Into the Impossible With Brian Keating - Earth Growing an AI Brain? Planetary Computation Explained with Ben Bratton [Ep. 486]

Episode Date: April 6, 2025

As a member of the Into the Impossible family, you get a special 20% discount on a subscription to The Economist. Visit their website at https://www.economist.com/Keating to get started. Pique is offering 20% off for life AND a free Starter Kit with your purchase—that's a rechargeable frother and glass beaker to make the perfect cup every time. Just go to http://piquelife.com/impossible Please join my mailing list here 👉 https://briankeating.com/list to win a meteorite 💥 What is planetary computation, and why does it matter? Are humans and AI truly separate, or are we actually co-evolving? And how did we end up building a megastructure that’s quietly reshaping governance, knowledge, and power on a planetary scale? Here to shed light on our evolving technological reality is Ben Bratton, philosopher of technology and professor at UCSD. Ben is the Director of Antikythera, a research program at the Berggruen Institute focused on the future of planetary computation. The program explores five core areas: planetary computation, synthetic intelligence, recursive simulations, hemispherical stacks, and planetary sapience. He is also the author of The Stack, a groundbreaking book that introduced the term “planetary computation” and mapped its political, social, and philosophical implications. In our wide-ranging conversation, Ben explains how computation has evolved far beyond servers and algorithms to become an organizing force of the planet itself. We also dive into the future of AGI, the ethics of machine valence, terraforming Earth as a philosophical imperative, non-biological life, and how speculative design and planetary sapience might reframe humanity’s role in the cosmos. — Key Takeaways:  00:00 Intro  01:02 Antikythera and planetary computation  07:50 AI and the limits of computation  18:04 Terraforming and the future of humanity  19:55 The role of language in AI  31:01 Planetary data centers and future energy needs 34:00 Techno signatures  39:10 Science and philosophy  41:09 The Stack: On Software and Sovereignty 45:52 Outro — Additional resources:  ➡️ Learn more about Ben Bratton:  💻 Website: https://bratton.info/  📸 Instagram: https://www.instagram.com/benjaminbratton  ✖️ X: https://x.com/bratton  📚 The Stack: https://amzn.eu/d/j80ENZW ➡️ Follow me on your fav platforms: ✖️ Twitter: https://twitter.com/DrBrianKeating  🔔 YouTube: https://www.youtube.com/DrBrianKeating?sub_confirmation=1  📝 Join my mailing list: https://briankeating.com/list  ✍️ Check out my blog: https://briankeating.com/cosmic-musings/  🎙️ Follow my podcast: https://briankeating.com/podcast  — Into the Impossible with Brian Keating is a podcast dedicated to all those who want to explore the universe within and beyond the known. Make sure to follow so you never miss an episode! Learn more about your ad choices. Visit megaphone.fm/adchoices

Today, we're exploring whether the Earth itself is developing intelligence through us. Soon you'll discover why the world's first computer wasn't built to crunch numbers, but to map our place in the cosmos. My guest today is Ben Bratton, a professor, a visionary philosopher of technology, and a man whose work has redefined intelligence at planetary scales. He's the perfect person to explore this mind-bending concept. Ben argues that AI isn't just another tool. It's the planet evolving a nervous system.

But here's a twist. Are we training artificial intelligence? Or, terrifyingly, is AI training us? First, we'll unpack computation as a natural force of the universe. Then, we'll ask, is Earth itself becoming conscious? Join us as we reimagine humanity's role in the greatest evolutionary leap, since life began. One of the best parts about doing a podcast

is that you get to invite your friends and people that you are so inspired by and that's no exception for you. I want to start with this new project that doptails so beautifully with what I do, which is possibly the existence of planetary scale, computing, and so forth.

And I thought we could start there with this concept that I'm trying to learn how to pronounce antichythera. Anticathera. Anticathera. So what is Anticathera? And why should we care about this as astronomy, you know, speaking to the most brilliant minds in the multiverse, a lot of them are

astronomers. I should say, first of all, that, you know, I call myself a philosopher of technology, right? I'm coming from the humanities side of the house. Visual Arts, right? That's your... In the Department of Visual Arts, you're at ECC, right? But I'm a little bit unusual in this regard, that is, I take a very intense and sincere interest in what science is doing and the way in which emerging technologies not only allow us to do new things, but to know new things. The relationship between philosophy and science, has been one that has been, you know, variously contentious, but also one that has a very tight genealogical relationship sometime, right?

And so you may have people like, I don't know, Lawrence Krause who sort of see them as strongly opposed to one another, whereas others like myself may recognize that all the sciences that we recognize today, one way or another, began as philosophies. And so like, you're going to say, like, when philosophy learns to ask the right questions, a new science is born. But when, so where are new philosophies? come from. They might as to say when new technologies force us to realize that the languages

and concepts that we've been using to understand the world are inadequate or anachronistic, moments like these when philosophy, I think, becomes most useful in a creative and generative sense, right? So it's less about how do I take these concepts and apply them to new things. What would Hegel think about driverless cars? What would Kant say about this telescope? But rather, how do we generate the concepts that allow to bring something new. So Anticathera. Anticathera is named after the, what is probably apocryphly the first known computer was from 200 BC. It was discovered off the island of Anticathera in Greece in the beginning of the 20th century. And it took a really long time to figure out what it was because it was this intricate, complex, geared mechanism that in the

course of technological evolution, and I should say I actually believe the technology evolves in a literal sense, it was completely anomalous. There wasn't a, you know, a predecessor technology. Yeah, it's a predecessor species before or after it when it were iPhone version one, you know. Exactly. Exactly. So what eventually was figured out that it was, in fact, a computational device. But in addition to being a device that allowed you to do basic mathematics, it also was an astronomical device. It allowed you to map your position in relationship to the stars and planets, Not only in the present, but also in the past and future. So it was a kind of, you know, a temporal mapping device as well.

