Can Humans Stay Smart in the Age of AI? (David Krakauer, President of the Santa Fe Institute)

[00:00] You've spoken before about your sort of concern that LLMs act as such a tempting mental prosthetic for people that if we're not careful, and perhaps even if we are careful, can't help but allow it to atrophy our own cognition. I think it's really much more dangerous than people realize. You just have to look at the opioid crisis or the sugar crisis. These are all essentially about easy [00:30] or materials that are addictive. I'm concerned that LLMs in their current incarnation are like sugar cocaine. The optimistic part of me hopes that we're going to see people find ways to use LLMs and AI in a more complementary way. What is a thinking person's best course of action? [01:00] here. [01:00] It's just not evenly distributed. [01:02] Each week, I sit down with the founders, investors, researchers, and visionaries who [01:07] who are living in the future. [01:09] to help you see what's coming next and understand it more clearly. [01:13] Today I'm speaking with David Krakauer. [01:15] a leading complex systems researcher and the president of the Santa Fe Institute. [01:20] David leads one of the world's most unique institutions. It's home to a dizzying, [01:26] interdisciplinary collection of scientists, artists, and thinkers.

[01:30] united by the mandate to search for order in the complexity of evolving worlds. [01:35] In our conversation, we explore why large language models may not be intelligent at all. [01:41] the risks of AI usage destroying our cognitive capabilities... [01:46] The surprising qualities of great minds [01:49] and what David's friendship with legendary novelist Cormac McCarthy... [01:52] taught him about the single-minded pursuit of excellence, [01:56] and the cultivation [01:57] of unconscious creativity. I walked away from this discussion fizzing with new ideas about the nature of intelligence, the unique representational value of literature, [02:07] and how each of us must work to maintain our human agency. If you liked today's discussion, [02:12] I hope you'll consider subscribing and joining us for some of the incredible episodes we have coming up. [02:17] Now, [02:18] Here's my conversation. [02:20] with David Krakauer. [02:21] This episode is brought to you by Brex. Fred Adler, the influential venture capitalist of the 1970s, [02:28] was known for displaying decorative pillows in his office that featured a signature business philosophy. [02:34] Corporate happiness is positive cash flow. [02:37] In today's post-SERP environment, Adler's wisdom feels particularly relevant. [02:42] as founders need to make every dollar work harder. [02:45] That's exactly what Brex delivers. [02:47] Their modern finance platform was built specifically [02:50] for startups like yours. [02:52] and designed to help extend your runway when capital efficiency matters most. [02:57] With Brex, you get global corporate cards with up to 20x higher credit limits

[03:01] and no personal guarantee required. [03:04] Their banking solution has no minimums and no transaction fees. [03:08] while letting you earn high yield from day one. [03:10] with same-day liquidity. [03:12] Best of all, Brex knows you were born to build, not juggle spreadsheets and finance tools. [03:18] Their AI-powered platform brings cards, banking, [03:21] expense management, and travel all in one place. [03:25] It's simple. [03:26] Scalable. [03:27] and designed to get you back to what you do best. [03:30] building. [03:31] More than 30,000 companies, including one in three U.S. venture-backed startups, trust Brex to help make every dollar count toward their mission. Join them at brex.com slash mario. [03:45] This episode is brought to you by Interpret. Interpret is a customer intelligence platform used by leading CX and product orgs like Canva, Notion, Perplexity, Strava, Hinge, and Linear. [03:55] to leverage the voice of the customer and build best-in-class products. [03:59] Interpret unifies all customer conversations in real time, from gong recordings to Zendesk tickets to Twitter threads, and makes it available for your team for analysis and action. [04:09] What makes Interpret unique is its ability to build and update a customer-specific knowledge graph [04:14] that provides the most granular and accurate categorization of all customer feedback [04:19] and connects that customer feedback to critical metrics like revenue and CSAT. [04:23] If modernizing your voice of customer program is a 2025 priority, like it is for customer-centric companies such as Canva, [04:30] Notion, perplexity, and linear, reach out to the team at interpret.com slash Mario.

