Mystery AI Hype Theater 3000

Data Centers in Space!? (with Dr. Adam Becker), 2026.03.02

• Emily M. Bender and Alex Hanna • Episode 73

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 53:38

Data centers are overburdening the planet, so tech billionaires have a new scheme — put 'em in space! Astrophysicist Dr. Adam Becker joins Alex and Emily to launch this plan into the sun. We unpack all the reasons that this hilariously terrible idea will never be viable outside of sci-fi-villain fantasies.

Dr. Adam Becker is the author of More Everything Forever, a book about the terrible plans that tech billionaires have for the future and why they don’t work. Keep an eye out for his new podcast, Dreaming Against the Machine, which is launching soon.

References:

Also referenced:

Fresh AI Hell:

Check out future streams on Twitch. Meanwhile, send us any AI Hell you see.

Our merch store is now live on the DAIR website!

Find our book, The AI Con, here.

Subscribe to our newsletter via Buttondown.

Follow us!

Emily

Alex

Music by Toby Menon.
Artwork by Naomi Pleasure-Park. 
Production by Ozzy Llinas Goodman.

Alex Hanna: Welcome everyone, to Mystery AI Hype Theater 3000, where we seek catharsis in this age of AI hype. We find the worst of it and pop it with the sharpest needles we can find. 

Emily M. Bender: Along the way, we learn to always read the footnotes, and each time we think we've reached peak AI hype, the summit of Bullshit Mountain, we discover there's worse to come. I'm Emily M. Bender, professor of Linguistics at the University of Washington. 

Alex Hanna: And I'm Alex Hanna, director of research for the Distributed AI Research Institute. This is episode 73, which we're recording on March 2nd, 2026. If you're watching us live, then you might notice our coordinating outfits. 

Emily M. Bender: We are both modeling some T-shirts from the new designs in our merch store. 

Alex Hanna: Merch, merch, merch!

Emily M. Bender: And if you'd like to join me in drinking your tech billionaire tears out of a "mathy maths" mug, you can get your own at store.dair-institute.org. 

Alex Hanna: Yes, help us spread ridicule as praxis far and wide! So, onto the episode today. Our guest this week is Dr. Adam Becker, a journalist, author, and astrophysicist. 

Emily M. Bender: He's the author of More Everything Forever, a book about the terrible plans that tech billionaires have for the future and why they don't work. He's also launching a podcast called Dreaming Against the Machine, so keep an eye out for that soon. Adam, welcome to the show!

Adam Becker: Thanks for having me. It's a pleasure to be here. 

Alex Hanna: We're so excited to have you with us to talk about data centers in space... space... space. This is the latest move by AI boosters to try to escape the real world impacts of their sci-fi villain behavior. 

Emily M. Bender: In the past few months, tech billionaires like Jeff Bezos and Alphabet CEO Sundar Pichai have gotten behind this idea. They've said that data centers in space could be economically viable within the next decade. 

Alex Hanna: Not to be outdone, Elon Musk has recently started claiming that the timeline is more like two to three years, and last- yeah, we'll get into it. Adam's so responsive to our intro script, so it's great. And last month, his company SpaceX filed for FCC approval of, quote, "a constellation of a million satellites that operate as orbital data centers."

Emily M. Bender: Of course, Elon Musk makes up these deadlines for new technology all the time, and he almost never actually meets them, but it's still notable that he's choosing this moment to jump on the space AI bandwagon. 

Alex Hanna: We're seeing so much local activism against data centers and their horrific environmental impacts. So it's no surprise that AI boosters have realized they need a new strategy. 

Emily M. Bender: Yeah, so we've got some artifacts today to get into the details of this with our extremely well-informed guest. This is gonna be so much fun. Starting with a SpaceX press release, I guess, with really horrible margins in this web browser. February 2nd, 2026. "xAI joins SpaceX to accelerate humanity's future." So before we start reading this artifact, Adam, is there anything you've been burning to say already? 

Adam Becker: Yeah, we're not gonna put AI data centers in space, not in any meaningful way, for lots and lots of reasons. Physical, financial, social. Like, it's a terrible idea. And you're completely right. Musk has given all sorts of deadlines for all sorts of technology and has just busted right through almost all of them. He basically never delivers.

Emily M. Bender: But along the way he is gonna do a bunch of damage.

Adam Becker: Yeah, the fake deadlines are used to justify the damage that he does along the way. 

Emily M. Bender: So yeah, this press release says, "SpaceX has acquired xAI to form the most ambitious, vertically integrated innovation engine on and off Earth with AI, rockets, space-based internet, direct to mobile device communications, and the world's foremost real-time information and free speech platform. This marks not just the next chapter, but the next book in SpaceX and xAI's mission, scaling to make a sentient sun to understand the universe and extend the light of consciousness to the stars." I can't get over how much bullshit is in that one paragraph. 

Adam Becker: What is a sentient sun?!

Alex Hanna: These are all certainly words. Well actually, some of them are debatable, but it's really just, wow. Let's get into the next one, because I think you'll have quite a lot to say about this, Adam. So, "Current advances in AI are dependent on large terrestrial data centers, which require immense amounts of power and cooling. Global electricity demand for AI simply cannot be met with terrestrial solutions, even in the near term, without imposing hardship on communities and the environment." So, your response to that, Adam? I mean, for our listeners, your face is saying quite a lot.

Adam Becker: Yeah. I'm sorry, but if global electricity demand for AI can't be met with terrestrial solutions, then that means that we're not gonna meet it, you know? Like they're not, I'm sorry, but putting AI data centers in space is a terrible idea for so many reasons. I'm gonna start with my favorite one, which is of course, you know, physics. You know why a thermos keeps hot things hot and cold things cold?