Interesting. And so for us thinking about how do we, what should be the foundation for a 21st century philosophy of computation, the idea that computation really begins, or artificial computation, I'll also make the argument that we discovered computation as much as we invented it, the foundational idea for computation that it orients intelligence in relationship to its planetary condition. as a starting point for not only what computation is, but ultimately what it's for, it was sort of the inspiration for this. And so I can talk a bit about the research program that we're sort of developing there, but to your question about what do we mean by planetary computation, more

recently, I mean, zooming out from 200 BC to the present time, the way in which we look at it is that we need to think about computation not only in terms of algorithms and mathematics and forms of calculation and inference and inference structures and so forth. And not only in terms of thinking of computation, in terms of what we recognize as computers, the kind of relatively energy inefficient appliances that we have managed to construct, but one, to think of, understand computation as a natural force in the universe, that informational system that the universe computes,

and that we essentially have learned how to discover and tap into this. but a computer is no more computation than a light bulb is electricity. Hmm. But to the planetary computation, over the last 50 years, arguably, you know, really beginning in the 1960s, artificial computation, an artificial computational infrastructure has wrapped the globe. From subterranean data centers to oceanic fiber optic cable to the miraculous glowing glass rectangles that we all carry around in our pockets and still call fiber. bones for some reason, to the satellites overhead. You could think of it this way. You know, the famous blue marble image, right? I just was speaking with Stewart Brands a few weeks ago. Imagine the blue marble image, not as a still image, but rather as a kind of super fast-forward

movie that sort of showing the whole several billion years of the evolution of Earth. What you'd see is it sort of goes from red to green to blue, the oxygenate life, appearance of atmosphere, panegia, all the rest. And in the very last few seconds, you would see something kind of remarkable that this glowing blue orb would sprout this. It would sprout this sort of exospraying into being. Sensory exoskeleton. The way we see this is that that's not only something that humans did very recently, ultimately we need to see it as something that the planet done. That we have one planet, at least that we know of, that has evolved a mineral-based sensory organ that exists not only on its surface but in its low-worth orbit,

and that it uses to know things really fundamental about itself. So, for example, I would argue that when we ask the question, what contribution can planetary computation make to the mitigation of climate change, which is a perfectly reasonable question, we have to understand that the very idea of climate change as a scientific concept is an epistemological accomplishment of planetary computation in the first place. Of the very entities that can comprehend and contribute to it, that's right.

That's right. And so planetary computation, as we see it, is both something that allows intelligent life, both human and non-human, arguably, in terms of AI, to do things that it wouldn't have been able to do. But it also allows us to know things that we wouldn't have known before. So for us, that's the stakes. That's the stakes of this. And just as astronomy is arguably like the way in which the planet has folded itself into the shape of astronomers in order to know things about itself. It's all part of a similar kind of problem. It's all part of a similar kind of problem. of the emergence of what we may think of the kind of planetary sapiens. It's only getting every customer's order right. It's only a point-of-sale system connected by Spectrum-fiber-powered business internet, helping you track hundreds of secure transactions. And it's all backed by 24-7 U.S.-based customer support and local technicians. It's only everything.

Get business internet advantage free forever when you get four mobile lines from Spectrum. Visit Spectrum.com slash free-for-life to find out how. Restrictions apply. Service is not available in all areas. This just, you know, stimulates the seven podcasts worth of ideas in my mind. But, you know, one of them that I love to tweet out every couple of months is, you know, kind of aping Einstein by saying, I don't know what, with what, you know, what hardware, you know, chat GPT-5 will run on, but chat GPT-6 will run on an abacus, you know, basically or an antique, you know, whatever. Oh, I can't pronounce it. But the point being, you know, when we think about what.

AI is capable of, it seems to me there's almost as AI of the gaps. Like, whatever we think that AI or computation or planetary scale things can do, it can do everything that we don't understand. And that's kind of paradoxical, because we don't actually fully understand the limits of this technology, of the, you know, computational limits, except via physics. And the first guest on the podcast was Freeman Dyson. And, you know, you're sitting in the seat basically that he was in. And the point that I'm, you know, kind of getting at is that are there ultimate limits, you know, to things like the singularity, to things like the simulation hypothesis, to global, not just planetary scale, but universal scale, at least galaxy scale. Are there limitations that we can actually set right now on the probability or the, let me say, on the idea that this is an evolving system and that we're just one, one genus on the way to, you know, who knows where we're going to evolve to?

But is it not true that we may be, you know, both the creator of this tool and its last users of it, effectively? The first one, let me talk about the AI of the gaps. And I think there is a way in which the AI, trying to determine the status of AI, is it intelligent, is it not intelligence, does mirror the kind of God of the Gaps. Think back to Turing's famous 1950 experiment. What Turing sets up is a highly functional definition of intelligence, right? But he also sets up one in which he asks, you cannot tell. whether Player A is a person or a computer. Or actually says if it's a woman,

which has an interesting valence in terms of his own biography. But then we have to grant that there's something going on, right? And so what he's positing is a sufficient condition of intelligence. Unfortunately, I think a lot of the ways in which during tests has been incorporated at a cultural level is more of a necessary condition. That is, unless the AI can perform thinking the way that we think that we think, then it's disqualified. It's a mirror.

