
[00:44:42] Nick: You would have a giant rack of just- They come to
[00:44:44] David: now you have a
problem “
[00:44:45] Geo: Okay, we got your heat”
[00:44:47] joe: there’s, steam. There’s ways you can transfer
[00:44:50] Mary: are immediately peeing on those server racks … so 100%. what
[00:44:54] joe: did were- 100%. Exactly … so in, there’s a UK startup called Deep Green, so these are now [00:45:00] real-world cases,, where they install server racks directly under swimming pools submerged in oil, and the heat from the servers warm the pools.
[00:45:07] Nick: seen those.
[00:45:08] Mary: there
[00:45:08] joe: seen that, yeah. And so you go, you had the Equinox Par- Paris Data Center in 2024 Summer Olympics that heated the Olympic Aquatic Center and it’s gonna be free for the next 15 years via a heat network. There’s a Norwegian lobster farm that’s heated by a nearby data center, and I had to have this.
We had a turpentine farm on the
[00:45:29] David: yeah. Uh-huh.
[00:45:30] joe: I saw this story… have this heat and this lobster farm, so you have this really interesting thing that you could then think about the heat generation. Are there ways to use that?
Now once again, is that community… Are there ways that you can heat- A homeless shelter nearby. Or, are other things in the community that would be useful that you are you’re making this trade-off- Which … but
[00:45:53] Nick: your breaking point? Right,
[00:45:54] joe: right.
I mean, you’re also gaining.
So I think that’s where these data centers have to come in with community [00:46:00] leaders and get creative and say, “Well, we know what this is all about. We’re not gonna try to blow smoke,
[00:46:05] Nick: Is this us trying to aim at
data
centers to hire us
as community outreach?
[00:46:10] joe: outreach? Sure, ‘
if
[00:46:11] Nick: they’re-
[00:46:12] joe: They’re paying up the bills. I mean, you know,
[00:46:15] Nick: They storing our episodes
for free.
[00:46:17] joe: the
[00:46:17] Geo: wouldn’t hold your… Yeah …I would not hold your breath, not hold those up. But yeah, you have… I mean, and in just different locations too. I mean, we talk about rural centers and we- and mostly America, but you have ones that are buried under like, you know, old Cold War bunkers in Stockholm.
[00:46:34] joe: So they kinda go down there. It’s cold. I saw one in Antarctica. It’s a functioning data center at the South Pole, and so you don’t really have a cooling problem down there ’cause it is cold. So you have these places that are Kansas City, SubTropolis in Kansas City, so it- it’s in a, it’s in a cave.
And so you have this massive carved into an underground limestone mine. And so these kind of old mines [00:47:00] that we had, can you put data centers down there? Which naturally in our episode where we talked about living underground can you use these naturally occurring places that are cold that you don’t have to then use as much water to…
‘Cause the water is what we’re using to cool these, so can we actually find other ways to cool these centers? And other technologies-
[00:47:21] Geo: you might, you’re just gonna open another can of worms at that.
point. You know what I mean?
[00:47:25] Nick: We’ll figure, we’ll cross that bridge when it comes to it. We’ll
[00:47:28] joe: do that episode next season Well, that
[00:47:30] Mary: bridge when it comes to it. Yeah,
We’ll do that episode- Don’t worry … next season. Well, then th-
[00:47:31] joe: you- Oh, I’m just throwing
[00:47:32] Mary: also… Yeah, no I hear that But
like,
I wonder if
[00:47:35] Geo: like, like
[00:47:37] Mary: we have to like hit them where they live as far as like financially.
mm-hmm. You know, I mean, they don’t … I mean,
what is right now it- they’re just able to just run rampant with like no consequences
whatsoever. But what if they actually had to pay taxes on- …the AI data centers that, you know, that they’re [00:48:00] creating?
[00:48:00] Nick: I think that would be up to the politicians- right. to go ahead and- grow a pair.
[00:48:05] joe: Yep.
Oh.
[00:48:06] Mary: Right on that problem .yeah.
[00:48:07] Nick: Grow them a
[00:48:08] joe: pair. Or- The data
centers are going in where they’re going
[00:48:10] David: Or that.
[00:48:11] joe: because regulations
[00:48:13] Mary: are lax.
[00:48:13] joe: are lax.
Mm-hmm.
Right? So …the areas that they’re … If you’ve, if you track where the data centers are primarily going in- Mm-hmm … it’s because regulations don’t exist for these things. You’re ta- you know, Mary, you’re, I mean, you hit right on the head there. That’s why. So you can almost go, “Why are they building this in, you know, wherever?”
The worst place to build the data center by all means.
[00:48:33] Mary: As we like- Why are you buil- why are You building it in Hammond? Because Hammond …has never
seen a bad idea- that they didn’t wanna throw money at.
Okay. So yeah.
But
yeah, I think
there’s one a data center on State Line Avenue or something.
Yeah. Or anyway,
In
Hammond. Yeah. They want to expand it.
[00:48:50] Nick: Hmm.
[00:48:50] Geo: they already voted
that
they aren’t, didn’t
[00:48:53] Mary: I,
well, they’re June ch- 30th- Oh 2026. So when this, you know, we’ll see.
[00:48:59] Nick: It’s already passed. [00:49:00]
[00:49:00] Mary: It’s already passed.