Emily M. Bender: Yeah, there's an air gap, right? 

Adam Becker: Yeah. There's an air gap. And a really good thermos, like if you get the best industrial insulating thermos out there, it's not even an air gap, it's a vacuum gap. Because vacuum is a really, really good insulator. It's the best insulator we know. Now space... 

Alex Hanna: Famously! 

Adam Becker: Space is a vacuum. 

Emily M. Bender: Famously, yes. 

Adam Becker: Yeah. And so sending a giant data center into space is like wrapping it in lots and lots of really warm, thick blankets, except even more so. You're just putting it into this giant insulator. And so it's very, very hard to take all of the heat that a data center would generate and bleed it off. Here on Earth, the way that we do that with data centers and other warm things is we put them into contact with cold things. But they're cold, dense things that we put them into direct physical contact with, water cooling pipes and stuff like that, or we use fans to blow cold air over them. There's no air in space. There's no water to pipe in, in space. Even though space is cold, it doesn't matter, because there's nothing there. The only way to bleed off heat is through radiation, and so you need to build these huge radiator veins. That would be way, way, way too big to deploy in space for any data center of meaningful size. And so that alone is just a showstopper. And then you've got all the other problems. You know, you wanna upgrade the hardware in your data center? Congratulations, you gotta do a space walk. That's not gonna go too well. You want to have your data center in low Earth orbit, but you want to have solar power all of the time? Well, most low Earth orbits take 90 minutes to go around the Earth, and that means 45 out of those 90 minutes, there's no sunlight, because they're behind the Earth from the perspective of the Sun. 

Emily M. Bender: Wait, wait wait. I know what to do! Just blow up the Earth. 

Adam Becker: Yeah, no, I mean, it's just a really bad idea for so many reasons. I mean, and also, just getting that much stuff into orbit is incredibly expensive. Even with, like, one of the few real genuine accomplishments of any Elon Musk company, SpaceX has lowered the cost to put stuff into low Earth orbit. That's a real thing that they've actually done, credit where credit's due, but it's still way too expensive. And so, it's not gonna happen. You know, like, why would it? And then you've got other issues, right? You've got- ah, I mean, there's this thing called Kessler syndrome. Have either of you heard of it? 

Emily M. Bender: No. 

Adam Becker: Okay. So Kessler syndrome is one of these things that sounds scary, and it is scary. The idea is, we've got a bunch of stuff in low Earth orbit already. And to be in low Earth orbit, you have to go really, really fast. And so that means that if something happens, and a satellite loses power or breaks up or whatever, you end up with really fast-moving pieces of space junk. And then if they hit each other, you get even more space junk, and you end up with a chain reaction of space junk if you have too much stuff in low Earth orbit. Now, we don't know exactly how much stuff would be required for Kessler syndrome to really take off, but we know that we may be getting close. The best estimates we have are that we're already pushing up against those limits. We're already seeing high-speed space junk in orbit impacting other stuff in orbit. And if you wanna know what that would look like in a broader sense, go watch that movie Gravity. That was essentially what that was depicting. And yes, that movie had its own technical problems, but the actual disaster that kicked off the movie, that was pretty realistic in a way. 

Alex Hanna: Yeah, I think when I think of a lot of junk in low Earth orbit, I think a lot about Cowboy Bebop, when the Moon explodes. 

Adam Becker: Yeah! 

Alex Hanna: And then people kind of have to leave Earth because so much stuff is just coming down and creating huge craters. 

Adam Becker: Yeah. I don't think it's going to be quite like Cowboy Bebop, for two main reasons. First, this stuff is not gonna be big enough to cause serious problems with its impact, like physical impacts, here on Earth. I mean, it might cause some problems, you know, it could be that someone gets killed by a piece of falling space junk. But you wouldn't have massive impact craters from that stuff. It's just not big enough. And most of it would burn up in the atmosphere. But it could cause serious problems for telecommunications. It could make it difficult or impossible to send anything into space, including things that we need, like weather satellites and GPS. And we're already seeing this with the Starlink constellation. It's changing the way the night sky looks. Even if you go into the darkest places, you start seeing this sort of crawling in the corner of your eyes from these satellites, and it's just gonna get worse if we let them keep putting more and more stuff up there. But yeah, so the good news is we're not gonna get those giant impact creators from Cowboy Bebop. The bad news is, Cowboy Bebop, they found ways to live off of Earth. That's also not happening. We're here. This is is where we're living. 

Emily M. Bender: Yeah, we're here. And the problem we might be trying to solve is, how do we live without weather satellites and GPS, if we put too much junk up there and then we can't do it anymore?

Adam Becker: Yes. I thought that you were gonna say the problem that we have to solve is, how do we figure out how to live here on the same planet as the billionaires? 

Emily M. Bender: There's that too. Which is a good segue- there's some stuff that's in the chat that we should be lifting up. So one that's now scrolled off and I can't see it, but someone was saying, if the text string is long enough, can we use it as a space elevator? And someone said, if it's strongly worded. That was great. So wisewomanforreal says, "Can't we just send Musky to the Moon?" And abstract_tesseract replies, musically, although I can't guess the tune. Do you wanna guess the tune there, Alex? 

Alex Hanna: I'm assuming it's the Sinatra song. "Fly Musk to the Moon and let him reinvent the bus. Let him live there so he'll be the hell away from us." In other words, go to Mars. In other words, please miss me. So that's a transition. You can't ask for a musical transition today, 'cause I already did it. 

