Benjamin Bratton on Planetary Computation: Why Our Technology Has Outrun Our Concepts

Good evening and welcome to The Long Now Foundation. My name is Patrick Dow and I'm honored to introduce myself as the new board president of The Long Now Foundation and host for your Long Now Talks going forward.

Tonight's conversation could not be more timely. As humanity's technological capabilities advance at an unprecedented pace, we need new frameworks for understanding them at larger scales of both time and space. Benjamin Bratton, our speaker this evening, offers exactly these kinds of deeper, broader perspectives for us to consider. Benjamin's work challenges us to think beyond current debates on AI toward larger time scales, where the very nature of intelligence, life and technology may fundamentally transform our world. Please enjoy.

Well, thank you everyone. I want to begin also just saying what a pleasure it is to be able to share some of this work with you all at Long Now Foundation, an institution that I've long admired, no pun intended. So take a moment perhaps to take a step back and away from the persistent weirdness of our times, one in which we get free AI from a hedge fund and $200 a month AI from a nonprofit. Never would have called that one.

But more broadly speaking, if you take one idea away from this, it is that I genuinely think we are in a kind of pre-paradigmatic moment: that a lot of ideas are floating around that increasingly look kind of similar to one another and are coming together into something that may constitute a frame of reference that may be more useful to us. But this is a difficult process; it's one that we'll kind of have to invent. I'll put it this way: there's certain times in history when our ideas of what it is that we would like to do are way ahead of the technological capacity to do that, and there are other times when the technology is in essence ahead of our concepts. And I think this is probably more where we're at, and in those moments what we call philosophy, its job is to invent, try to conjure those concepts and bring them into being so they may be put to some use.

Unfortunately, I don't know that that's entirely what's going on. A lot of our most important epistemic institutions, such as universities, I fear to report, are a bit of a holding pattern, in that the sciences are focused on a kind of do now, think later mode, and the humanities, where I spend most of my time, it's more of a critique now and perhaps act later, but maybe not. In other words, this is exactly the wrong time to not be inventing those concepts and not be setting the initial conditions for the society to come, and yet here we are.

So let's begin. Our topic is computation. As you'll see, we sort of mean this in a somewhat idiosyncratic way. It's less to do with mathematics and algorithms, or with these kind of little appliances that we've constructed, but rather computation as a planetary phenomenon, not just something that humans do but indeed, through humans, something that the planet does. Our first proposition for the evening would be this: that computation was discovered as much as it was invented, and so we can think of natural computation, artificial computation. Stephen Wolfram recently published a paper in which he argues that the entire universe is a computational hypergraph, and that time is the rate of refresh, and dark energy is the heat exhaust of the big computation that is the universe. And I don't know, could be. I... sure.

Okay, so we think of computation in terms of planetary systems. This is not only sort of a new thing, it's also really where it comes from. Computation was born of cosmology, and I mean that in both senses of the term cosmology, which I'll talk a little bit about later. This is the picture of the Antikythera mechanism. It was, as far as we know, from about 200 BC. It is probably apocryphally understood as the first computer, but it was not only a calculation device, it was also an astronomical device. It was used to orient its user through the stars in relationship to her situation spatially but also temporally, allowing a kind of simulated movement back and forward in time. And so the idea that computation begins with this orientation of intelligence in relationship to its planetary condition seems to us a good starting point for this.

But computation quickly became something more than this, also a bit more practical than this. It also became calculation as a kind of world ordering. As societies became more complex, the necessity to calculate and compose not only what was happening right now but indeed what would happen in the past and what could happen in the future gave rise to forms of computation like this Sumerian cuneiform. You find this earliest form of writing, you try to imagine what could possibly be the first thoughts of humans put down in writing, and it turns out it's mostly receipts. And so in a way everything we've done since then, all written language, is sort of variations of accounting, which I like to remind my friends in the literature department.

So when we talk about a kind of school of thought, philosophy, where might we expect it to come from? Well, a few clues. One way of thinking of this is that, as I say, sciences are born when philosophy learns to ask the right questions. Most of the things we call sciences, at least historically, began as philosophy. And philosophies are born when technologies force the birth of new languages, which is, I hopefully, where we are. So what can happen in our present situation, one in which, as we might put it, the new things with which we are surrounded have outrun the available nouns that we have to contain them? This is Stanisław Lem, the Polish science fiction author, in terms of a distinction that Lem makes between what he called existential technologies and instrumental technologies. We think of instrumental technologies in terms of tools. Their main impact on society is what they do as a tool. So bulldozers move dirt very well, and so you can make cities faster. There are other kinds of technologies, much more rarefied, that when used properly change how we understand how the universe works.