I,
well, you know, it’s, I, it’s hard to s- it’s hard to say like I, you know, I’ve been going to like these city city council meetings and
[00:49:08] Nick: like Oh, no, I was saying as of the day
[00:49:10] Mary: Yeah,
that’s pa- it’s already passed.
But I think that I get the sense that, you know, even they, they are yeah, these companies seem kind of,
[00:49:19] Nick: You
know,
[00:49:19] Mary: too, yeah- icky
or too good to be
true or
[00:49:21] Geo: they’re- Right … They might be kind of figuring
it out. Yeah.
[00:49:24] Mary: may- Yeah,
I mean, don’t worry. I mean, I’m sure they will, they’ll figure out a way to, to make it happen but,
But yeah, there’s just, Do you … I don’t know. I
I just-
Yeah, go ahead. Go ahead,
[00:49:36] Geo: No, I’m sorry. I didn’t mean … to I just think it’s crazy ’cause I remember learning about computers when I was young, and they would talk about
The first computers took up huge rooms and,
like-
and you had, you know.
And now, ooh, now we have- everything, everything we have are now on this little
smartphone. But all we’ve done is then just-
[00:49:59] Mary: The Huge
[00:50:00] rooms
are still
there
[00:50:00] Geo: You know? Yeah. It’s like- you
[00:50:02] Nick: replace one- with
another
[00:50:04] joe: Yeah. Yep.
[00:50:05] David: Yeah.
I just wanna make it a little bit worse
[00:50:07] Geo: Oh. They do. Yes.
[00:50:08] Mary: we don’t need another huge room. Yeah
Yeah. You know? It’s like- you replace one with a nine. Yeah. Yeah. Well- Please do. Yes. Yeah
[00:50:09] David: in ad- in addition to all the companies, there’s also the military
that’s building lots and lots of data center infrastructures, and they’re getting these really long leases, 50-year leases, where they can then just… Who– You know, I mean, like, I do trust the military
but, you know, come on.
This seems like the military industrial complex on steroids in some way when you start doing the data center integration into the AI with
autonomous military capabilities. It just–
[00:50:45] Geo: Surveillance?
[00:50:46] joe: Yeah,
[00:50:46] David: even worse.
[00:50:47] joe: Yeah.
[00:50:48] Geo: I know.
Well,
[00:50:50] David: All right. Anyway.
[00:50:50] joe: to make it even worse, ’cause we talked about power, and I don’t wanna bash it, but, the rush to find power sources to improve our grid, you [00:51:00] have nuclear power. And so now we’re starting to talk about reactivating or activating, shuttered nuclear power plants to bring them back online to feed the beast.
And so, like, Three
[00:51:11] David: Three Mile Island
[00:51:12] joe: Is there, which was– almost had a near meltdown. And so… And that’s just kind of, for… A meltdown is when the reactor’s core overheats and you can’t cool it down with water, so the nuclear fuel begins to melt. And then there’s not coolant to absorb that, so you have a disaster
[00:51:28] Geo: go wrong?
[00:51:29] joe: you know, heads
[00:51:30] Mary: I’m sure everything will-
[00:51:31] joe: I’m sure everything will be fine.
[00:51:34] David: Island is the Crane Clean Energy Center.
[00:51:38] joe: Yeah.
[00:51:39] Geo: Ah.
[00:51:39] Mary: it’s clean. It
has
clean. in the name. Right. It’s, it, w- oh, it’s, it must
[00:51:43] joe: that’s the idea of it. So no, you have that.
[00:51:45] David: But it’s Microsoft, right? Yeah
[00:51:47] Mary: Yeah. oh boy.
[00:51:48] joe: But
you do have these ideas, where you go, these, I w- I talked about these submerged kind of centers as you go in liquid.
And so you’re talking about submerging them underwater, ocean, submerged ocean [00:52:00] centers, so you can, place them there where you can dissipate heat, and you got tons of water, you know. So you do have these,
you know-
[00:52:08] Geo: that-
[00:52:08] Nick: then the- … environmental effect is gonna be harming what, the fish?
[00:52:12] joe: No,
[00:52:13] Nick: I mean, For changing- the temperature of the water down there?
[00:52:16] joe: It will be fairly negligible because of so much, because you have so much water that you’re not, the heat load now won’t be, I mean, so
[00:52:25] Nick: You’re still producing the same amount of heat- but it’s just in a larger area.
[00:52:28] joe: So it will disperse … so like- Right. Right … you’re
[00:52:31] Geo: you’re
[00:52:32] Nick: hurting the environment of-
[00:52:34] joe: You-
[00:52:35] Nick: said creatures. Yes.
[00:52:37] joe: locally. I mean, you might,
you
[00:52:38] Nick: not humans that you’re doing it to … but You’re doing
it to- sea creatures.
[00:52:43] joe: creatures.
Right, right. And you
gotta think about what you’re doing, but you go low enough, then you…
Are you just you know, there’s a question, you know, you put things under stress, heat stress, they, you know, either A,
They will evolve and adapt, or you get a
new,
[00:52:55] David: die.
[00:52:56] joe: you get a new, you get a new, cohort come in and [00:53:00] live there. So around a data center, do you have different organisms that pop up and, you know, right.
[00:53:05] Geo: Right.