[04:36] That's E-N-T-E-R-P-R-E-T dot com slash Mario. [04:41] I must say, I've been really looking forward to this. I followed your career from a distance for a long time and some of your writing and and [04:50] public thinking. I can't say that I can fully grasp all of it, but I've certainly enjoyed it. And maybe the place to start is with a brief introduction about [05:01] the Santa Fe Institute of which you're the president. For folks who might not have heard of it or really just heard of it in passing, how would you describe SFI's mandate and place in the [05:17] is in Santa Fe, New Mexico. [05:19] He will ask. So I have to say that. And so the, um, [05:24] up on a mountain. We have two campuses. We were founded in 1984. [05:30] Thank you. [05:32] essentially to study complex systems. [05:35] And, um... [05:38] Our mission statement is [05:40] is searching for order in the complexity of evolving worlds. Okay. So it gives you a sense of the modesty of the enterprise. We're not particularly interested in disciplines. It's not how we think. One way to think about it is we're principles first. [05:58] So, [05:59] Collective intelligence, scaling, robustness, that sort of thing, information, computation. So we make those the central hubs.

[06:09] and then disciplinary [06:12] Domains become spokes. [06:15] And the wonderful thing about that is you get these very natural affinities being established between researchers from very different fields. [06:24] And so in a way, [06:26] were on the one hand very foundational, [06:30] new principles for the complex world, we can get into that. [06:34] And [06:35] but we're also interested in the systems of the world. [06:39] So how does the economy touch the climate? How does that touch psychology? And so on. So in our senses that if there's anything that's changed in the 20th century, [06:51] It's the density of connections between the systems of the world, or among the systems of the world. [06:56] And so how do we think about that in a principled way as opposed to just bloviate [07:01] or have opinions without a rigorous foundation. I think the dividend of that is that SFI seems to have collected and sort of moved through this rotating cast, as you said, of some of these really fascinating interdisciplinary minds. And in reading more about the history of SFI, I was reading how one of the founding fathers and legendary physicist Murray Gell-Mann [07:31] Babylonian figures. And on the one hand, I know what he means, but on the other hand, I'm curious how that really shows up. And so I wonder, how do you think about what that sort of Odyssean figure looks like, especially in the modern world?

[07:44] Yeah, so that's interesting. There's a long tradition going back to classical antiquity of creating [07:53] binary oppositions to establish [07:56] if you like, forms of disposition. The most famous comes from Archaeolocus, and it was made famous by Isaiah Berlin. That's the fox and the hedgehog, right? That's what most people know. Sort of the hedgehog is interested in one big idea and pursues it. The fox is restless and pursues multiple ideas, but somehow with less depth. [08:20] Both are important, right? And he was interested in Dostoevsky and Tolstoy. That was the contrast that he was making, where in that case, Tolstoy was the fogs. The other one that's famous, and I'm getting there, is Aristophanes. [08:34] the birds and the frogs, and that was made famous by the physicist Freeman Dyson, [08:39] and in particular as a means of classifying mathematicians. [08:44] And the frog has a very aerial view and everything looks vast. [08:48] and they pursue problems right in front of their face. They're very difficult problems, whereas the bird has a synoptic view of mathematics as a whole, an aerial view. [08:59] And you know, you can think about [09:00] landlands as maybe bird-like and coxeter as maybe frog-like. Both extraordinary. So I think Murray was a bit envious and wanted his own version. So he stole from Friedrich Nietzsche. In Nietzsche, in his first book, The Birth of Tragedy, Nietzsche talks about the Dionysian and the Apollonian, the interest in chaos on the one hand and the interest in order, the timeless and the timeless

[09:30] the immortal versus the mortal. And so he said, [09:35] If in some sense complexity is about the balance of the order and disorder, what comes in the middle? What's between Apollo and Dionysus? He sort of in some sense, [09:46] fused Hesiod and Homer and said, well, let's take Odysseus, who's kind of in another domain to Apollo and Dionysus. And the idea of the Odyssean is... [09:59] the complex, which is the [10:01] the world that is ordered and disordered in equal measure, [10:04] And Odysseus is coping with that world, the world of the gods, the classical world, the world of order and the world of humans, the world of disorder. [10:14] He felt that would be a good model for the kind of scientist he wanted. [10:19] on this long sort of hopeless voyage back [10:24] to Penelope, where in this case Penelope stands for truth. And what that turns into in practice is my problem. Wow. So interesting. I'm so glad I asked that. So much of SFI seems to have these tendrils with Los Alamos and the history of that group of scientists, with the history of the desert. [10:54] because there's sort of these, you know, this obvious intellectual transference maybe that happens. But when we're talking about things like order and disorder, I imagine even at this like