Telescopes and microscopes are kind of obvious examples, but as I will argue, and as a kind of key thesis of our program and my work, computation very much is both, and that preserving the space for computation as an existential technology... Think about the role of the telescope for Galileo, and ultimately for the deduction of heliocentrism. Without this technological alienation of seeing the world, perceiving the world in a way that would be otherwise impossible, we really wouldn't know where we are. And so there is a fundamental relationship between technology and what Freud called Copernican traumas. Copernican traumas are those priceless moments, priceless accomplishments really, by which we deduce that the world, the universe, doesn't work quite the way it looks like it might work, and we decenter ourselves or get outside ourselves and figure out again who, what and where we are in some way.

Now the cycle of this is that because we have a complex model of the world and how the world works, we build technologies based on the implications of that model that would allow us to measure something or see something or perceive something or calculate something based on the logic of that model. But when we use that technology properly, we figure out that the model that made that technology possible is wrong, and there needs to be then a kind of resolution of the implications with the model that ultimately gave rise to them. This, in a more general framework, we might say is the role of technology more broadly.

Now we're speaking of this in sort of more historical terms, but the implications of this, of Antikythera and the forms of planetary computation as an existential technology, are actually quite pressing. I would make the argument that the scientific concept of climate change itself is an intellectual accomplishment of planetary computation. Without the sensors and satellites and oceanic temperature sensors and so forth, and ice core samples, and most importantly the supercomputing simulations of climate past, present and future, we wouldn't have been able to perceive these temporal dynamic transformations in planetary systems in which we are embedded. This, in a rough sense, is really what planetary computation is for.

And now this obviously then has not only scientific import but also philosophical and ethical import as well, because as we understand climate change, Crutzen and others came to reckon with the concept of the Anthropocene, problematic as it may be. But for sure the Anthropocene, to the extent to which it arrives from climate sciences, which is predicated on planetary computation, is a kind of second-order concept derived from planetary computation. It's a good example of how it is that computation as an existential technology can give rise to really fundamental shifts in our thinking and understanding of our agency as a species. In one way or another we are transforming the planet, and only through the deduction of this, by understanding and measuring how much we had artificialized the planet, did the possibility of recognizing agency become possible. But I think this is actually an important lesson. We tend to think about it in philosophy that first you train subjectivity and that this will give rise to better or other forms of agency. In many ways it often works the other way around: the subjectivity, and the possibility of a subjectivity as a planetary subject, only becomes possible once agency is mapped.

All right, let me speak a little bit about planetary computation, this term that we use, both what it is and what it's for. The Lunar Orbiter image from 1966: this is the first image of the Earth taken from the Moon, and it was on the cover of every newspaper in 1966. It's a little bit forgotten, but in 1966 it was quite a big deal, so much so that when Der Spiegel was interviewing the notorious German philosopher Martin Heidegger, they had presented him this image and asked him to comment upon it. Heidegger said that he was horrified by what he saw, literally shaken, and he said we don't need nuclear weapons to destroy the world because this image has destroyed the world already.

And what he meant was that a kind of intuitive, phenomenological, egocentric, perspectival understanding of the world and being, as something that is properly manifested by de-technologizing its relationship, has now been overwhelmed and overcome by this allocentric perspective. We can never quite believe that the world is the same way it was before, now that we understand a little bit more where we are. Now for us this is a feature; for him it was the biggest bug, I suppose, of all.

Now another little thought experiment though: imagine the Blue Marble to this image, 2018, the black hole image. An image, right, is a reconstruction from a terabytes-of-data set, but I would argue that this is actually just as important an image in terms of its existential implications as Blue Marble, and perhaps even more so. Part of it has to do with how the image was constructed. Now if you want to take an image of something very, very far away, in this case 50-some million light years away, you're going to need a very high resolution image. Resolution is dependent upon the aperture, the size of the camera that you're using. And so what would be the widest aperture camera you could possibly build on Earth? It would be one that's the same diameter as Earth itself. So the Event Horizon Telescope networked together several telescopes from the North Pole to the South Pole and linked them together into a kind of new optical sensory organ that, over a period of time, would even use the rotation of the planet as the kind of timing mechanism by which each of these would work. And so the planet itself not only grew this new sensory surface, but it even became a part of the machine.

The implication for us is that this locates humans, who obviously were the creatures responsible for the construction of the Event Horizon, in a rather different position than Blue Marble does. Blue Marble implied a global village by putting apex creationists in charge of a mythical garden. Black hole demands a different planetary regime by rendering humans as a privileged mediating residue that sets in motion further generalized cognition. The two worlds could not be more different. This is a new profile for us, and one that we'll take some time getting used to.