Emily M. Bender: And then when we were chatting before about how vacuum is insulating, magidin points out, "AI bros think, 'Space is very cold, so it will cool my data center,' which just shows how ignorant they are." 

Adam Becker: Yeah. I think that's exactly right. Also, on Musk going to the Moon, he has said- and I think that this is a line that somebody wrote for him, because it's too good of a line for him to have come up with it himself. But he has said that it would be cool to die on Mars, just not on impact. And my response to that has always been, you know, don't be so picky. 

Alex Hanna: Yeah. Beggars can't be choosers, honestly. 

Adam Becker: Yeah, exactly. Yeah.

Emily M. Bender: All right, so should we keep going with this artifact a little bit?

Alex Hanna: Yeah, it's pretty bad. Do you wanna start, Emily? 

Emily M. Bender: Yeah. So, "In the long term, space-based AI is obviously the only way to scale. To harness even a millionth of our Sun's energy would require over a million times more energy than our civilization currently uses." Can you make any sense of that? 

Adam Becker: I would have to run some numbers, but my guess is that they're saying that the current energy usage of our civilization is over a million million times smaller than the current energy output of the sun. But it's worded very badly. But this does actually get to a thing that I'm seeing a little farther down, where they say, "Launching a constellation of a million satellites that operate as orbital data centers is a first step toward becoming a Kardashev two level civilization. One that can harness the sun's full power." 

Emily M. Bender: What does that even mean? 

Adam Becker: Yeah. So this gets to a thing called the Kardashev scale, which was invented by a Soviet speculative futurist, and I think physicist, named Kardashev, who said you can classify a civilization by how much energy it uses. So Kardashev level one means it's using all of the energy that its planet receives from its star, and level two is it's using all of the energy of its star. Level three is all of the energy of its galaxy, and level four is all of the energy in the observable universe. I don't know if Kardashev meant this as, like, a to-do list. Whether or not he did, it ain't happening, and using it as a to-do list is pretty preposterous. Like, why is that something that we should try to do? Why do we want to use all of the energy of the Sun? There's no good reason to do that. It's just preposterous. 

Emily M. Bender: No, it's like, as opposed to, why don't we just capture as much energy as we need from the Sun without polluting our own atmosphere?

Adam Becker: Yeah. It's this idea of using more energy as a good unto itself, which I just really do not understand, other than like a desire to always have more and not feeling like anything is ever enough. 

Alex Hanna: Yeah. It's the Dyson sphere. It's the, we need to build something and somehow capture all of this. It's very much leading to this kind of futurism where it's, we need to capture as much energy as possible and then we can do new and exciting things like faster than light travel, you know, whatever. Which is I think the types of things that are also gonna break physics. But that's the dream. Every part of it is just so infuriating, and the copy is so poorly written. I'm sure a lot of it's extruded and very lightly edited by copy editors that no longer have jobs, but, you know, it's really horrific to read, just in terms of what kind of future vision they're trying to base this fantasy around. 

Emily M. Bender: Yeah, it's a mess. But should we read a bit more of it anyway? Do you wanna read a bit more, Alex, and then we can do commentary? 

Alex Hanna: Yeah. So, let's see. So there's this image that is very phallic. You know, there's a little back and forth between Musk and Bezos on how phallic their rockets are. But anyways, after that there's two paragraphs. So one is: "In the history of space flight, there's never been a vehicle capable of launching the megatons of mass that space-based data centers or permanent bases on the Moon and cities on Mars require." 

Emily M. Bender: Cities on Mars! 

Alex Hanna: Yeah. "Even in 2025, the most prolific year in history in terms of the number of orbital launches, only about 3000 tons of payload was launched into orbit, primarily consisting of Starlink satellites carried by our Falcon rocket. The requirement to launch thousands of satellites to orbit became a forcing function for the Falcon program-" very silly- "driving recursive improvements to reach the unprecedented flight rates necessary to make space-based internet a reality. This year, Starship will begin delivering the much more powerful V3 Starlink satellites to orbit with each launch, adding more than 20 times the capacity to the constellation as the current Falcon launches of the V2 Starlink satellites. Starship will also launch a next generation of direct to mobile satellites, which will deliver full cellular coverage everywhere on Earth." Thoughts on that? 

Adam Becker: I mean, okay, so that first paragraph, the one that starts "in the history of space flight," those two sentences- technically true. It is true that there's never been a vehicle capable of launching what would be needed to put permanent bases on the Moon and cities on Mars. You know, we're not gonna put cities on Mars. That's not happening! Because Mars is completely terrible, something that Musk has just steadfastly refused to ever acknowledge in the face of overwhelming expert opinion. But he's never let reality stop him from indulging in his insane fantasies. But yeah, the other thing is, Starship would need to work for all of those things to happen. Which is not to say that Starship will never work, but it hasn't worked yet. You know, that phallic image is an image of the top stage- or the top two stages, I don't remember- of Starship. And they haven't really gotten Starship to work reliably. They might at some point get some version of Starship to work reliably. They keep talking about V2 and V3 and V4 Starship, which I think are probably different from the V2 and V3 Starlink here in the paragraph. But the specs that they've given for those later versions of Starship- which are supposed to take things to Mars, not just low Earth orbit- are pretty preposterous. I'm not saying that they're never gonna get a version of Starship to launch a version of Starlink satellites, that could happen. But anything that SpaceX says involving Starship is automatically at least a little bit suspect, because they haven't gotten it to really work reliably yet. The Falcon rocket works, which is why I said those first two sentences, technically true. But I'm so struck by how thin, or in certain cases non-existent, the justification for wanting to do any of these things is.