It's a mirror. And so what that means is that there is a way in which the question is framed in a kind of either or opposition, that it's either human or AI, that the relationship between humans and AI is one of a boundary condition. That is what humans are is what AIs are not. What AIs are is what humans are not. And so what you're beginning to see is first a kind of receding border that more and more of the things that we may have said AI will never be able to do. It does, right? And so, you know, And, you know, people keep moving the goalpost as fast as they can possibly run to keep this going. But it also means that what it means to be human is defined in relationship to what we think AI is. AI might be training us as much as we're training in, right? I don't think we've ever been in charge of our technologies. I think, you know, we've co-evolved with them from the very beginning. Right.

Now, to the other question, is there limits of, you know, limits of this as a larger computational thing? I don't know. It's a sort of a question. is there is there a limit to how large and complex and informational structure can possibly become? Is there a limit to how much of a kind of, you know, intensive energy event you can, you know, artificially produce? And then you get back to, you know, Dyson's form kinds of questions, right? And you begin to scale things out to a galactic, a solar or even galactic scale. And you're asking, you know, what is Cardishab two?

How far are we from Cardishab one? Cardishab two, so forth and so on. I want to connect this with the gap and opposition problem. I don't see it as opposition. I don't see humans and technologies opposed. I see them deeply, the evolution of both is deeply intertwined. And so when I think about this, I don't know if you read James Lovelock's last book, Nova Scene, where Lovelock, of course, posits the original guy hypothesis by which the emergence of simple life transform, terraform the planet to make the conditions for complex life and that this is this terraforming is an ongoing process. He concludes the book that with, you know, looking at the emerging and accelerating evolution of machine intelligence and says this is a similar kind of phenomenon. It's going to terraform the image in its own, you know, the earth and its own image. But, you know, the book doesn't end on a pessimistic note. And it's not one of, well, we've had a good run, you know, we might as well, you know, drink the hemlock and be done with it.

I think humans will survive really long, you'll continue to survive longer, but they won't, they won't necessarily be what we imagine them to be today. I just think of it. All biological systems become, you know, I know you've had Sarah Walker. around the show many times. Yeah. All forms of, you know, complex biologic systems become components in yet more complex systems. That's right.

Humans will persist. They've always been part of systems larger themselves, and they will continue to persist in systems larger themselves, some of which of their own making. So one thing I've brought up in the past is that there's sort of this fallacy of iteration that, you know, there will come upon us AGI, and then it will do things that will be detrimental to humans. I'm just painting very, very broad brush. And I think this fear is sort of misplaced.

well, if it's as smart as us, which it has to be for super intelligence, it has to be smart of us, right? But it will know at least that fact that we're concerned about it, AI, replacing us, right? So then it will be afraid. It will know that it's not the final evolution of its own technological advancement, right? I mean, presumably chat GPT, you know, infinity could exist the same way iPhone 18 is sure enough to come down the horizon as, as, you know, as the earth will make another lap around the sun, we hope. But the point being, if it, if we're smart enough to conceive of this as a challenge,

surely AI would say, well, look, this can either be a threat to me as an AI, or it could be something so beneficial, in which case, I have to let it be created, in which case, it's not a threat to humans to begin with, right? Because otherwise, the AI wouldn't allow it to be created. How do you react to kind of this solipsistic definition where, you know, the infinite regress of iteration where eventually there's a final iteration, right? I mean, I always say a computer chip wasn't created with a computer chip, right, by definition. An operating system wasn't going to know. They are now. No, no, but the original one, the first one. So I'm saying the first AGI will not be created with an AGI.

I mean, there'll be some boundary. If we get there, I'm not convinced that it's right. I am, you know, good thesis back to the 60s that the AI, if we make true AGI, it'll be the last technology we ever make. Again, I don't, I see this as a, you know, if you zoom out far enough, I see this really as a kind of a process that is continuous, but one that it can be is marked by forms of, you know, of threshold events in the Maynard Smith sort of sense, punctuated equilibrium. There are step functions in which complex forms of biological, technological, convergence have these moments where they radically increase in their information processing capacity, which was Maynard Smith's sort of key argument. I think we're in the middle of one.

I don't think we're caught up to it paradigmatically. I think we're in a kind of pre-paradigmatic moment where then terms and concepts we used to even describe this is to recognize technological evolution or something we're part of. I think we have to grow into that very, very quickly. I don't think there's a final event, right? I'm not an eschatologist. Right. You know, I think the future is much more contingent than it is, you know, that we are,

there's certain path dependencies that are built into forms of convergent evolution, but for the most part, you know, it can go a lot of your ways. I think in the short term, the emergence of AGI, you know, I think we're seeing a kind of slow take up of that over the last few years, is one that's going to be highly disruptive, really disruptive in terms of how, what an economy is, what a society is, how we relate to, agents, AI agents that are trained on us and our relationships with simulations of ourselves as a fundamental tool. This is going to be beyond future shock. It's going to be a Copernican existential trauma. And it's going to be a really bumpy ride for a while.