This is a quote from The Terraforming book. Imagine the Blue Marble image not as an image but in essence as a movie, a movie that spans the entire 4.7 billion year career of the Earth, but thankfully on a kind of super fast forward. What you would see is the Earth spinning, volcanoes, Pangeas. In the very last instance of this you would see something extraordinary: this little organism would sprout a kind of exoskeleton, this sensory epidermal exoskeleton of satellites and various other sort of mechanisms by which the surface of this organism can relay information from one point to another. This is where I want to sort of think about the location of planetary computation: it is not only something that humans do in their industry but indeed something that the planet has done. It needs to be understood as part of the evolution of the planet as a dynamic system.

Well, to sort of dive right into AI and in terms of our work: AI has co-evolved with the philosophy of AI quite closely. From Turing's and Searle's thought experiments, the thought experiments about AI have driven the technology, and of course in turn the technology drives the philosophy. There's a kind of double helix in the conjunction of these that you don't find in other sorts of ways. So another way of putting it is matter thinking about matter making matter that thinks. That's sort of what we're up to.

Now our position on this is something like this: that ultimately AI will teach us as much about what thinking is as we will teach it. To artificialize something is indeed to discover what it is. The lesson of climate science is that to artificialize a climate is in fact the way it becomes eventually possible to even know you have that agency. A more recent piece called The Five Stages of AI Grief is a map of the different ways in which the discourses around AI are finding different ways to be inadequate to the task: AI denial, anger, bargaining, depression and acceptance. You can sort of read it and just begin to think about it like, oh yeah, I know who's in this category. It's one of those things that kind of started as a bit of a joke, and then you realize actually those are usually the best ones.

Another sort of area of the positions we take on this has to do with the question of alignment, and a somewhat contrarian position on this is: alignment to what exactly? A lot of the discourse around AI alignment kind of presumes, as I've seen, a relatively naive image of what human needs, desires, values, ethics are, sometimes a very spoken, sometimes unspoken presumption that by simply amplifying unidirectionally the manifestation of those needs and desires, everything will sort of work out in the end. This strikes me as a kind of inversion of the Copernican implications of AI and a reversion to an unnecessary anthropocentrism and even anthropomorphism. One way to think about this is in terms of Turing's famous thought experiment, where what Turing had initially proposed was a kind of sufficient condition: that is, if the computer can fool player A, then at a functional level we have to sort of grant that there's something going on in there, and that's a sufficient condition. Unfortunately, in many ways I think, the Turing test as a metaphor has become more like a necessary condition: that is, unless the AI can perform thinking the way that humans think that humans think, it is disqualified. And this sort of over-normativism, or the reflection of the human as a model, I think it certainly

...centers too much of that AI. The alignment conversation also, I think a lot of the alignment discussion here, and again, I'm not really talking about, like, yes, if you ask AI to do something you want it to do that thing, and you don't want it to make chemical weapons. These are sort of unproblematic, and I'm not really arguing against these. It's the presumption that, like, really, that by making AI reflect human, exactly sort of human culture, human values, human dependencies, like what humans are most likely to do and think, as the kind of North Star for the artificial evolution of AI. It's like, have you ever met humans? Are you sure that that's kind of what you want?

Another basic, just sort of, what we call reflectionism. That is, there's a part of discourse where on the one hand you hear people say the problem is AI is not enough like humans and human decisions, and we need to bend it towards that. You have others who would basically say, like, no, AI is only the manifestation of the socioeconomic systems of power of human society, and that's the problem. So it's either exactly like us and that's the problem, or it's not at all like us and that's the problem, and somehow people will sort of say both at the same time, which is usually a sign that there's something a little bit afoot.

Now, if you hold that AI is an existential technology as a value that you would want to hold on to in terms of these conversations, that is, there are ways in which it will teach us what thinking is, ways in which it will disclose the workings of the world and the universe and our own dynamic processes, our agency, objectively to us in ways in which we cannot possibly have imagined yet, it means that, and once we discover these things it would transform our cosmology in the anthropological sense in important ways, it implies at the very least alignment needs to be bidirectional, both in terms of AI to where we want to direct it, and us to the outputs that it might have. So another way of putting it is that the presumption that the greater societal risk comes from not regulating AI may be quite wrong.