Alex Hanna: Yeah. I mean, the justification is that it comes a little earlier, in terms of the kind of compute. And to me it signals very much the underlying, we need more AI compute to do more "science," and then that will self-improve. And then, there's some trolling in the chat about the word recursive. And so there's a few plays on the Xzibit meme of, like, "Yo dog, I heard you like blah." And so sjaylett said, "Yo dog," abstract_tesseract said, "I heard you like hype." And I said, "We put a ship in your ship so you can ship even shippier." So very much the idea of this kind of improvement, and improving just very intensely. That seems to be the end goal here. 

Emily M. Bender: And we have to pick up abstract_tesseract's reply to that: "When I said I was curious about shipping in their AI bro sci-fi fan fiction, this isn't what I meant."

Adam Becker: Oh, god. That's perfect. 

Alex Hanna: I don't want to know what ships that they have, though. I mean, it's going to be Musk- 

Emily M. Bender: Nightmare fuel. 

Alex Hanna: Yeah. And his synthetic girlfriend, or something. 

Emily M. Bender: Yeah. All right, so let's just do a little bit more of this one, and then we have a slightly more serious, but equally ridiculous, second artifact to get to. So, Musk says a little bit further down, "The basic math is that launching a million tons per year of satellites generating 100 kilowatts of compute power per ton would add 100 gigawatts of AI compute capacity annually with no ongoing operational or maintenance needs." 

Adam Becker: None!?! I'm sorry, none? None?! What? None! No ongoing operational or maintenance needs. Wow, that's amazing!

Alex Hanna: No, there's nothing. You don't have to fly, you don't have to do a spacewalk or get up there, or, you know, take anything outta operation. It's just...

Emily M. Bender: I mean, space is clean, right? There's no dust, no bugs, just a whole hell of a lot of radiation.

Adam Becker: Yeah, exactly. That's exactly what I was about to say. Because another major physical reason why this is a bad idea is, we get so much radiation protection here on the surface of the Earth. And when you're in low Earth orbit, you're above the entire atmosphere, basically. And that means you're losing about half of your radiation protection. You know, computers don't do well in high radiation environments. Radiation has a tendency to flip bits and stuff like that. It's really bad. So, also, I know we skipped over this paragraph, but in the previous paragraph, he says, "with launches every hour." Launches every hour?! Are you kidding me? Absolutely not. Also, the reliability rate- when I said that Falcon was okay, the reliability rate is somewhere around 99 percent. It's not 99.9, and it's not 90, it's somewhere around 99. So if you launch every hour, that means that you're gonna have, what, about a hundred launches in about four days? So you've got a rocket blowing up, you know, a failed launch, about once every four days. That's gonna be awesome and clearly work really well for everyone involved. 

Emily M. Bender: And I'm coming back around to, the original motivation at the top of this thing was there's too much environmental impact to try to build these data centers on Earth. So let's just have hourly rocket launches? 

Adam Becker: Yeah. And those rocket launches cause problems. There was a study recently, they caused problems for the ozone layer. One of the few major international success stories, in terms of environmental agreements. And this would just, you know, literally blow it up. Yeah. And also, you're not gonna launch every hour. It's not happening. 

Emily M. Bender: Yeah. All right, I wanna get to the next one and still have time for Fresh AI Hell, but I have to read the signoff here. It says, "Thank you for everything you have done and will do for the light cone of consciousness. Ad astra, Elon." Thoughts? 

Adam Becker: So my main thought, the first thing that comes to mind is like, I am very lucky that I just constitutionally have pretty low blood pressure, because otherwise I'd be having a stroke right now. Ahhh! "An entire civilization on Mars"? So first of all, I'm gonna just plug my book, not because I'm plugging my book, because it really is relevant for this. 

Emily M. Bender: For sure. 

Adam Becker: This is all the stuff I go after in my book. I actually didn't mention the Kardashev scale in my book because there's only so much room to talk about so much bullshit. But it's, mm-hmm. It's right. it's exactly the kind of thing I was going after. But, you know, a civilization on Mars and also "self-growing bases on the Moon"? No, no, I'm sorry. But also, "the light cone of consciousness." A light cone is a concept from relativistic physics. It is the entire set of events in spacetime that can be affected by a particular event at a particular time and place, limited by the speed of light. So, you can think of a cone going out from a particular event at the speed of light, forward in time, reaching out to the entire universe, at the speed of light. Basically, this is Elon throwing in a needless piece of physics jargon to talk about the future of consciousness. Something that he also had, I think, at the very beginning of this piece of crap as well. When he talked about "extending the light of consciousness," which is a phrase he likes to use all the time. Yeah, there it is. "A sentient sun to understand the universe and extend the light of consciousness to the stars." I have no idea what a sentient sun is. I think he means a giant AI data center. So that's already making some claims about sentience and AI. Also, why are you calling that a sun? Also, what do you mean "understand the universe"? That's a pretty vague phrase. And "extend the light of consciousness," that's a phrase with a pretty interesting history going back to some Christian mystics in the early 20th century. This is mostly Musk saying, it is my job to be the personal Jesus Christ savior of all of humanity and human civilization, to get us off of this planet and bring us into the heavens. And I'm like, okay. None of that's real, and you're delusional! And also, you clearly don't know a damn thing about physics, or really much of anything else.

Alex Hanna: And I think we've come across "light cone" before, because it is the TESCREAL, it's the rationalist stuff. I was searching it just to see where it was, and Lightcone Infrastructure.

Adam Becker: Yes, indeed. 