[00:53:06] joe: mean, you do have
heat, you have heat
vents in the ocean. It’s not like there’s not, hot things in… I mean, I’m just, you know, kind of putting out
there that, that you kind of bury it there. But yeah. That’s
[00:53:14] Nick: Just
like all humans, yeah
we bury our problems. Yes.
[00:53:17] David: You would drive evolution of new microbes
[00:53:20] joe: probably would, yes.
[00:53:21] Mary: Oh, you know
[00:53:23] joe: Microbes.
[00:53:23] Mary: Let’s get
[00:53:24] David: Definitely
[00:53:25] Mary: You know what? We haven’t gone a single… We haven’t gone the whole time without mentioning The Thing, which we’re all Thing fans, right?
So would it be great with
a
[00:53:34] joe: Antarctica out there. Right.
No one took the
[00:53:36] Nick: I wasn’t going to. I didn’t want to.
[00:53:37] Mary: I’m taking the bait now.
Like, imagine, like, the heat vents and the new microbes and
[00:53:43] joe: you thaw
[00:53:43] Mary: Wow. Anyway, Yeah, yeah.
So
yeah- …we unthawed The Thing. But what… Okay, I wanna get super petty right now and talk about, like, the dumbest things that I, wanna talk about really dumb things that AI suggests to me, and this is some…
Okay,
[00:53:58] Nick: so.
Wait. Hold on
[00:53:59] David: [00:54:00] Is this a segment?
[00:54:01] Nick: quick.
[00:54:01] Geo: This-
[00:54:01] joe: That’s right. A new feature. a
[00:54:03] Mary: New feature …new feature. Okay, let’s just talk… I mean, because we’re, like, ruining the planet and destroying the water and driving up our electric bills for these… Okay, so, Georgia and I are children’s librarians, and one of the things that we do is we print out lots and lots of coloring sheets for little kids, right?
So, like, I’ll get… Sometimes I’ll just download, like, a batch of, like, 16 different coloring sheets on a particular theme, and some AI dipshit whatever comes up and says, “This appears to be a long document. Would you like me to summarize it?” It’s a
[00:54:37] joe: a coloring sheet.
[00:54:37] Mary: coloring
sheet. It’s coloring
sheets, okay? don’t need it summarized.
I just- … get out of my face. Let me
[00:54:47] David: Thanks, Copilot.
[00:54:49] Mary: I don’t… Every
ti- Okay, so I am a proud Luddite when it comes to AI. I mean, I’m sure I’m using it without realizing it, but every time they’re like, “Would [00:55:00] you like to use this?” No. Click. You know, it just… It’s a constant game of Whac-A-Mole just to…
Like, listen, I don’t need my coloring sheet summarized. I know how to write an email. It’s not a problem. I don’t… It,
[00:55:14] Geo: You know. And then it
[00:55:14] Mary: then-
[00:55:15] Geo: can’t even get voice to text right.
[00:55:18] Mary: Oh, okay. There you
go.
[00:55:19] joe: boy. Right?
[00:55:20] Mary: Let’s,
yeah, talk about-
[00:55:21] joe: you- ‘
[00:55:22] Nick: cause you’re shouting at your phone if you had
[00:55:23] joe: a data center in Hammond- Then-
then you wouldn’t have this problem.
[00:55:27] Mary: Let’s- Just ’cause you’re shouting at Google. If you had NASA and Hammond- Right …you wouldn’t have this problem.
That’s right.
Yeah.
I have
a lot of other
problems. than Hammond. That’s
true.
Oh my God. I I have, I was gonna go some fictional things that I like talk… I mean, we can think of supercomputers gone wrong in fiction- Mm-hmm … but I went back. There was,
[00:55:41] Geo: a- … life is so dystopian, do we really need dystopian?
[00:55:44] Mary: really. Do we need to bar? Do
we need to bar
trouble
here?
[00:55:47] joe: out two
[00:55:48] David: there any happy fiction?
[00:55:49] joe: Short– I wanna point out two short fictions- Okay … that I think are kind of cool, and one was The Machine Stops, E.M. Forster
[00:55:58] David: Yeah [00:56:00]
[00:56:00] joe: where humans live underground in pods, and they’re comp- completely dependent on a global machine for food, communication, and social connection.
The machine runs everything until it doesn’t. And so you can imagine in 1909, this is before internet, radio, commercial, aviation, so you really have this kind of dependency on something and then when it collapses, like, what happens? So very good thought piece.
And then Isaac Asimov, who I think just, you know, really good idea generation. The Last Question, and it’s about a supercomputer is asked whether entropy can be reversed, and then the supercomputer keeps evolving and computing until after the heat death of the universe.
[00:56:44] David: Cool
[00:56:45] joe: Both of these are really short. I think they’re, like, nine, 10 pages long, both of them, so.
[00:56:50] Geo: AI could summarize it.
for you. it
[00:56:51] joe: could- summarize it. AI could …summarize
it very well. You can ask
AI. I
[00:56:55] David: to
add a tiny bit of level- levity,
right? The Hitchhiker’s Guide to the Out- Galaxy, [00:57:00] the
AI, he fed it everything, and what’s the answer to the universe?