So just for the sake of argument, you could think of the area that we are exploring as a bit like this. We're familiar with the quadrant of less alignment, more bad; more alignment, more good. It's the less alignment, more good that we would like to account for, at the very least, not only because it's probably underexplored, but also because I think we make the case that alignment overfitting, making AIs really do exactly what people want in every sort of case, is actually kind of the real risk. The pinnacle of human-centered design, arguably, is the slot machine, a mechanism that does exactly what the human wants it to do, in a sort of wonder. Like, this is something to be avoided at all costs. And so the cultivation of what we call productive disalignment is what we see as sort of part of the agenda here as well. And it also, I think, has to allow for otherwise unpredictable cascades of causality, that the kinds of causal relationships between one thing and another in terms of those productive disalignments are not always kind of what you would expect.

Now, thinking of this in relationship a little bit to the DeepSeek news from last week, you might also think of some of these kinds of cascades. Cheap energy produces cheap complexity. This is sort of a truism of, like, the Santa Fe Institute, for example. Cheap complexity allows for cheap inference, cheap inference allows for cheap intelligence, cheap intelligence allows for cheap energy. These kinds of cascades are exactly what I mean by the kinds of productive disalignments to be protected. It's also one in which a sort of understanding of AI and its relationship to governance and the composition of society is not necessarily one that, you know, the question of control and agency is a little bit undecided. As we say, AI is less a tool of industrial policy than industrial policy is a tool of AI. But also in terms of the consolidation issues and where power lies with this, these are serious issues in terms of centralization and decentralization. Thinking about other kinds of means of production that have structured society, AI is one that's available for a monthly subscription. That may really be, that fact may really be the democratizing factor.

A couple of other productive disalignments I want to put on the table. There's been an enormous, interesting explosion of work and insights in non-human cognition, animal cognition, plant intelligence, and this is happening at the same time by which we are artificializing intelligence in a mineral substrate through AI. And we're beginning to see a similar kind of comparative non-human cognition discourse, in ways that are just beginning to come together, and sometimes rather explicitly sharing and mirroring these, in ways I think that'll be increasingly very important.

It'd be hard to get away from, in the last few months, the claim that next year it's going to be all about agential AI, that it's all about agents that will be making things for us, and also the kind of consensus prediction that AGI will appear somewhere around 2027. So you put these two things together and you have a potentially very complicated scenario by which, if you have an explosion of AI agents that are roughly AGI level, whatever you take that to mean, you can imagine a scenario by which you may have 8 billion human-level minds that are human and 80 billion human-level minds that are not human, a ratio of 10 to one, and later perhaps 100 to one, a thousand to one. And in those cases, what even constitutes a society kind of goes back to first principles.

So for us, it is these sorts of edges, what we call the weirdness right in front of us, from which we try to gain some sort of insights. I want to make the case to you that technology literally evolves, and it does so in ways in which it's never really ever just a tool. All technologies are built out of earlier technologies. There's a kind of scaffolding process by which one technology becomes the scaffold by which a yet more complex technology develops. Even thinking more in terms of anthropogeny, that with the beginning of humans there's always been a coupling between biogenesis and technogenesis. Literally the shape of our anatomy, like opposable thumbs, for example, is a kind of imprint of earlier forms of making use of the world for directed kinds of purposes, and this structural deepening becomes more complex over time and ultimately becomes a way of mapping evolutionary time.

So again, when we say evolves, what do you mean by evolves? How one, via its component scaffolding towards more complexity, a complex thing emerges, it becomes a component of something yet more complex, which becomes a component of something yet more complex. We see adaptation and exaptation, that not only are there niches into which certain kinds of technologies were fit, but also technologies that were designed for one purpose then become very useful for completely other kinds of purposes. Nevertheless, they become components for things that are well beyond their original intention, just like biological adaptations. And looking at the maps, we see both convergent and divergent path dependencies in terms of the directionality by which all these things may be moving. Now, everything that I've described in terms of technological evolution is also true of biological evolution, which is kind of the interesting fact. Or as V neor puts it in one of our works, what if humans are a phase in the history of technology? All right.

Now, evolution of computation, or evolution as computation, to move this a little bit further along. Computation as a kind of technology is itself evolving. We can sort of map forms of intelligence, complex cognitive intelligence, in this way. This object, this is not my refrigerator, by the way, don't worry, this complex object and the prefrontal cortex that wraps around it is one of the most remarkable accomplishments of biological evolution, to produce an object that is capable of these feats of predictive information processing. But in other words, the planet, over long periods of time, folded itself in such a way to produce this object by which it ultimately came to deduce things about itself. But now the substrate of complex intelligence includes both the biosphere and the lithosphere. We, the fire apes, by folding bits of metal and rock and running electric currents through it, figured out how to make the rocks think. This is news. It's also then part of not only our evolutionary trajectory, it is also part of the evolutionary trajectory of the rocks.