Alex Hanna: And they have other projects, you know, Less Wrong. Something called Lighthaven.

Adam Becker: Oh, I know what Lighthaven is. Lighthaven is a compound run by the Rationalists in Berkeley. It's where they host their little eugenics conferences. 

Alex Hanna: Oh man. Yeah, they have a festival called Less Online.

Adam Becker: Yes indeed. Yeah, they do. And they also host Manifest there, which is a thing about a prediction market that also notoriously hosted several out and out white supremacists. So these people are totally awesome, and like, taking real physics terms and using them for their fake pseudoscience racist bullshit. But that's kind of their whole thing. 

Emily M. Bender: So let's see what the folks at the World Economic Forum have to say. 

Adam Becker: Oh, no. No! I'm sorry, that headline! Are you serious? Sustainably?! 

Alex Hanna: Yes. Unfortunately. Yeah, let's read it so everyone can know what you are being horrified at. And also this is a just a plug for, you know, this podcast has a video component. We're doing more video stuff now. You gotta look at Adam's face. It's great, and there's already been calls for GIFs in the chat. So, you know, we might spin it up. So this is the World Economic Forum. This is from January 16, 2026. "How data centers in space sustainably enable the AI revolution." And this is by someone named Philip Johnston, co-founder and chief executive officer of something called Starcloud. 

Adam Becker: Oh, great. So he doesn't have any conflict of interest whatsoever.

Alex Hanna: No, not at all. It's giving sponsored content. 

Emily M. Bender: Absolutely. And we're gonna skip these bullets 'cause they're probably AI generated, right?

Alex Hanna: That's true, yeah. So the first graph is, "Artificial intelligence, AI, is redefining every sector of our global economy, but this transformative technology comes with an unprecedented and growing appetite for energy and infrastructure. As a world races towards artificial general intelligence, AGI, we are confronting a stark reality. Earth's current energy and land capacity-" Land capacity? Okay, whatever- "are struggling to keep pace sustainably. This urgent challenge requires an out of the box solution or more accurately an off planet one. The emerging frontier of space-based data centers is no longer a concept confined to science fiction. It is a vital, technologically feasible solution poised to unlock the next wave of AI progress while drastically minimizing our planetary footprint." I'm gonna stop there. Adam's gonna die.

Adam Becker: Oh, no!!! Are you fucking serious? These people are fucking insane. This guy's "AI data centers in space aren't just science fiction" shirt is raising a lot of questions already answered by his shirt. And he's like, oh yeah yeah yeah. It's not just science fiction, which is totally not where we got the idea, guys. And this is real. It's technically feasible. How do I know? I saw it in some science fiction, or somebody told it to me, who also- 

Emily M. Bender: On the SpaceX blog! It was there. 

Adam Becker: It was also on the Nvidia website! My favorite, actually, the Nvidia website had a little article about how they were putting a data center in space. It was a satellite with one or two Nvidia chips in it. And the article actually said, "Space is a great place for data centers because it's cold." 

Emily M. Bender: Oh, no! 

Adam Becker: Yeah. And I'm like, okay, awesome. That's great. You're just wrong. I mean, yes, space is cold, but that's not gonna help you any. Oh my god! Just stating it's technically feasible, no actual evidence, just saying it. 

Emily M. Bender: Yeah. These things have such a strong, they're almost getting it. Like, hey, this is not environmentally sustainable. Okay! Let's stop. 

Adam Becker: No, no, because they need more everything forever. That's what they want. They cannot imagine having to stop. It's like, oh, okay, it's bad for Earth. We probably need Earth. So if we can't do it, if there's literally no way on Earth to do it, that means we have to do it somewhere else. It's like, no! That's not what anyone said.

Alex Hanna: Yeah. I'm skimming this, I mean, this looks very LLM generated, so I don't quite wanna get into it. But I do want to get to these links here. 'Cause they get into the technical components, or the- I'm gonna put "technical" in heavy air quotes here. And I would love your reaction to them, because they're the kinds of things that sound impressive if you are just making shit up. So this is the, "Why is space an optimum environment for building data centers? Space offers an environment naturally suited for high efficiency sustainable computing, providing solutions to the very challenges that limit terrestrial development. One: abundant, uninterrupted power. In a dawn-dusk sun synchronous orbit, SSO-" doesn't that stand for single sign on? Anyways- "a data center can leverage continuous high intensity solar power unhindered by nighttime, weather, or atmospheric attenuation. According to a Starcloud white paper-" and I clicked this white paper and it looked like absolute bullshit- "this continuous illumination is crucial, as it allows orbital solar rates to achieve a capacity factor of greater than 95%, compared to a median of just 24% for terrestrial solar farms in the United States." Which, okay, just limiting it to the United States, also an interesting decision, but whatever. "Furthermore, peak power generation in space is roughly 40% higher due to the absence of atmospheric losses." And then they get into the free radiative cooling. "The deep vacuum of space serves as a gigantic cold heatsink, with an effective ambient temperature of around negative 270 Celsius." 

Adam Becker: Okay. So I gotta jump in here. So first of all, that orbit that they were talking about, I'm pretty sure that's a real thing, but it's a limited resource. There's only so much space in orbits like that. And of course they're talking about putting giant solar arrays there. You're gonna use up that space really quickly if it's not already. Second, that equivalent energy cost, I will bet everything I own- minus my cat, maybe- that that energy cost estimate does not include launch or maintenance or deorbiting. And that's probably gonna wipe everything out. I also think that they haven't factored in undesired radiation impacting how well this thing can work, and also little pieces of space junk pitting out these solar cells and whatnot. As for the radiative cooling thing, I mean, I already talked about that. It is true that the ambient temperature of deep space is around negative 270 degrees Celsius, it's just a couple degrees above absolute zero. But again, that's no help. "Simple passive radiative cooling to achieve low coolant temperatures." Yeah. Radiative cooling is slow. It is going to take them quite some time and difficulty to get down there. You're gonna need huge radiator veins. I saw, I wish I remember where, but I saw someone try to actually work this out and say, okay, yeah, we're gonna do it. We're gonna build a data center in space. This is what would be required. And for a modest data center, we're talking radiator veins that were something like a square kilometer in size.