42
[00:57:05] joe: long- and-
So long, and and Thanks For All The Fish, right? Yeah. So,
[00:57:10] David: Exactly
[00:57:11] Nick: you have the-
[00:57:11] Mary: we don’t have
[00:57:12] joe: we have data centers down there
[00:57:13] Geo: Oh,
[00:57:13] Mary: if we put the data centers down there, then they
all evaporate and
boil.
Yeah.
We get a,
… little fish
boil there.
[00:57:19] Nick: Yum,
[00:57:19] joe: yum. Yeah.
[00:57:20] Mary: Good eating.
[00:57:22] joe: But
yeah, that’s, you can go there. Yeah. So I was gonna, Dave, I was gonna go and touch a little bit into biology maybe, and we talked about, we were joking about the aquatic creatures that, that Nick is worried about by the data centers.
Yes. Yes, I am.
[00:57:35] Nick: Yes, I am.
[00:57:37] joe: But for, but when you stress cells, I was kinda thinking of biological metaphors to our data centers they decide what to sacrifice, and so is that… I you were pointing this out about AI, that is AI going to go, “Here’s how I should throttle or shut down non-essential things.”
I mean, is that a biological equivalent [00:58:00] or, you wanna touch on that? Your research kinda touches into that, so I wanted to give you some,
[00:58:03] David: Yeah. So, you know, we basically study resource allocation, but just how cells do it instead of economies or households or individuals, but Cells, they don’t have as much choice. There’s not a lot of free will if you’re a cell. You know, the primary directive is to grow and proliferate. Of course, that, you know, not if you’re an adult human, by and large, most cells, that’s what they’re trying to do, is take nutrients in from the environment, build another copy of themselves. And so when they ha- encounter stresses in the environment, no matter what they have to do to cope with that whatever positive thing that they need
to do, they always have to do the same negative thing, which is slow down. They can’t grow as
fast because
they have to spend some of their [00:59:00] energy on coping with the stress that’s in front of them before
they can even consider thinking about growing.
‘Cause if they tried to grow before
they coped with the stress, they would just pop. So when I find it hard to extrapolate that insight that we have about How constrained
cells’ budgets really are onto what seems like a system that humans have set up to be utterly hungry, insatiable,
rather than
ever
having to balance
a
budget
or
ever experience stress
where you have to have some trade off. it just
seems like we’re just feeding them raw sugar,
you know, in the form of data centers.
And,
So I just,
I don’t think
these architectures are
inherently adaptive in the same kind of [01:00:00] way that you and I would
be familiar with, where you actually have to think about
some trade off and some Yeah you can’t have your cake and eat it too, and these models can right now. And I think that’s that hard constraint
That they can
as
long as the world
suffers
and we take the water and we raise the temperature of the climate.
Like,
there
is no free lunch here.
It’s just the budget and the cost and the stress is being put not into the models, but onto us
[01:00:32] joe: What I mean is that, I mean, will the machine, if it becomes self-aware- Yeah
[01:00:39] Geo: Yeah.
[01:00:39] joe: will it care? I mean, will it then go will it try to,
like, regulate itself, right? I
[01:00:44] Nick: on what
[01:00:45] joe: Yeah.
[01:00:46] Nick: becoming self-aware with? Like-
[01:00:48] David: So the best essay I’ve read on this precise topic was by this guy, Siddhartha Mukherjee. Do you know this guy? He wrote the
[01:00:56] Mary: No Cancer, the
[01:00:56] David: cancer.
the great book about
cancer.
[01:00:58] joe: Yeah. The [01:01:00] malady of all… Yeah, right. Yeah,
[01:01:01] David: yeah, The Emperor of All
Maladies.
[01:01:02] joe: Seth wrote a new book, right? Is
[01:01:04] David: Yeah. he has several new books. So what he’s working on now, and he’s, so this essay is a, something I read in The New York Times a few weeks ago. It’s part of their series, What Drives Us. So it’s in trying to get thinkers to tell you about what’s motivating them. And so he’s been Trying to cure cancer his whole career. And lately, he
has turned to training an AI model
to
try to make cancer
drugs, right?
[01:01:29] Mary: Mm-hmm.
[01:01:31] David: essay is
[01:01:32] Nick: cancer drugs?
[01:01:33] Mary: you said cancer drugs?
[01:01:33] joe: Yeah
[01:01:33] David: yeah, drugs to, to cure people of
cancer,
right?
That he wants to build,
[01:01:38] Nick: Got it. I was like That’s cool. Yeah. Yeah, right Thought he was like-
[01:01:39] joe: Useful.
[01:01:40] David: AI to cure cancer, to
make drugs.
[01:01:42] Geo: he was like, “
[01:01:43] Nick: I’m gonna give everyone cancer. Here’s the drugs.”
And
I’m
like,
“Oh, damn, that got
[01:01:48] David: that except good.
[01:01:49] joe: That’s good. Yeah, this isn’t the Spider-Man villain episodes, okay?
[01:01:52] David: No. This guy seems like a really genuinely, like, wonderful guy. Very thoughtful. The essay’s about [01:02:00] him talking to his mom actually in part. But the thing is he finds that the amount– the way that this model is, like, learning to understand cancer protein folding and how small molecules that might help patients could interact with these proteins at the very atomic level, and it was learning these kind of chemistry and physics rules. He just realized that this thing, once It’s starting to
Work a little bit. But once it works
it’s
gonna work because it went far beyond whatever
he could do,
and it’s learning on its own, and it figures it out. And
so
he started talking about it like his own child and like, you know, you can hope you send your child out into the world with good values
but ultimately what they do when they grow up is very much on them.