Especially for certain populations, right? Yes. But as you zoom out, I really kind of think of this is part of a larger process that can be mapped that we can see ourselves inside of to sort of understand that we have certain degree of, you know, as a sapient species, we have a certain degree of, of agency. I mean, one of the discovery of the Anthropocene is which we actually have, we actually can terraform the planet in a relatively short period of time. And we need to do it again in a way for to to keep things going. That would discover forms of agency we didn't know that we had, but we're going to take a long time for us to mature into those, the subjectivities that are needed to actually

have some degree of control over those things. So I'm optimistic in the long run and buckling up in the short row. We need the answer to the question. As a professor and a podcaster and a scientist exploring the boundaries of knowledge in the lab, I crave independent sources of knowledge and wisdom wherever I can find them. One of my go-to sources is the economist. I've been reading it literally for decades. The economist consistently delivers independent journalism for independent thinkers like you and me. But the best part for me is that their science coverage is nothing short of extraordinary. Just this past week, I was captivated by their deep analysis of the recent claims of quantum computing breakthroughs that could revolutionize how we process information,

exactly the kind of forward-thinking coverage that I discuss on the podcast and that I crave to expand my own intellectual horizons. The economist's rigorous approach cuts through the noise and BS in a way that you can truly trust. And what sets the economist apart as their simply science weekly newsletter. It's a subscriber-only exclusive benefit. Each weekly edition delivers beautifully written explorations of nature, technology, the cosmos, the human body, and all this comes alongside stunning scientific photography selected and curated by their editors. I don't just do science. I love to explore complex ideas.

The Economist analysis is a perfect compliment. It provides the insight and clarity needed to navigate challenging intellectual terrain with confidence. And to get everything the Economist offers, the Economist is offering a special 20% discount just for viewers of the Into the Impossible Podcast. To claim the special 20% discount, visit Economist.com slash Keating. Your pursuit of knowledge shouldn't end when this podcast ends. The economist ensures that doesn't have to. I think you've written that, you know, terraforming in this way in the context that you've described

is sort of an imperative. What did you mean by that? What does that mean? Yeah, there's a book I wrote, it was really an essay that became a book that got translated very quickly. There's a longer version of this book that's going to be coming out with MIT Press next year called the terraforming. And the argument, in a nutshell, is that we need to think of terraforming. forming of the Earth, not of Mars or something to make those other planets habitable

for Earth-like life, but the terraforming of Earth to ensure that Earth remains a viable host for Earth-like life. Right. And so you could think of really there are three terraformings, I think, that are relevant here. One is thinking about what Kruitsin and others called the Anthropocene as a kind of planless, headless terraforming initiative that we fundamentally transform the planet. Now, you know, now we call an Anthropocene event. it is supposed to an era, fire, but that it was a terraforming scale event and that you were

born inside of it, that we were all born inside of a terraforming event. And that's like welcome. And meteorase, right? The second terraforming is the one that's inevitably going to happen no matter what we do, that we could all go full Unabomber tomorrow and the ongoing effects in relationship to everything from oceans to land use to planetary geochemistry is going to, the momentum will outlast us. And the third one, is potentially the more deliberate process by which open-ended, non-falsifiable questions,

like what kind of planet should we make? Where should the people go? How do you get, what should be the cycle? What do you work? Deep utopia. Neatutopia needs to be defined, but it needs to be designed and composed. And I think to your question,

what is the role of machine intelligence within that? I think it's probably quite clear that there's a lot, but it also may be quite nonlinear. But anyway, that's what I mean by the term for me. How much is AI, at least these LLMs, I mean, the L, you know, for language, it seems to me that is incredibly powerful. Humans are language machines. It's how we, you know, establish dominance and how we preserve technology, science, culture, through language. But it's not all we do, right?

You could say that you might be able to code mathematics and physics in terms of a language, but it's different from the types of language that these LLMs are trained on, for example. Yes, they can be trained on my PhD thesis. But what I'm getting at is the type of science, say just restricting it to a search for a so-called theory of everything. How AI is or is not going to be capable of that. For example, I don't think it's a matter of prompting. I don't think it's a matter of the corpus of training data that is precluding AI from right now coming up with a theory of everything that Edward Witten and all of our friends that we know and love here and beyond have been unable to do. I don't think it relies on the same, you know, vector space, you know, dot products and matrix manipulations that have been locked in effectively by the confluence of GPUs and, you know,

matrix algebra and neural networks. In other words, there could be, I'm not saying there is, but what if there is, a fundamental difference in the types of language, if you will, that we use to communicate and so forth, that these elements are uniquely, you know, well qualified. They're already IQ 100, 120 in every subject matter from law to. medicine and so far. But so far, you know, the stubbornness of, say, finding a theory of everything or discovering some missing matter, the constituents of missing matter, is that going to permanently evade at least LLMs? And because we're so invested heavily into it, we're locked

into LLMs as far as the I can see. I mean, there's almost no, you know, there are alternatives, but they're not at scale of $300 billion, you know, valuations, et cetera. And I just don't think that the thing of preventing us from getting a, theory of everything is the fact that the latest, greatest generation of GROC or whatever, hasn't been trained on the Fast and the Furious, you know, 12. In other words, is there something uniquely different about some types of what humans can do that will make us preserve what we do uniquely in perpetuity? To the last question, for sure. For the last question for sure. Like, again, I don't think

it is necessarily should be, there's a limit to thinking about this as a replacement dynamic, right? I think that in the long run, AI will teach us what thinking is as much as we will teach it how to think, that the demonstration of another substrate of intelligence, you know, like the planet Solaris was another substrate of intelligence, that we can compare ourselves to and understand where we fit within this larger spectrum, I think, as part of the understanding that will come to this. But in terms of the more specific questions you get there, like, you're right. I think there's good reasons why it turned out language was the path. Right.