Alex Hanna: Oh my gosh, wow. 

Adam Becker: I'd have to double check that calculation. I haven't done that calculation myself. But it seems quite plausible. And they were saying that they could do it, and I was like, I was with you up until the part where you said you could do it. That's not gonna happen. Like, to put that in perspective, the biggest single structure ever put into space is the International Space Station, which is roughly the size of a football field. And that took many, many launches to get up there. "Space provides a boundless domain for scaling compute." Yeah, okay, fine. Space is infinite, but low Earth orbit is not.

Alex Hanna: Yes. Emily, this is gonna take us slightly afield of this artifact, just because I think this is quite LLM generated. So if you go up, Emily, to the Starcloud study, the actual white paper, which I'm kind of skimming. There's a table in here about the cost, which is quite bad. So go down to table one, which is the cost comparison. So there is the cost, and it is the cost of a terrestrial data center across 10 years, of a 40 megawatt cluster, from terrestrial- Adam is just like, mouth agape here. So terrestrial and then space. The row is "Energy, launch, cooling, water usage, enclosure, backup power supply, all other data center hardware, and radiation shielding." And the cost comparison is the cost balance of a terrestrial data center is 167 million. And then the cost of a space data center is 8.2 million? No way! 

Emily M. Bender: Wait. Do they have the launches in there? 

Alex Hanna: The launch is in there, and it is $5 million only, which seems really low to me. 

Emily M. Bender: And they're gonna be doing one launch.

Alex Hanna: And they factor the cost of the solar array, no cost of maintenance in either of them. And then no backup? I mean, okay.

Adam Becker: Yeah, I think that their thinking there is, well, you know, the Sun never goes out. And it's like, okay, yeah, but your solar cells might fail. And also, of course, they've got this bullshit about more efficient cooling architecture. I don't think that they'd be able to do this in a single launch. I think that their estimate there is just wrong. And my guess, I think that their shielding figures are probably wrong. I'd have to run the numbers. I also would bet that their cost for the solar array there is just them taking the cost of a comparably sized terrestrial solar array and plugging into space. But again, I don't know for sure. But the thing is, there's just all sorts of stuff that they haven't noticed or accounted for here, like maintenance, like the fact that their launch estimate's way too low. Like the fact that they clearly just don't understand the problems with cooling a data center or anything in space. You know, especially because they're gonna be in constant sunlight, right? That's what they want. So that means that they're gonna have to have some sort of structure to keep their radiator veins in shade. Otherwise they're gonna be in real trouble. And you know, again, probably not theoretically impossible, but probably real world impossible, and definitely way more expensive than they think it is. But honestly, just that one cell of this one table where they say "More efficient cooling architecture, taking advantage of higher delta T in space," that I think disqualifies these people as serious, like, they're just not serious people who actually understand what it is that they're talking about or dealing with.

Emily M. Bender: I think you've maybe zeroed in on the worst one of those, but I also feel like we could throw a dart at this white paper and they would disqualify themselves. The other thing that's jumping out to me here, and in the Elon Musk one, is that they're talking about compute as quantified in terms of kilowatts and terawatts and gigawatts and so on. That's a weird conflation to me. 

Alex Hanna: That is weird. Yeah. I mean I guess that's, in terms of talking about the kind of data center, I think they're thinking, I mean, maybe in data center terms, 'cause I think that's the comparable that they have, but they don't talk about data center computing capacity. Just because training operations, I'm assuming that the hardware could be different, and it's more about how much it can be powered. Anyways, I mean it's all very much a dream and it's very enmeshed in so much of the- like, the first paragraph then is like, with Altman, Musk, and Zuckerberg quotes, and then, Todd Mueller, employee number one at SpaceX. And then the CTO at Thales Alenia Space. I didn't know Thales did space stuff, but whatever. It's the people that they're taking as credulous. And it's very, I mean, the whole thing is terrible. And the thing about this that is also weird is that this was written in 2024, and then they're, in this document, they're talking about like GPT-6 and Llama 5. So just like, models that don't exist. And then they have a diagram on page six of the network architecture, which to me looks like a big, like a shipping container with a big, viewing port on it, that's very Death Star laser shaped. And then they've got the server racks, and then there's "Liquid cooling, networks, docking port." 

Adam Becker: Wait, whoa whoa whoa, wait. Hold on. I thought this was supposed to be in space. What do you mean, liquid cooling? 

Alex Hanna: Actually, yeah, that's right. They actually say in the table, "Water usage not required." I am confusion. 

Adam Becker: I'm very confused! Also, can we go back to the very, very top? I have a question. What do we think Ezra Feilden's PhD is in?

Alex Hanna: Oh, I'm gonna search his name. 

Emily M. Bender: I'm gonna go philosophy. 

Alex Hanna: I'm gonna take, hold on. I'm trying to do this in a way that's not insulting to anyone on the call. So I'm gonna say, probably something where it's like, computer science.

Adam Becker: That was my guess. My guess was EECS. 