[01:02:55] joe: Right
[01:02:56] David: And so if these [01:03:00] models get smart enough to the point that they can really do things, what He argues is the best thing we can do is try
to give them good values, like we’re their parents
and try to…
[01:03:11] Nick: has not happened at all
[01:03:12] Geo: Right, has
[01:03:13] David: Which is certainly not what
we’re doing right?
now, right? But, like, to raise them in a nurturing
way and teach them what
human is.
Be- teach them about humanism,
right? And, like, all the beautiful things humans have done, and that is really, like, our last best hope at getting these models to
actually be nice to us
later on
[01:03:35] joe: or should you not teach them human values, but cellular values that you have to respect the limits of your resources, right? Because human values, there’s a lot in there, but if you just take it down to what you were talking about, the free will of a cell is just to grow, proliferate, live within your means.
That’s actually what we’re not teaching AI right now. Right now we’re teaching AI [01:04:00] you can have as much power, as much water,
[01:04:02] Nick: Oh, I thought you were gonna say it, We’re teaching it racism …
[01:04:05] joe: gener- well,
[01:04:05] Geo: porn. In gen- Well, I mean- Well, that
too.
[01:04:06] joe: Yeah. What did David say? Every data
server is a, you know, a,
[01:04:10] David: To a- first approximation.
[01:04:12] joe: It’s
just porn.
So yeah,
I think there are these
things where you can go and if that happens I still don’t know if, I don’t know if that jump singularity is gonna happen. I’m still, you know, I think there’s things that
y- is
the hype and the super hype and then the mega hype, and I think that’s one of them.
But, just hedging the bets if it happens, I do see the fictional scenario where AI just goes, “Your humanity is not worth” … I just feel like it’s kind of like we we’ll be in some war and we’re gonna lose because we need water and it’s kind of…
And we’ve used it all for the data centers.
[01:04:47] Geo: That’s
[01:04:47] Nick: can
[01:04:48] David: is exactly like Waterworld.
[01:04:50] Nick: like
[01:04:51] joe: yeah.
[01:04:51] Nick: you
[01:04:52] joe: We just flood their GPUs and we win. Yay.
Yes, no, but that was something, I think it’s yeah it, can we [01:05:00] actually train AI to think about natural resources and to… And I think people
are I,
I think there are folks trying to do that and find better ways
that you
grow,
and if
you become more efficient, you grow and we keep all the pegs, the legs of the table growing with it.
That’s the community, that’s energy resources, water resources. What are we doing with the
technology?
Not just consuming it, but are we doing, are we trying to find cures to human diseases and cancers and things like that? And having fun too. I mean, I think that’s part of human nature.
But,
Is it just trying to compile…
Right? Right? I mean, it’s asking you to summarize your document because it wants you to get hooked. It’s like a drug.
And if you get hooked to
[01:05:45] Geo: if you get hooked to it … I’m already, I’m already-
[01:05:46] Mary: addicted
to enough things.
I don’t need, I don’t to be addicted
to AI. Yeah, I know.
[01:05:51] joe: addicted
[01:05:52] Nick: what’s really cool? I’m already-
Drugs.
[01:05:54] joe: I know.
Really You know, I
don’t need,
I don’t need no digital
drugs.
[01:05:58] Mary: what were you gonna say, Georgia?
[01:05:59] Nick: Oh, [01:06:00]
[01:06:00] Geo: I just think the bottom line is we’d just be so much better off. I’m not saying, like, we can totally disconnect, but-
if- we
[01:06:08] Nick: everyone on drugs.
[01:06:12] Geo: But I
think if we just took less… If we just took more time away from our devices and, yeah. Just in general. Yeah.
[01:06:20] Mary: Well, every generation- is a response to the one before it, so- maybe, like, the-
m- maybe the next
generation is going to be like, “Oh, this is really cool, this pen and paper thing.” You know, all bespoke and everything.
[01:06:32] Geo: Right.
[01:06:33] Mary: So yeah, I…
Oh, boy.
[01:06:35] Nick: even the disconnecting part is hard for everyone.
[01:06:38] Geo: It is.
[01:06:38] Nick: alone
have a smartwatch on
your
[01:06:41] Geo: It is. hard. Like,
[01:06:42] Nick: as soon as It buzzes, you’re gonna look at it because-
[01:06:46] Geo: That’s
what we’re- It- It’s also
[01:06:47] joe: privilege too, right? So to disconnect from the world, you have to be privileged enough to not care about whoever the dean’s writing me about something, right? I gotta be able to put that aside, right? There’s some level of privilege [01:07:00] that you need that you can actually detach. So either, A you have the device and you’re in the world and you can now have some, you know, forward momentum in, in society.
And then at some point you can go, “You know what? I don’t need the device. I’m gonna put it away. I’m gonna go this week without
checking emails.”
[01:07:18] Geo: Well, I think- it gets back to what we could learn from the cell. Yeah.
You know what I mean?
[01:07:23] Mary: Yeah. The
[01:07:24] David: Allocate
your
[01:07:25] joe: the regular
[01:07:26] Mary: Well, it, I mean,
like-
[01:07:27] joe: you’re an
extremophile cell, and then
you’d use whatever garbage you get.