Species that are good at artificialization do well. It's part of what evolution selects for. If a species is good at finding ways of building technologies, for example, that allow that species to capture more energy, information and matter by artificializing its environment, that population can grow. The capacity for artificialization is an adaptive process. Another way of thinking of this is the distinction between autopoiesis and allopoiesis. So autopoiesis, we learn from cybernetics, is how it is that a system uses the external environment to reproduce itself. Allopoiesis is the way in which that agent uses the external environment to produce something extrinsic to itself.

Let me make the case then here that would try to locate AI within this a little bit more explicitly and give a bit of a timeline: that AI is actually the artificialization of artificialization itself. One of the arguments that Sarah Walker is going to make to you when she comes to speak is that natural selection doesn't begin with biology, it actually begins with chemistry, that certain kinds of molecules are stable and are able to reproduce each other. So the capacity for selection and evolution ultimately stabilizes into certain forms of life, that is, entities that are capable of autopoiesis and the internalization of energy, information and matter for reproduction of themselves. To get really good at life, it is selected for to get really good at artificialization. Like, in order to do autopoiesis you need to get really good at allopoiesis, and the better you get at allopoiesis, the better you will be at autopoiesis. By making things that allow you to capture more information and matter that's extrinsic, you can therefore take more that is intrinsic.

Now, to get really good at artificialization, you kind of have to be smart about it, not only individually but collectively. You need to be able to imagine future possible states and mechanisms by which you can do allopoiesis. And so to get really good at artificialization, you need to evolve intelligence. This, rightly, I'm not really proposing this as kind of like the new scientific method; just imagine I'm drawing you a napkin sketch, okay? So that's really how this is meant. Now, to get really good at intelligence, you need to be able to communicate abstract ideas between the nodes in the intelligence system, that is, each of the individuals. And so to get really good at intelligence, something like symbolic language becomes very, very useful. And again, each of these things, like the ability to do one, sort of feeds back on the ability of the other. Now, to get really good at symbolic language, and to use symbolic language as a way to accelerate and collectivize, to make into a kind of generative dynamic, the artificialization of intelligence itself becomes quite useful.

So there's a kind of cascading sequence. Again, each one of these things goes in a particular order, and even the accomplishments of one in essence become scaffolds for the other. I would also say that there's a kind of feedback loop here as well. Like, once you get good at this one, it actually changes how that one works. So once artificialization becomes robust, it actually changes the dynamics of symbolic language. Inevitably, over the next few years, it will transform the kinds of symbolic languages that we speak and work with and that exist. And in turn, the kinds of symbolic languages that we have have transformed the kinds of intelligence that we work with; we think in languages and ways we may not have before. And indeed, this sort of recursive feedback would likely continue, and such, back to here, that the ways in which AI changes the language will change the forms of intelligence, will ultimately change the capacities for artificialization itself. This is what I mean by cheap intelligence equals cheap energy in that way as well. So this is what I mean by AI is actually the artificialization of artificialization.

Now, locating it then in sort of past, present and future, and where to situate this present point, it's important to remember that this is not the end of the cycle. As life becomes a scaffold for autopoiesis, autopoiesis becomes a scaffold for allopoiesis, becomes a scaffold for ultimately symbolic language, so forth and so on. What is all of this a scaffold for? And in turn, what is that a scaffold for? And in turn, what is that a scaffold for? And we'll have to wait for the end of the 10,000 years in the future. We have some idea, but this is a way to sort of map in advance a little of what this might mean.

Now, when I mentioned earlier that I think we're in a kind of pre-paradigmatic moment, let me give you one example of that, and that has to do with both functional and intellectual definitions of life, intelligence and technology. One of the things you notice, when you spend a lot of time thinking about this, is that increasingly contemporary definitions of life, as a kind of autopoietic, allopoietic phenomenon that uses predictive modeling to use the environment recursively, kind of look a lot like our most contemporary definitions of intelligence, which kind of look a lot like the more evolutionary theories of technology. They're beginning to look a lot like each other, and there's something to that. Like technology and like intelligence, life is based on evolutionary scaffolds built on past scaffolds, in which itself is a scaffold for forms to come. If both life and technology are not just kinds of matter, categories of matter in the Rylean sense, but rather processes that produce kinds of matter, then in what ways and at what level of abstraction are they in fact the same process? I don't know. We might presume it's a way that any sufficiently advanced technology is indistinguishable from life, and perhaps vice versa. We'll find out.