You know, Hasivas had bet on games because he was a chess prodigy. And so for him, games was going to be the path. And it turned out, actually, language is a repository of all different kinds of intelligences. It's something that evolves over time. It's something that speaks its users, more than the users speak it. And that producing a kind of a model that is able to think and to respond to the world generatively using language was a shortcut that shouldn't really be that surprising in hindsight. First of all, the LLMs are evolving very, very quickly, right? I mean, just there were no real, the kind of reasoning models that we see today just didn't exist while ago.

And so I think LLM will evolve into other things that would be unrecognizable to what they are today. But you're right, language isn't the only thing. I think, in a way, you know, whether models can be trained. I mean, anything can be tokenized, right? Now models are constructed on that which can be tokenized, which could be anything, principle, right? And that is made into something, a kind of language in a way. And so I think the movement, you know, Faye Faye Lee's new startup is about spatial intelligence. Yon Lecun at META is focusing on spatial intelligence.

And he, I should say, is quite pessimistic about, you know, the organization. He is. I'm just focusing particularly on the spatial dynamic of the question, which suggests that other implications of spatial intelligence models for robotics. And, you know, Hubert Dreyfus' old critique of AI is that it can't really working the work because it's not embodied. It doesn't have embodied phenomenological experience. Well, like, that's no longer the case.

the physicalization of AI, where you've got spatial models and linguistic models that are together interacting with the world and ultimately in the long run through forms of active inference where their experience is retraining the model almost in real time, then you have a different kind of takeoff that may take this in a really different direction. But the second about the limit. While you were speaking, I was reminded of a – I should say Anticathera also has a journal that we edit with MIT Press in a book series. First book in the series is a book called What is Entail? by Blaze Aguirre I work with also at Google, at Google, fantastic book.

Another piece that we're publishing the journal later this year is a piece by, who you may know, Risa Weschler at Stanford. Stanton Feren't. Yeah. She's all along the law of Simon's Observatory in Columbia. That's right. And so Risa is doing a piece for us on the role of computational simulations in her work. And then the simulations as a kind of epistemic technology.

And I don't want to like, you know, oversimplify the point. I'm sure I will, but I mean, ERISA would say that one of the reasons that we, you know, can't figure out exactly what dark matter, dark energy is we don't have the compute. That if you think about, if you think about all the available computation on Earth at this moment, like if you network everything together into one giant machine, let's call that one unit of planetary compute as of 2025. How many units of planetary compute would be necessary to produce a simulation at a granular resolution sufficient enough to make, you know, important conclusions about the network?

People say the same thing about the weather, you know, similar in climate. But, yeah. So, Riesu would say, like, no, there actually is, like, a brute, there is a, like, a brute force scale to scientific ontology. The more compute, the more you know. Right. That would be one answer to your question. Like, is there a way in which are there things that we can't know, even if we had a lot of compute? Probably. Yeah.

Are the things that we don't know that if we had orders amounting to a more compute, we could know for sure. Yeah. Mm-hmm. One thing that is sort of fascinating to me, and you did bring it up, and I brought out many times, is, you know, this, the famous statement by Einstein that an observer in free fall would experience no gravitational force, right? Such a lovely thought. It is. And it called it his happiest thought. He called it the thought that titillated him more than any other thought he ever had.

But that brings up, you know, the exact, you know, topic that that you hinted at before, which is embodiment. To what extent can a computer have happy thoughts? And to what extent can it experience this early or no? In robotics with a gyroscopic sensor and attitude indicators and inertial reference, you know, laser gyro, even such a device. Can it visualize what your, you know, my toddler can visualize as going on a roller coaster driving over a bump and that feeling in the pit of your stomach? And was that not, you know, the sine qua non that allowed Einstein to have the happiest thought, as he called it? So these two things. Can they have happy thoughts?

If so, can they have painful thoughts? Do we have ethical requirements and imposition on us? as they're programmers and then when they exceed us. And then second of all, can they do it truly without embodiment? Are there things that are, you know, to use, I forget the gentleman that you mentioned, but this is something I've thought about a lot. Do we need this right embodiment? Hubert Dreyfus was a philosopher.

I said Chomsky had a similar... Berkeley in the 70s, yeah. He wrote a famous book called What Computers Can't Do, which was a really important critique of sort of Simon era good old fashion AI and kind of laid the philosophical foundations for what Brooks picked up later as a kind of bottom-up model. So your question about... Is there, can a sufficiently advanced form of machine intelligence have valence?

Yeah. Can it feel bad or good for something to happen to it? Qualia that we experience, right. I wouldn't say qualia. I said violence. Yeah, I'm more on the Dennett side than the Chalmers side of that one. Dennett was his final interview, yeah. That's almost a year ago.

His biography, his latest autobiography, is fantastic. But can it have valence? I think it's an interesting question. I think it would be a mistake to presume too quickly that it can't. And I just, the lesson I think to learn there is how badly Western philosophy underestimated the complexity of animal intelligence. You know, as early to 20th century, you have people say animals can't feel pain. Right. Bent them. Well, today we would think is insane, right?