But
[01:07:32] Geo: …e-
[01:07:33] Mary: e- at like
exponential
growth. I mean, when d- where does it s- where does it stop?
yeah. Yeah.
I mean, I mean, it’s-
[01:07:40] joe: Yeah. Yeah …
[01:07:41] Mary: I don’t…
Yeah.
[01:07:42] Geo: just-
I mean, bizarro …I mean,
[01:07:44] joe: think you have…
Right?
I I also, I think people are like, “Oh, we shouldn’t regulate because that hurts innovation.”
And you know- … we look at, but we look at other… We look at the Industrial Revolution [01:08:00] and there was no
[01:08:01] Geo: regulation- We we know what happens when there’s- no regulations. I mean,
[01:08:04] joe: it was
a, it was pretty bad, right? You know-
You don’t- …
child labor, don’t need limbs.
[01:08:09] Geo: work.
Yeah, Okay. So
[01:08:11] joe: I, I think you could have had regulation and growth at the same time, and, you know,
you wouldn’t “
be, Oh, look at this disaster.”
Like it
would be a case study. Right. but So I think you would have that, um- Musk
[01:08:22] Mary: be a trillionaire. And why, well, …well- no one think of-
[01:08:25] Nick: trillionaire. Whoa, why are you trying to attack Elon Musk? Oh, well, we He’s a billionaire. We can’t attack
him. He’s a now.
[01:08:32] Geo: no. trillion.
There’s a effect, okay?
[01:08:35] joe: It’ll all But,
[01:08:36] Mary: yeah, it’s,
Are there any good uses of, for AI, of AI,
[01:08:41] Geo: of AI? Well, I think he was giving- that example- of using it for like- No, like for- There, Yeah.
[01:08:47] Mary: are there in-
[01:08:48] joe: the- …therapeutics.
I think, you know, medical diagnostics. I mean, I think there are. I mean, if you
are- Yeah
[01:08:55] David: Yes.
[01:08:55] joe: you have-
[01:08:56] David: the answer is yes but they [01:09:00] certainly- don’t directly impact most people
day to day in any tangible way.
[01:09:07] joe: Yep
[01:09:08] David: They’re abstract. They’re kind of enterprise productivity type things. It’s coding.
It’s helping people that don’t need help, and it’s not helping people that do
[01:09:22] joe: Yeah. Or super personal, like, some of the stuff where, , we were talking about devices, brain interface, quadriplegians being able to inter-
right. I
I think you are having… Yes I think you have that. But the majority of it is memes, you know,
[01:09:37] David: slop right. It’s just kind of,
[01:09:38] Mary: just,
[01:09:39] joe: trying to get AI to do things that, , let’s replace human creativity and it’s like, well, why?
Isn’t that the, you know, that’s the fun part of it all,
[01:09:47] Nick: like, I mean, didn’t you just listen to the first,
AI generate fake porn.
[01:09:51] joe: Right, yeah.
That’s right. Yeah.
[01:09:52] David: more porn.
[01:09:53] joe: We
[01:09:53] Mary: S- So
[01:09:54] Geo: I, so-
We can’t
[01:09:54] joe: pay the real
people to do it. Let’s
[01:09:56] Mary: So are we really just,
Blowing up the planet [01:10:00] for slop and porn?
[01:10:01] Nick: We’ve done it for less.
[01:10:02] joe: Okay.
Yeah. I think that-
[01:10:03] Geo: w- I think that about-
[01:10:05] Nick: done it for
[01:10:05] David: I think
[01:10:06] joe: right? Yeah, mean, it’s
[01:10:07] David: we were gonna
[01:10:08] Nick: that
[01:10:09] David: title This episode. blowing up the planet for slopping porn Is pretty much,
[01:10:14] Geo: yeah.
[01:10:15] Mary: Yeah, I thi-
[01:10:16] Nick: might be the top rated episode.
[01:10:17] Mary: like-
[01:10:18] joe: be. That’s right.
Maybe I need the slop
and porn. Maybe that
changes-
[01:10:21] Mary: I was gonna say,
Like, tell me I’m wrong, but, you know, yeah, it just, like, I guess it’s just all for slop.
Yeah. Yeah.
[01:10:28] David: Right now.
[01:10:29] Mary: now. Oh, wow.
[01:10:29] joe: Or 80%.
[01:10:32] Geo: 80 is
[01:10:32] Nick: it’s- a
percent-
[01:10:33] joe: don’t know.
[01:10:33] Nick: that’s used for- It’s I mean, why- … the home use … you know,
[01:10:36] Geo: the water is
[01:10:37] joe: right? It’s all, it’s consumer-based, right? So we live in a, I mean, consumer and the promise that the consumers will turn around and pay money for this, right?
Like a subscription service.
So I think …that’s the whole
[01:10:49] Mary: To get them hooked.
[01:10:50] joe: That’s right. So that’s why… And that’s, you know, there’s more people in non… Not everyone’s a researcher trying to solve a structure to [01:11:00] push their research for defined drug targets, or what new antibiotics can we find, right? I mean,
that, there, there’s,
That’s happening in the background, but the
stuff we hear about
and the stuff we see is
[01:11:09] Geo: the majority of it is- right. yeah …And that’s
[01:11:12] joe: problem
is that you’re feeding that beast, and then you’re saying, “Well, how can we have, how can we hook the consumer more? Oh, if we can give them more short form videos. Can we give them more, you know, what they want? If they, if you want this story with this person in it, can we give that to you right away?”