Last two bits I'd like to share with you this evening. One is a very quick run-through of the Antikythera program. Its rationale is that, as I've already sort of suggested to you, there's a rather potentially disastrous gap between our capabilities and our concepts. New epistemic institutions, forms of epistemic institutions, are needed in order to try to develop what that vocabulary would be. We're incubated by the Berggruen Institute, which for a decade or more has supported cutting-edge thinking in philosophy and politics. Some of the things we do: we host conferences. Just a few weeks ago we were at MIT Media Lab, where we brought together a lot of our key researchers over the course of the whole day. We host salons, which are shorter one-day, very intensive discussions with like-minded confederates in different cities.

The main area, though, arguably one of the more important areas of the program, is the studio. The key idea here is, like, well, architecture as a discipline has benefited from this exploratory, experimental studio culture. To the extent that society now asks of software things that it used to ask of architecture, the organization of people in space and time, software needs a similar kind of studio space to organize and do projects from first principles. Our other sort of major initiative is a new partnership with MIT Press, a book series and a journal. What Is Life? has come out; What Is Intelligence? will be the first major title. We also have a peer-review

journal. The impetus behind the Journal is to pair some of the most interesting thinkers with some of the most interesting designers and to develop definitive versions of those ideas in a context by which we can take some of this visual intelligence and include it as part of the ways in which we tell the story. And also, again, exhibitions, so we will be at Palazzo Diedo in May with an exhibition showing a lot of the work.

Antikythera is very much a collaboration. We are astrophysicists, zoologists; we have faculty as part of our network from all of the major universities and companies, and if you want to see the full roster of the people we got to work with, you can also see our site.

Okay, last point. One of the issues: you sort of imagine the way in which the future was understood in the 20th century was something to be accomplished. If things come together, we will have accomplished this idea of the future. My son is 16, and watching the ways in which the future has been part of the pedagogy that he's been exposed to, it's presented more as something to be prevented. When we talk about the year 2050 in IPCC reports, it's how do we make sure that future doesn't happen, which there's some validity to, but I also think it's important to think about the ways in which another future faces forward.

I just want to leave with this in terms of the situation being remedied. One is, as I have intimated a few times, there is a kind of dire disconnect at present between cosmology in the astronomic sense and cosmology in the anthropological sense. When I talk to my friends in astrophysics about cosmology, it's about black holes and dark energy. When I talk to my friends in the anthropology department, it's about how cultures understand their position and significance. Traditionally, as we imagine it, they sort of go in lockstep. At this point there's, as I say, a kind of dangerous disconnect. We know so much more about how the universe works than has been absorbed by the intelligence of our cultures.

Now, one way to think of this is that the science needs to figure out how to bend to the dispositions of the culture. Another is, in essence, that we need to update the culture to what it is that we actually know to be true. You might suspect I recommend the latter.

So where really are we, then? Well, you could put it this way: lithosphere makes a biosphere that makes a technosphere that is now part of the noosphere. There you go, that's the history of the Earth in 15 words. That's the where and when, that we might say where we are in terms of where the agency constitutes going forward.

The argument I'm making is not one of mastery. It's not one of total control. It's not one of, if we maximize intelligence, that the normative decisions about what needs to happen, or even the controllability of the outcome of the cascades of those conditions, is something we can entirely control. But nevertheless, it's something where, in essence, it's not optional, because humans are doomed to compose their own evolution and blessed to never truly understand or control that process.

So back to the issue we wrestle with: what are the conditions by which complex planetary intelligence could exist and grow and thrive for that 10,000-year span? What in essence would make it adaptive? You could think of it this way: in the short term, complex intelligence, as I showed you on the timeline, is very, very evolutionarily adaptive. It allows for species such as ourselves to do things that would otherwise be impossible. But the situation we're facing now is that there may be a point by which the continuance of complex intelligence in the form that it has taken may be the thing that ultimately undermines the possibility of the continuance of complex intelligence. It may be maladaptive in the long term, which is an idea that many, including Arthur C. Clarke, had posited.

So the question to be posed is: what are the preconditions for the long-term adaptiveness? What would have to happen, what would need to be worked, what would be the preconditions by which that long-term capability would be possible? This is where our work and the Long Now Foundation's work so profoundly overlap.

One way, I think, to start is to understand that those preconditions are ones that will have to be artificially realized for the most part. They may be ones to be discovered, but they also may be ones that need to be brought into existence. And understanding both the necessity of bringing them into existence and the impossibility of controlling the implications of bringing them into existence is traumatic as an idea, but this may be really the most important Copernican trauma at hand.

Let me leave you with one piece that I'll read here as a way to tie this up, and then I appreciate your patience.