And so now you have this, you know, just in the 20th century had this very interesting convergence of human neuroscience and AI, where they're sort of bouncing back between each other and even neuroscience and flow. You know, the church lens here, for example, where this feedback loop between AI models and the way we should talk about this was very important. I think now you've got one between studies in animal cognition and studies in machine cognition, where it's a kind of comparative non-human mind studies. That's very, very interesting. So you've got people like Peter Godfrey-Smith who are looking at, you know, cephalopod intelligence. I think you have to think this is a kind of parallel discourse with what are, how would we know? How would we test?

This is a real question. There's a couple people that I have collaborated with Winnie Street, who was part of the Anticathiric Program, her colleague in the group Paradigms of Intelligence, where I'm a visiting faculty researcher at Google, another Jeff Keeling, philosopher there. All their work is on how would we know exactly the answer to this question? Right. And if we don't know, what are the rules by which we may have learned from animal welfare sort of standards by which they might apply this as well?

It's an open question. Yeah. But I think it's one that will become increasingly relevant. But I think it's a different question to the real gist of this. First of all, I think AI is embodied. It's just embodied not in a tetrahedal-hedral body plan. It doesn't have a stomach, so it doesn't have a pit in its stomach. But I think there's a limit to anthropomorphization. And to the extent of which we cannot, we are frustrated in our attempts to fully anthropomorphize it, that's not the same thing as suggesting that, you know, across this different substrate can may have other kinds of things.

Like, there may be forms of pain and pleasure that it can perceive that we cannot proceed, and vice versa, and that's fine. That sensation may, you know, may not be the critical, you know, quantity. It may just be one. It's a script for it. Right. Yeah. It's important. And many other scientists don't have any sort of physical and, you know, embodiment.

That's what I'm saying. Yeah. Yeah. I mean, some of the, you know, theories of mind before we go on to, you know, think about, do artificial intelligence feel pain or whatever? We have to look at our own colleagues and understand, you know, how does it feel? Yeah. Now, I mean, what is painful about not? failing in an experiment or not getting

some data or predicting something that's more simply. You know, you know, more simply you know, Damasio's work on the importance of emotion in these kinds of things and, you know, constructions of other forces of other forces. I think what we call, I think the distinction between emotional and rational intelligence is probably overblown.

And don't see any reason why we wouldn't consider, the forms of valence would not be able to operate across different kinds of sub-trade, even beyond just simple knee-jerk reward systems. This episode, is brought to you by Netflix. Most valuable promotions in Netflix are hosting a blockbuster triple headliner Saturday, May 16th.

Rhonda Rousey returns to face fellow woman's MMA pioneer Gina Carano in the main event. Plus co-main's Nate Diaz versus Mike Perry. And the best heavyweight in the world, Frances Ngano versus Felipe Lenz. Watch Rhonda Rousey versus Gina Carrano, live only on Netflix. Saturday, May 16th at 9 p.m. Eastern Center time, 6 p.m. Pacific time. So you mentioned just in some correspondence we had before this. We were talking about planetary data centers and computation, some work you're doing with Google. Can you explain a little bit more for the audience?

What does that involve? What interest is Google have in other planets? I mean, there's some of it. There's some things that are being worked on. I probably can't talk too much. I can't talk too much about. But I would say, like, you think about the idea what I suggested before that has to do with, if you think about all of the aggregate computation that exists on the planet now,

versus how much exists on the planet 10 years ago, versus how much existed on the planet 50 years ago. Clearly, we're in a kind of exponential spike in terms of the amount of the complexity of, the amount of capacity for artificial information processing, which is something that we turns out very good at, Grito. As I say, like, the fire apes have figured out how to make the rocks think fast.

You say big rocks, right? We take bits of rocks and minerals and we electrocute them just so, and the other rocks can do things that only primates could do, which is, you know, big news. Yeah. But if you begin to think about, you know, where this goes in the long run, you know, what is 10x planetary compute, 100x planetary compute, 1,000x planetary compute,

which are all kinds of things that we may see in our lifetime, you probably will see orders of magnitude increases in efficiency in terms of, you know, from neuromorphic chips to other kind algorithms that make, you know, the amount of energy that goes in for a certain unit of compute, dramatically, certain of it. But you also see, you know, all of the big platforms. forms, buying up every available electricity contract they can find, starting of three-mile island. And so it turns out AI may be the economic solution to nuclear energy. So you think, like, okay, AI is what's driving the economy of nuclear energy, GPUs is what's driving the economy of AI,

and Fortnite what's driving GPUs. And so by it transfers, Fortnite is saving, it's saving a lot of breath. But a minecraft, it could even be. For, yeah. But if you really get to like, okay, now we're beginning to, we're not quite Cardish have won, but we're getting, we're closer, it becomes a relevant question of like, what are, how would a planet make functional use of the available energy that it has? Not only, Jim, minerals and so forth. So it brings up a chance for me to give you your gift of appearing in person on the podcast. This is a genuine meteorite.