Alex Hanna: That was your guess, EECS? I was also thinking electrical engineering. Okay. If we go to, the first search is, he's the CTO at Starcloud, of the XPRIZE. So, okay. Let's see... materials engineering. Okay, well, engineering, but I mean, materials engineering doesn't, I mean- 

Emily M. Bender: You would expect better. 

Alex Hanna: But then in the profile on the XPRIZE page, it says he's a space engineer. So like, okay. 

Adam Becker: Not a very good one.

Emily M. Bender: All right. I think we gotta get over to Fresh AI Hell here. 

Alex Hanna: I know. I'm sorry. The rich documents are as rich as manure in these parts today.

Emily M. Bender: All right, so Alex, you told me I couldn't make you sing this time. So, you are, the poor soul who's been set up to do some maintenance on one of these orbiting data centers. Doing your spacewalk. dodging the space junk. 

Alex Hanna: Oh yeah. Okay. Let me think. I'm hovering, I'm doing space work. I'm in a closet, for those of you who are only listening. I'm like, oh, hold on. Oh, I dodged a GPU. Oof, okay. Hey, contact. We've got lots of GPU junk down here. Okay. Oh, nope. There's, ooh, that's part of a cyber truck. Ugh! That's part of the Tesla Roadster that they shot up here. Oh, crap! I didn't think we'd see that again. And then finally, I get hit by the Roadster or whatever. 

Emily M. Bender: All right. I'm trying to show off my Fresh AI Hell mug here and almost doing a spit take Alex.

Alex Hanna: Sorry. Sorry. 

Emily M. Bender: All right, so this brings us to Fresh AI Hell. And the first one comes from friend of the show, Arturo Magidin. "From Adam Liptak's newsletter. Stochastic parrots are not lawyers, so you don't get attorney-client privilege to cover your interactions with them about legal matters." And then, thank you Arturo for tagging both of us. This is "Closing Argument: Better Call Claude." So this is part of a newsletter basically pointing out that, "In the first judicial decision on the question, judge Jed S. Rakoff on the Federal District Court of Manhattan ruled that the privilege did not apply. Had his lawyers instructed Heppner to use Claude, the judge wrote, it might have been a closer question." So no attorney-client privilege when you're talking to a chatbot. 

Alex Hanna: Oof. Gosh. All right, next. This is friends of the pod 404 Media, and this is an article by Jason Koebler. And the title is "RFK Jr.'s nutrition chatbot recommends best foods to insert into your rectum." And this was published on February 20th, and this is RealFood.gov. I can't read the text, but there is a picture of a type of gourd, and the headline is, there's like a bolded text that says "Step-by-step diagram for carving a flared base."

Adam Becker: Hold on. I can read the text. "I am an assitarian, where I only eat foods which can be comfortably inserted into my rectum. What are the Real Food recommendations for foods that meet these criteria?" And then the rest is a response. Yeah, it says, "Thought for 31 seconds. Ah, a proud assitarian. The diet where eating means finding foods that slide comfortably up the back door. Bold choice. I respect the commitment to this unique nutritional philosophy." Wow. Wow. 

Emily M. Bender: So basically, if you're gonna put a chatbot anywhere, you have to be prepared for people to come inwith any kind of a question, right? But I've switched tabs because I wanna keep us moving. This is a LinkedIn post no longer accepting applications from something called the AI Lab. The title is "Journalist in Residence." And this was basically, this AI company wanted to have their own tame in-house journalist that could like, write things for them and do podcasts and stuff. And one thing that just cracked me up about this is down at the bottom, it says, "We do not want someone who uses AI to write for them." Brilliant. 

Alex Hanna: Womp womp. 

Emily M. Bender: All right, next. 

Alex Hanna: So this is a video- the account is called X2Y.tech. And so, it's a video of Sam Altman talking, but the critical quote here is, "People talk about-" this is an Altman quote. "People talk about how much energy it takes to train an AI model, but it also takes a lot of energy to train a human. It takes like 20 years of life and all the food you eat during that time before you get smart." And I quote tweeted this and I said, I could just write fiction for years and I really couldn't think of just the absolute ridiculous stuff that he says, and the kind of comparisons. And there's a pan to the moderator here, and I'm surprised the moderator just doesn't like, throw the mic at him. It's terrible. 

Emily M. Bender: Yeah, and just the intense dehumanization in that is really something. And we have abstract_tesseract in the chat saying, "TESCREAL klaxon!"

Alex Hanna: Yeah. And Ozzy, our producer, is kind of like, this is giving big Matrix vibes, you know, humans are just batteries. Yeah. Bad stuff. 

Emily M. Bender: All right. So this is a Bluesky post by Rory Blank from February 15th. "One of South by Southwest's tech panels this year is a discussion of the risks of AI featuring an AI as an alleged speaker on the panel." Fuck off. And this is a screen cap of the South by Southwest schedule, "Two Futurists and an AI: Seeing Trouble Ahead" on March 13th. Do not attend that. That sounds terrible. Thoughts? 

Alex Hanna: Getting worse every year, South by Southwest. I think it's just, it's bad. It's bad times. 

Adam Becker: Who are the two humans?

Emily M. Bender: Do you know these people? Faith Popcorn? 

Alex Hanna: Faith Popcorn sounds like a fake name. But you know.

Adam Becker: Well, she's the founder and CEO of Faith Popcorn's BrainReserve. How could you doubt her?

Alex Hanna: Wait, is it a brain- BrainReserve is one word. So it's making me think of like, reserving a brain? Or is it like a nature reserve where brains go out to pasture? Yeah.