James Lovelock knew that he was dying when he wrote his last book, Novacene: The Coming Age of Hyperintelligence, and he concludes his own personal life's work with a chapter that must startle some of the more mystically minded admirers of Gaia. He calmly reports that Earth life as we know it may be giving way to abiotic forms of life and intelligence, and that, as far as he's concerned, that's just fine. He tells us quite directly that he's happy to sign off from this mortal coil knowing that the era of the human substrate for computational intelligence may be giving way to something else, not as transcendence, not as magic, not as leveling up, but simply as a phase shift in the very same ongoing process of selection, complexification and aggregation that is life, that is us.

Part of what made Lovelock at peace with this conclusion is, I think, that whatever the AI Copernican trauma means, it does not mean that humans are irrelevant, are replaceable, or are at war with their own creations. Advanced machine intelligence does not suggest our extinction, neither as noble abdication nor as bugs screaming into the void. It does mean, however, that human intelligence is not what human intelligence thought it was all this time. It is both something we possess but which possesses us even more. It exists not in individual brains but even more so in the durable structures of communication between them, for example, in the form of language.

Like life, intelligence is modular, flexible and scalar, extending to the ingenious work of subcellular living machines and through the depths of evolutionary time. It also extends to much larger aggregations of which each of us is a part and also of which each of us is an instance. There's no reason to believe that the story would or should end with us. Eschatology is useless. The evolution of intelligence does not peak with one terraforming species of nomadic primates. This, I think, is the happiest news possible, and so, like Lovelock, grief is not what I feel.

Let me end there. Thank you.

Thank you, thank you for that incredible presentation.

Thank you, my pleasure.

Benjamin, we were so excited to invite you to give the first talk of our first quarter century, because your work with building this new school of thought is truly a long-term project, and I want to ask you some questions about building a school of thought in this era. In particular, what do you see as the role of human thought and machine thought, and how might the interplay of that evolve over the next 25 years as your project continues to grow and blossom?

That's a big question. Let me answer it rather directly and honestly. This term school of thought is something, when people ask us what the Antikythera program is about, what it tries to accomplish, the answer I will sometimes give is that we want to establish a new school of thought for this pre-paradigmatic moment, one that would not necessarily have all the answers but would change the kinds of questions that are asked, such that the right, better answers are more likely to emerge.

And a lot of that has to do, as I said, with the role of generating the philosophy from the direct encounter with the technology rather than projecting the philosophy onto the technology, of which we have a lot, from certain forms of an AI ethics discourse, or what would Kant think about driverless cars. Much more important is to essentially invent the concepts bottom-up from the technology itself and, in doing so, constitute this language. And I think it's a goal for the program: if we can get people speaking our language and using our language to define the problem space, then even if they disagree with us, we win.

Now, in terms of the second part of your question, about this alignment of human thought and machine thought: I guess I don't really have a really strong dichotomous thinking of this. I kind of take as a given that in anthropogenesis, how humans became human, and technogenesis, how technologies evolve, there's been a deep coupling of these going deep in time. And to think of human thought as something that's separate from the technologies of thought is probably the wrong starting point. So it's not so much that you've got human thought here, machine thought here, and what happens when they come together. It's more like, now that they are coming together in this very amazing and explicit way, how does this change our understanding of that long-term trajectory of how we got here? In essence, it forces us to rewrite history, but then perhaps gives a little bit of an arc of this going forward.

In general, I would say I learned enough structuralism and post-structuralism as a graduate student to presume that we speak language, but more so language speaks us. And the fact that language turned out to be a repository of intelligence, that if you can model language you have this general-purpose capacity for intelligence to be derived from it, why it was language that turned out to be the trick, not games, for example (DeepMind had a big bet on games; it turned out to be language), shouldn't be so surprising. In other words, we've constructed this language, the language is the model by which the language models are working. It's all tied together.

It's interesting what you mention about language and language models, because we're living in a moment where people all around the world are discovering the magic of generative AI for creating language and, in many instances, finding that they can outsource their thinking to these tools. So people are thinking less and less, and at the same time you are making space and time with your school of thought to think and create new ideas, which seems novel in the context of these LLMs being able to kind of see and think of everything. So what is the value of thinking now, and how do you conceive of that as a leader?

Am I a leader? Yeah. I don't know that people are thinking less. I don't know if I agree with that. I'll speak to generative AI in a moment, but when you were talking, it reminded me of one of the dialogues of Plato, the Phaedrus dialogue, I think it's the one, where Socrates is critiquing this newfangled technology called writing. And he was really dismayed at this because he thought it would destroy our capacity for memory, because you're just outsourcing memory to this sort of thing, as well as that you couldn't have a direct Socratic dialogue with the person who wrote this, and so you're susceptible to all kinds of deception. And even more, he was concerned that you could actually communicate with dead people, which was horrifying for him.