So this is a fragment of the early solar system, about 4.3 billion years old, older than the earth. I give them out on my website, Brian Keating.com, slash, or slash list. But you're guaranteed to win one if you have a .evu email address like you do, and you're guaranteed to win one because you are such a gracious intellect in front. If you're a platform that operates at the scale of a state, which is what Google does, and you're thinking 50 years out, 100 years out, about how you would, in fact, power a thousand X or, you know, or is it 10,000x a version of the sort of thing. You begin to, you begin to have crazy

the ideas about where you get the energy from, and I'll leave it at that. Okay. Yeah, fair enough. And those are some of the most exciting things from a physicist perspective, right? I mean, we'd love to tie into the late-grade Freeman Dyson's ideas. But it does bring up not only energy, but also minerals and so forth. So the average enrichment of this metals 100%, you know, it's basically pure iron, nickel, cobalt, very, you know, rare, you know, in terms of the overall crustal, you know,

the composition of the Earth, but by no means unique. And so some of the questions I have at least about the planetary scale, maybe you'll have to do a part two eventually about this. But on the trajectory of Earth's evolution, you know, is the, is the a tinkera, I mean, is the planetary sapien, so to speak, that you talk about? Is that truly, you know, is that meant literally or is that more metaphor? I'm asking for the following reasons. We have a very Earth-like planet in a very habitable zone planet called Mars, and it's

very close to us. It's completely barren, sterile. It had water on it. It had water on it when life existed on the Earth. And there's a phenomenon called panespermia. Fred Hoyle popularized it were planets exchange material and other one's genetics. Biological or mineral, like planets are not closed systems. That's right. They're exchanging continuously. And yet we see pretty much everywhere we look.

Doug and, you know, we stuck probes. We violated Mars and basically every way we could possibly do it for an animal. And we see no life. Now that's not proof that life never existed. It's not proof that life couldn't exist in the future. That we recognize as life. As we recognize as life. As we recognize as life. And Sarah, you know, it always correct me whenever. Right. It's not as we know it. Exactly. A wonderful new look. So my question is, what would you look for? My colleagues upstairs that use Keck Telescope, 30-meter telescope coming out in the next few years, the UC project. What would we look for to see planetary compute? My friend Adam Franks spent on many times, he talks about global warming as a techno-signature, right?

Agriculture, at least, right? What would we look for? What sort of techno-give my young viewers out there, listeners, some inspiration? What would you look forward to validate or falsify this notion of planetary sapiens? How could we do it? What are the pieces that we're also publishing at the key of the journal as a new piece by Sarah and Lee Cronin,

her bloggerer on. It's kind of like what, it's sort of the technological version of the question of life. We don't know it. Like, how would we find forms of technology on the other words that we don't recognize as being Earth-like technology? Or natural technology.

Or natural thing. Or first object. So let me make a sort of wild conjecture here, sort of for the end here as well, is that one of the things about the pre-paradigmatic moment that we find ourselves in, is that functional definitions of life, like the one that Sarah and Lee developed, which is a kind of, you know, you've got this increasing complexity,

some way. Componentization, one thing and not done this study is generally. And our contemporary definitions is definitions of technology, which are also kind of evolutionarily, you know, Brian Arthur and the sort of like complexity inside of further complex, all technologies are built at previous technologies, so forth, and our definition of intelligence of this sort of consistent that's able to, you know, both auto-poetically and alopoetically transformed the world

in a certain goal-directed way, are starting to converge. One way to think about, that is, what we define as life in the Aristotelian sense, and we define as technology, we define as intelligent technologies, are looking more like each other than they did before. I'm not saying they're all exactly the same thing,

but they may, in some respects, be different ways of describing a set of very similar phenomena, and that may be the way we, what we may arrive at. So the difference between a check-no signature and a, I mean, biosignatures is one thing. Signature of life is something else, right? I think in many ways technology, I don't think life needs to be biological.

And I think the forms of technological evolution around us, which are evolutionary, that do follow these paths, is exemplary of that. I don't mean hypothetically, life doesn't need to be biological. I think we're surrounded by non-biological life and always have been. So the difference between finding a signature of life and finding a technocidure may be academic, maybe, maybe academic. But what's an example of a non-biological way? AI.

Okay, so I don't know. Literally. Okay. Yeah, no, literally. So you would consider that biolai, like, I have not to be studied in the biodeform is light, but not a lie. Right. And so, and I keep trying to make, sir, but there was a piece that we published in Noamia called AI is Life where she makes a similar kind of argument here as well. So what would we look for?

Let me think, let me try to think, of what we would try to fight this as well. One answer to the question is you may, it may be a kind of sampling problem where you need, it's not looking for a thing or a categorical. of matter, you would be looking for a process of transformation over a longer period of time where there's a certain kind of state change. Teleological. From one thing to another that may be the thing you want to find. There's something that demonstrates sort of like moments of increasing complexity that we're

not there before. And this intensification of localized complexity may be the sign that there's something going on there that may be technological, maybe life, maybe something, maybe another thing we don't we don't quite recognize. But I don't think, and I think, you know, a lot of the discussions in astrobiology, both official and non-official, let's say, scenes are sort of concluding that, you know, what we think of as life is, again, not a, not what we recognize life to be, and nor is it a kind of matter, potentially something like quite alien.

It's still the same table of elements, right? Right, there aren't any dotted periodic. I think it could be a dark periodic table. But what the substrate might be we don't really know. But this is, to me, like, what's really wonderful about, you know, things like, Fermi paradox as a philosophical problem, because they force you to think about what is the possible boundaries, what are possible conditions of what a civilization might be, what communication might be. That's right.

Is there some mathematician? To me, this is where science is doing good philosophy, and hopefully philosophy can contribute to the science an interesting way, too. Yeah. I mean, as Galileo, my hero said, that's not Galileo, that science study. Galileo said, he was a natural philosopher. And so he said with his telescope, which is technology, he didn't invent it, but he perfected for that time. He said that he had, quote, and I quote, let's see if I can remember that.