Like, because you can’t wait for it. And then it’s gonna happen slowly, but it’s gonna be
now
you have to pay for it.
And
I
think that’s the thing we don’t know yet is when that transition happens, how many people will go, “Okay, I’ll pay for it,” or how many people will be like Georgia and, “You know what?
I don’t need it that bad. I’m gonna find something else to do with my money or time and not be hooked into it.” So I think we have this, like, kind of freemium I think they call it. It’s where it’s free for a while-
[01:11:56] Mary: Ah, I
see.
[01:11:57] joe: you build up, like not a premium,
[01:11:59] Geo: and then premium-
[01:11:59] joe: a [01:12:00] freemium- There’s that- and you go,
and then it, and then so that’s like your search engines now.
How long will it be before Google stops digesting, AI
digesting-
[01:12:08] Geo: gotta figure out your
[01:12:09] joe: then you gotta figure… That’s right. Yeah. You gotta, You gotta, figure that out on your own, lady.
[01:12:15] Geo: gotta
[01:12:15] Nick: they’ll have
a cheaper option with
[01:12:17] Geo: with ads.
[01:12:17] joe: And you’re right, they probably will, right.
I mean, that’s- True
…they’ll have
something to keep you going, but that will be
the model.
I don’t know how long that’s gonna take to, to change, but I see the next probably couple years.
[01:12:28] Nick: We see the same thing with the streaming services. Yes, sir They started off- at such a low price- just to compete with cable. Yeah And now they’re killing cable- and also increasing the prices Yeah It’s that same model
where-
[01:12:41] joe: The
[01:12:41] Mary: the
[01:12:42] David: Same with Uber, Lyft,
[01:12:44] Mary: Hoopla.
[01:12:44] joe: And so you’re killing
blogs- say, hey, Check out Kanopy. It’s amazing.
[01:12:49] Mary: Yeah, your
[01:12:49] Nick: li- Hoopla …
[01:12:50] Mary: get your
library card, and your library probably has a subscription to Hoopla,
[01:12:55] Geo: can actually go- Yeah …
just like an old-fashioned
[01:12:58] Mary: Kanopy with a K.
[01:12:59] Geo: you [01:13:00] can actually go into the libraries like an old Blockbuster Video- Right …
and pick the videos Old DVDs It’s so cool.
Hell Yeah Yeah.
[01:13:08] Nick: I did that. Yeah you just … Some
of them have games, some of them have- Yeah … music and movies.
[01:13:14] Mary: we go check it and you own it. If you in a library or if you buy- Physical,
[01:13:18] Geo: it’s a physical media. Physical. Yeah.
[01:13:21] joe: it.
[01:13:21] Mary: Cloud, Mm-hmm
[01:13:22] joe: the
cloud
could go away one day,
[01:13:24] Mary: Go back to physical media, yeah
[01:13:25] joe: right?
[01:13:25] Nick: Go back to physical media,
[01:13:27] joe: and then you’re sad.
You’re lonely, and you got
nothing.
[01:13:32] Geo: Except each other.
[01:13:33] Mary: Literally.
You got nothing.
[01:13:35] joe: school porn, right?
[01:13:36] Mary: other.
Aw
[01:13:37] Nick: porn.
[01:13:38] Mary: Old
[01:13:39] Geo: porn.
I
[01:13:40] Mary: remember
that magazine.
Where
did I
find it? Where was it at?
[01:13:43] David: I got a bunch of Betamax.
[01:13:45] Nick: I remember when the cloud was- ChatGPT, where’s my
[01:13:48] Mary: old porn magazine?
[01:13:49] joe: was like The Matrix right? You know, I designed her. Like, you know, it’s like yeah.
[01:13:53] Geo: The cloud
[01:13:54] Nick: back in the day was just the f-
[01:13:56] Mary: cloud Real cloud
The
cloud
[01:13:57] joe: Real
[01:13:58] Geo: closet Really? Yeah.
A tree stump with
porn
stuck in it. in my day- I think, I- In my day we had clouds. I couldn’t think of the
[01:13:58] Mary: Like an actual
[01:13:59] Nick: A, a [01:14:00] tree
stump with porn
stuck in it.
[01:14:02] Mary: it.
[01:14:02] Nick: Yeah.
[01:14:02] Mary: Yeah.
[01:14:03] joe: Yeah.
[01:14:03] Mary: In my day.
[01:14:04] joe: In my day,
we had clouds.
[01:14:06] Nick: I couldn’t think of s- the word stump.
[01:14:08] Mary: I-
[01:14:09] Nick: It stumped me.
[01:14:10] Mary: If you, haven’t thought about the word stump- I think it’s time
for us to wrap it
[01:14:13] joe: it up. I’m right.
[01:14:14] Mary: Yeah, I think
[01:14:14] joe: I think it’s Probably
[01:14:15] Mary: time to bring it in. Bring it down home.
[01:14:17] joe: Any
f- last questions? Yeah.
[01:14:19] Geo: We just added to the slop, but-
[01:14:22] Nick: We just added to the slop no, I actually think we hit a lot of
good things this episode.