And so the term he used to describe this new thing, which was both amazing and horrible at the same time, was pharmakon, from which the term pharmacy comes. And what pharmakon means is something that is both remedy and poison at the same time. It's not that we'll figure out whether it's remedy or poison at some later date; it will always be both. For sure, AI is pharmakon.

And now for generative AI, I think the discussion around generative AI needs reframing. A lot of this is, I think, very short-term thinking. More generally, imagine that everything you do from when you wake up in the morning to when you go to sleep is training data for the future's model of the past. It's a big responsibility that you have. It's a bit of living in the third person, I suppose, but it demonstrates a way to think about generative AI not as a kind of instrument or tool but rather as a way in which collective intelligence is modeling itself over the long term.

One of the terms that you used that I found really compelling was extrinsic philosophy, and this being related to allocentric as opposed to anthropocentric. What are some of the ways that you could see culture being updated, as you put it, to reflect allocentrism instead of anthropocentrism?

I mean, maybe the first thing, again I'm thinking in terms of my son, I think the issue is we know so much more, in a way that hasn't really percolated in. There are so many ways to think about how this would work, even just very obvious things like pedagogy. I find it completely amazing, bizarre but amazing, that for the most part (UCSD is a bit of an exception, Pittsburgh too) most philosophy departments don't teach any neuroscience. Why would you have a whole department about how it is that we think and how it is we might think, and completely ignore what we know about how we think and how we might think?

For example, I think you kind of look around and you find more of these sorts of things, just looking at, you see my point, how we teach astronomy, how we teach neuroscience, how we teach genomics. At my son's high school, they're forbidden to use AI. I remember when I went to high school in the 14th century that if you used a calculator in class you got in trouble, and now if you don't use a calculator in class you get in trouble. But the end result of putting calculators in the calculus class means that students are taking calculus a year earlier than they were before, because they don't do this busywork with the arithmetic; they can focus on the concepts. And so I don't think they're thinking less.

Now, the question, I think, for AI is: okay, given this certain degree of inevitability, what's the analogy of that for all the things we teach the next generation? What's the putting-the-calculator-in-the-class version of this? And I also happen to be on a University of California committee for what should be the policy for AI in the classroom at the UC, and this is also a committee where, when I spend time in there, I feel like pulling my hair out, because most of the discussion is what are the ways in which we can prevent and forbid this, as if that's even possible.

The way I see it, as a teacher, when I teach my undergraduates, it's up to me to presume they're using the large language models and to build better assignments. With the types of things people are capable of doing with these kinds of tools, we need to rethink this. So there are bigger-picture ways of answering your question, and I think there are other ways in which a lot of it's just right in front of us.

I'm curious why you have constructed your program with Antikythera in the way you have, instead of within a traditional university context, for instance, and how is this working for you?

Yeah, I mean, it's not entirely outside. We're a bit of a kind of pirate ship that kind of

moves sort of in and out of different ports in different ways and kind of can shuttle things, hopefully not rats, from one port to another. But you know, I am in academia, right? I'm a professor at University of California San Diego. You know, it's a thing to be, but it would be absolutely impossible to answer the question, it would be impossible. There's just no way. There's no economic format for doing the kind of work that we do within the university. I tried.

I'm curious, what would be your response to our community member Jessie Kate's question? She asks: as an archetypical pattern of relative association, how might you read institutions through the lens of planetary computation, and what could be the future of institutions?

It's a good question. Yeah, I mean, the value of institutions to a certain extent is their durability, that there are ways in which people can come together to construct a system for the solving of particular kinds of problems, the cultivation of particular kinds of questions, that grows over time, and that every iteration by which that institution runs its cycle, it evolves. But it evolves in a certain way that it also has scaffolds internal to itself, so it does something that becomes a component to something it does later, that becomes a component of something it does later that's more complex. That is an evolution, and that takes time. That takes time.

And so yeah, I don't think it's so complicated. In a world in which things seem to be liquefying at such a pace, pushing against that with things that have a bit more durability remains sort of important. But there's different kinds of institutions. I would extend the institution not only to be things like institutions that are built of humans and institutions that have boards of directors, but there are technical institutions, there are forms of structure that operate in a very similar kind of way.

We were in the lobby before talking about the metric system as a kind of platform. What this allows you to do is you don't have to decide how wide things are going to be. It becomes a way in which you don't have to do that work, so that you can get on with it, one way or another. And I think the best kinds of institutions are ones that not only are doing the work, but they're doing the work so that everyone else can get on with it.

Well, Benjamin, thank you so much for being with us tonight to introduce Antikythera to our community. We look forward to collaborating with you and cheering you and the school of thought on over the next quarter century. Thank you, Ben.

Appreciate it very much. Thank you.