That's made irrelevant the disputations that have, for generations, vexed philosophers. In other words, by observing that the Milky Way was comprised of individual stars rather than this discrete, you know, flowing plasma, whatever, and certainly with the moon surface that he was able to reveal that it wasn't this idealized thing. He was then instantiating a new hypothesis, right? That's what philosophy does. Now, he was a philosopher himself, except they, He's called physicists, natural philosophers. But his ultimate goal, as he said, was to measure what's measurable and make measurable what is not yet so.

So, yeah, I mean, vision in the future, how could we develop a Drake equation, you know, for these technological biosapians and these technological sapiens? Where does that fit in and kind of in your research profile? I mean, we didn't even talk about visual arts and so forth and speculative design increasingly alienated relationships. That's fine. But that was originally, you know, what brought you here.

And I think that there's, you know, your work on how visual arts can influence philosophy and there could be the symbiosis between them in the past. We started a new major here. Yes, speculative design. That's right. He's trying to bring together a kind of the creative, open-ended exploratory sort of approach of art studio. But starting with questions that are informed by what's going on the labs around here. Well, I'd be remiss if I didn't at least have one question about the stack.

You've got it there. You brought it. You're so kind to bring it. Tell me about the stack in the upcoming. reprinter or second edition. There's an obsolete wrap here to the Canada. There you go, Marshall. There's a 10th anniversary of it coming out later.

So this was a book that I wrote 20 here at UCSD. It was a book that I went to sort of came here to write. It was published in 2016 originally. And then I was making what was that, at that point it was making was a kind of somewhat science fiction sort of argument was that what we recognized as planetary computation was transforming the entire structure of political geography in its image. that the emergence of planetary computation was transforming this relationship between technology and state, between the political and governance, and that the forms of geopolitical dynamics that we would see over the next coming years

would be fought over the competition to produce large VAC systems, to model societies through these large VAC systems. And again, by then this was a bit kind of trippy of uncons. And now it's sort of taken for granted. I think that's just kind of obvious. from chimp wars to everything else. The other argument, though, is that we can see this thing called planetary computation, not as a single undifferentiated mega machine in the way in which Louis Mumford might have thought of it,

but rather as something that is literally like a stack. It is comprised of modular functionally defined layers, where for Earth, Cloud, City, address, interface user, they each have their own dynamic, each have their own sort of process, each is replaceful by different kinds of technological systems. and becomes a kind of schema by which from energy sourcing to service provision to, you know, the images that we use to make sense of that space excrucients to the interface that's right, that it becomes a kind of discontinuous accidental megastructure. That's what we have produced. And so the book is the story of where that came from, how you build a society with it, what its integral accidents are, and indeed, more importantly, how we think about the stack to come.

Because the great thing about stack systems is it designed to be replaced. and not his species as infrastructure. That's right. So that's what the book is all about that. I was thinking yesterday it was good that, you know, they didn't burn the ships when it came to Theseus. But also, I can't think it was an accident that the vice president recently made a mention of the stack

or maybe not specifically of the book, but I think he's been influenced. AI staff. AI stack, which is a component of what the book is all. I'm not going to take any connoisse. But it's interesting to see how sometimes percolated to the zeit guys. That's right.

And that's, I mean, that, you know, To me, I mean, I'll just put this, you know, make this last point. I think the relationship between philosophy of technology is traditionally, at least in the 20th century, be one that's like been quite hamstrung by different, you know, from, let's say, the Hidegarian School of Thought, which always seen saw technology as sort of the enemy of deep thinking as the thing that withdrew us or alienated us from being. I think the Galileo example that you show is like, no, it's actually through the alienation of technology of perception to getting outside our idea.

All the Centristian, and this is through technology that we know who we are, where we are, when we are, any of these kinds of things. So there's this epistemic role of technology, and now our fundamental technologies are computational, that we need to protect it all,

we need to protect at all cost. One of the, I should say, we are also, antiquated, there's also curating an exhibition at the Venice Architecture would be an alley this May. One of the artifacts that we're presenting, well, two artifacts I mentioned are presenting there.

One is an original copy of the Harmonia macrocosmica from 1662, that 17th century astronomical atlas. And another is an image from 1966 called the Lunar Orbiter image, which was the first fit picture of Earth taking a form. It was kind of forgotten these days, but it was on the front page of every newspaper in 1966, but Blue Marble kind of right, came and took the glory.

This was also an image that was shown to the German philosopher Heidegger, who was sort of a founding figure, 20th century philosophy of technology for mostly worse, in my mind. And he looked at this image and said, in his typically hyperbolic fashion, we don't need nuclear weapons because this image has already destroyed the world. And what he means by this is that the world of a kind of intuitive understanding of the horizon is flat, that I am located in this sort of egocentric, kind of, like, that everything operates according to a kind of a Trubian scale. The basis of a kind of phenomenological essentialism was now forever exploded, that innocence was gone. I could not, right, unsee it. It could not offering it.

I agree with that. I just see this more as like the point, a feature not a bug, let's say. And so I would invite your viewers to come to Venice and see it for themselves. Don't need many excuses to go out and see the original doge where the original doges were hanging out. Ben Vradden, so great to meet you, not to meet you, but to be with you in person again. And to hopefully be the first of many conversations. I would love that. Thank you, man.

Awesome.