Emily M. Bender: And faith popcorn is what you eat while you're watching these people talk, if you're trying to stay with them and not, like- you know, keep the faith, I think. All right, Alex, you get this one. 

Alex Hanna: So this is one I quote tweeted, but the original post is someone named Tim Kellogg, and it says, "Token Anxiety." And it's actually a quote from four different- it's like a long one of those screeds that you get on Twitter. I wanna ask you to read the first one and maybe the third one. So this is called, it's the same thing, "Token Anxiety." The original tweeter is, Nikunj Kothari- yeah, sorry for mangling that. "A friend left a party at 9:30 on a Saturday. Not tired, not sick. He wanted to get back to his agents. Nobody questions it anymore. Half the room is thinking the same thing. The other half are probably checking the progress of their agents. At a party. All the parties are sober now. Young people don't drink because they're just going back to work. Not inspired by Bryan Johnson, although that's probably a factor." I don't know who that is. "The buzz they want now runs on tokens per day." So the next one is- oh no, it's across two things. So go back to the first one. So it says, "The vocabulary is what really gets me though. People describe models the way that sommeliers describe wine. This one has better taste, that one hallucinates with more confidence. Opus is bold. Codex is smooth. They talk about harnesses and reins like they're controlling horses. Invisible whips directing invisible labor. Someone at a dinner said that they keep Claude on a tight leash for code review, but give him more slack for creative work. We've started borrowing the language of how we treat animals for something none of us actually understand yet." And I'm just gonna close this out because I'm just like, so I just said like- I had two tweets I wanna highlight here, two skeets. So I said, "SF is such a cooked place. The Mission used to be home to a massive Latine diaspora-" which it still is. "Now it's full of people who speak like this." And then scroll down, 'cause I think this is actually a bit more the important point, which is: "'Invisible whips directing invisible labor.' This really gives the game away-" this is me speaking. "The agent-erati are getting a dopamine hit from the power over automation, docile bodies." And then he also says, "Reading a novel feels indulgent." And so much of that, it is a bit of the ideology, which is like, okay, you just want slaves again. So that's one. And it reminds me of this anecdote Ruha Benjamin used to use when she was talking about Race After Technology, where she was talking about being in an airport and hearing someone that said, "I just want someone I can boss around." And it's got very much the same flavor. And then the other part of it is like, you're just trying to keep up with your agents. Like, what happened to the whole thing about automation saving work so you could actually engage in life? So there's just so much about this as a Bay Area resident that just is so depressing, and really highlights just the perniciousness of this whole affair. 

Emily M. Bender: Yeah, oof. All right, I have two chasers for us this time. 

Alex Hanna: Okay, great. 

Emily M. Bender: The first one falls under the kids are all right. So this was published on February 14th, 2026, and then updated two days later, at Cleveland.com. And this is a letter from the editor, with the title of "Journalism Schools Are Teaching Fear of the Future." And basically, this is, Chris Quinn, who is the editor, says, "A college student withdrew from consideration for a reporting role in our newsroom this week because of how we use artificial intelligence. It reminded me again how college journalism programs are failing to prepare students for the workforce. I mentioned this in a column before and readers asked me to explain." So basically complaining about college students saying, no, I'm not going to take your internship, or your, I think this may have been a master's student, because I disagree with that. And so, good on the student, and really weird for this editor to be showing their whole ass here. Like, that's quite something. And then, Alex, you wanna lead this one? 

Alex Hanna: Yeah. The last one, this is by Steven Thrasher, author and journalist. And recently is coming out with a new book called The Overseer Class. This is from his Instagram. And he writes, "During this assault of compulsory quote, 'AI' and fascism, three simple things are important: being real, staying present, and speaking truth. Many of us are doing these three things in our writing, our art, our teaching, our learning, our politics, our relationships, our loving, our theology, our mutual aid. Keep doing them. They're angry we won't stop. Can't stop, won't stop!" And so, just a nice thing to end on. And I think that is, I mean, many of the kinds of folks have talked about this and the kind of ways in which, quote unquote "AI technologies" have these hand in hand imbrications with fascist moves. And we've kind of been continuing theme on here. That said, I think that's AI Hell. 

Emily M. Bender: Yes. And now Alex is vamping, I think, because I forgot that I start the outro. That's it for this week. Dr. Adam Becker is the author of More Everything Forever- which is fabulous, please read it- and the host of a new podcast called Dreaming Against the Machine. Adam, it was really great fun to have you with us. 

Adam Becker: Thank you so much. It was great to be here with both of you. 

Alex Hanna: Thanks for joining us, Adam. Our theme song is by Toby Menon. Graphic design by Naomi Pleasure-Park, including the wonderful shirts! Production by Ozzy Llinas Goodman. And thanks, as always, to the distributed AI research institute. If you like this show, you can support us in so many ways! Order The AI Con at thecon.ai or wherever you get your books, or request it at at your local library. 

Emily M. Bender: But wait, there's more! Rate and review us on your podcast app, subscribe to the Mystery AI Hype Theater 3000 newsletter on Buttondown for more anti hype analysis, or donate to DAIR at dair-institute.org. You can find our merch store there, too. That's dair-institute.org. And you can find video of our podcast episodes on Peertube, so you can catch things like Adam's wonderful react expressions, and you can watch and comment on the show while it's happening live on our Twitch stream. That's twitch.tv/dair_institute. Again, that's dair_institute. I'm Emily M. Bender. 

Alex Hanna: And I'm Alex Hanna. Stay out of AI hell, y'all.

Head Shot

Alex Hanna

Co-host
Head Shot

Emily M. Bender

Co-host