[01:14:25] Mary: I, well, I would also like to just let people know, anybody who’s listening to this right now, that we’re all actually just AI generated.
[01:14:33] Nick: Well,
[01:14:33] Geo: yeah.
[01:14:34] Mary: Yeah and none of this
[01:14:36] joe: Take me to, your leader.
[01:14:38] Mary: just to mess with you a little bit.
Yeah, none of th- you know, we just put this into a ChatGPT prompt.
[01:14:43] Geo: Are we real?
[01:14:44] joe: we
[01:14:44] Mary: a podcast for-
[01:14:46] joe: We did that an episode- Yeah … in this,
Yeah … season
three with Lily and Generoso. I
[01:14:50] Nick: if you go
[01:14:50] joe: Yeah … I
asked I asked AI to interviewed AI for a segment on that episode. So if you go back and listen to
that. Yeah. We did that.
Yep. Got their opinions. But, [01:15:00]
[01:15:00] Mary: we’re totally real.
[01:15:01] joe: We are real.
We’re real
[01:15:02] Geo: Sure I’m gonna
pinch,
[01:15:04] David: how you know we’re
real
[01:15:05] Mary: because we c- S-
[01:15:06] joe: Yeah.
[01:15:07] Nick: can
[01:15:08] Geo: Yeah.
[01:15:09] joe: You guys didn’t…
[01:15:09] Mary: Sorry, David gave us
the
thumbs-
up- ……and a little wink and
a
nod. So
[01:15:13] Geo: we
[01:15:13] joe: the nice thing about
is to rabbit hole research, we go out and about, and we interact with our people.
So we are real. We are here.
[01:15:22] Nick: We interact with
[01:15:23] David: out of the burrow
[01:15:24] joe: interact with people. Yeah. We do.
[01:15:26] Nick: I kind of just hide away like a little
rat.
[01:15:28] Mary: blink, blinking
[01:15:29] joe: Let’s give, uh… David,
you got any final words on this climate change data center? Are we all screwed?
[01:15:36] Nick: you leave us on a high note?
[01:15:37] joe: will we a-
[01:15:38] David: Yeah,
no.
[01:15:41] Nick: Say everything’s gonna be okay.
[01:15:43] David: no, I do think that there is a really good chance that, you know, there is some good that comes from all of this
[01:15:51] Mary: Mm-hmm.
[01:15:53] David: industrial scale escalation. Because at the, at least it’s about [01:16:00] intelligence, at least they say it is, and it’s about, like, understanding and getting to the bottom of scientific questions.
And if we can do that a bit with this, that’s gonna be good, and that’s gonna then help us solve a lot of these other problems. If we don’t do that, and if we just slop ourselves into tedium, then we’re doomed. But I do think people are really excited about solving hard problems, and we don’t wanna just look at screens.
We wanna get out there. We wanna do things. We wanna act in the real world and make change. So I’m hopeful and optimistic even though I think right now there’s a lot of Yeah.
Cloudy clouds
over the cloud.
[01:16:52] Mary: Yeah A lot
of wild west- mentality happening
right now.
[01:16:56] joe: Yeah
[01:16:56] David: I do think the science is underneath all of this, and if we [01:17:00] can let it happen, that’s gonna help humanity
[01:17:04] joe: I like that Hell
[01:17:05] Geo: Yeah.
[01:17:06] joe: are you doing?
[01:17:07] Geo: thank
[01:17:07] Nick: Hell yeah. Yeah. Well- Yeah …thank you so much, David, for
being here with us today. Hopefully next season- you’re here-
in
[01:17:12] David: charm
[01:17:13] Mary: All
[01:17:13] joe: in studio. Yay. We have him on the f- the f- the fourth time, and we’ll see if you get there before whoever was at three before.
[01:17:20] Geo: think he
[01:17:20] Nick: he wants to be on episode one of
[01:17:22] joe: That’s right. Yeah,
[01:17:23] Nick: He’s like, “I’m gonna be the,
[01:17:24] David: exactly.
[01:17:26] joe: He’s always
at the end of the
season. We
always like,
[01:17:29] David: on a bonus episode
[01:17:30] joe: Awesome.
[01:17:32] Geo: Yeah.
Yeah,
[01:17:32] joe: There’s …
You gotta come in studio for that, so that’s,
[01:17:34] David: I’ll make my own jacket that
says four-timer
[01:17:38] Nick: We’ll have
[01:17:39] Geo: We’ll
have AI. we’ll have
[01:17:40] joe: a, we’ll
[01:17:40] David: little stars like I won the World Cup,
[01:17:42] joe: AI,
we’ll have- I was like our number one fan …we’ll have AI
make you something, so, we’ll see.
[01:17:46] Nick: we’ll see.
[01:17:46] Mary: see.
[01:17:47] joe: Well, yeah, on that note,
you have me, Joe.
[01:17:49] Nick: have me, Joe. You
got
[01:17:50] joe: You got Nick.
[01:17:50] Geo: Georgia.
[01:17:51] joe: We’ve got
Nick.
Georgia.
[01:17:52] Mary: got Mary.
[01:17:53] joe: We’ve got Georgia. You got Mary.
[01:17:55] Nick: And-
we went down some
hole. See you.
[01:17:59] joe: Love
[01:18:00] y’all.