[11:04] there are connections there. [11:06] Yeah, I mean, it's a whole part of our... [11:11] mythic foundation. [11:13] Um... [11:14] On the one hand, yes, that we're born from. [11:18] the most destructive force ever invented by humanity. Yes. In the Manhattan Project, that's significant. So that as backdrop is important to us. [11:31] you know, [11:32] Talk about creative destruction. [11:34] There you go. There you go. And so, but of course, that project was supported by an extraordinary cast of minds. And so to the extent [11:45] that [11:48] independent almost of the particular project that they were working on. If you look at that cast of extraordinary, mainly Hungarian, immigrants, [11:58] you know, [11:59] Wigner, Szilard, Uhlamp, von Neumann. That mindset, super polymathic, unbelievably rigorous, very historically rooted, [12:11] aware that there was an issue in the world they had to apply their mind to that was important at the time of course at that time it was world war ii that conjunction of a sense of importance [12:25] and a belief that your rigorous ideas that seem detached [12:29] from application were important has always informed our sensibility. So

[12:35] both dimensions that the [12:38] the grandeur and scope of the project, and the kinds of minds those projects require. [12:44] Hmm. [12:45] In your role as president, I imagine you're sort of [12:49] have to play the part a bit of a talent spotter or perhaps a talent wrangler. Have you come up with your own heuristics of what [12:58] an Odyssean looks like or a bird or a frog or something else that you think, "I'm able to turn over all the shells on the beach and find the one that is inflected in just this magical way that really fits us." No. I mean, although there are characteristics, there are characteristics. And, [13:23] So, [13:24] A deep appreciation for expertise. [13:28] We all know that when we see it. If you become expert, you spot experts. That's what you're good at. That's true in sport, it's true in mathematics, it's true in music. If you're a good violinist, you know what a good violinist looks like. It takes an expert to spot an expert often, not always, but often. That's one. And that has all sorts of sequelae, what that looks like, what that feels like. And the other one is a kind of interesting one, which is, [13:57] Analogical reasoning. [13:59] Thank you. [13:59] I feel very often talking, you know, that I've spent too much time in a bath of Nobel laureates. And one of the things that I...

[14:09] Note is... [14:12] Oh, this is like that. Oh, let me explain. This feels like that. That desire to communicate via structural analogy seems very strong as an impulse. [14:22] It's not always illuminating, right, and it's not necessarily always exact. [14:27] But... [14:28] I think there's this kind of looking at the Rubik's Cube from many different angles. [14:34] and rotating it into new configurations. [14:38] seems to be something. That's an empirical observation, really, more than anything else. [14:43] Wow, how interesting. [14:45] one of the members of the Institute, I don't know if you call them members or not, but one of the cast of characters was the great [14:55] writer Cormac McCarthy. And in your essay that you wrote about him a few years ago, you talk about how he sort of almost considered some degree of inexperience important in generating novel ideas. And I think you also at some point cite Montaigne and how he talks about the minds of critics being bogged down with sort of almost too much context. And when you talk just now about sort of [15:20] expertise, experts recognizing other experts. I wonder how you find people who both have expertise, but somehow have enough of that ingenuity or creativity to not be [15:32] sort of, uh, [15:34] bogged down by it. So in relation to core marking in particular, [15:39] He was very interested in

[15:42] Naivety. [15:44] as a principle. [15:45] and he was very interested in [15:48] Ancient Greece. [15:49] and [15:51] their belief that everyone else was a barbarian. [15:54] Um, [15:55] that interested in, what did that, why was that? [15:59] and their enormous [16:00] naive belief in the excellence of their understanding. So there's a kind of juvenile characteristics to antiquity that attracted him. And I think it's the same... [16:15] Drakshire as his interest in [16:18] early quantum mechanics, where a lot of very young gifted minds [16:23] you know, Heisenberg, Schrodinger, all these, de Broglie, all these people, [16:28] I think he was drawn to the same thing. So when we talk about Montaigne and... [16:32] and those interests of Cormat, I think he associated that with a certain attitude towards [16:39] being limitless [16:42] and that knowledge would become a hindrance to your life. [16:48] Intellectual ambition. [16:50] That's one part of it. I think that's very strong in him. And I think it just... [16:54] it recurs throughout his interests. And I guess we'll talk about that concern at some point, but it's there in him too. It just keeps coming up over and over again. [17:03] I guess it's kind of a profile of... [17:06] combination of a sort of a extraordinarily wise [17:10] infant. I mean, it's just weird combinations. And you feel it in his novels, right? I mean, many of his characters have that character too. And I think-

[17:19] Um, [17:20] Now, with respect to the other question of being bogged down in knowledge, I mean, it's [17:24] You know, there's this quote I quite like, which is that... [17:29] you can become knowledgeable [17:32] with other people's knowledge. [17:34] but you cannot become wise with other people's wisdom. [17:38] And I don't think you can be too knowledgeable. And I think it comes down to how you use it. So here's an, here's an analogy for you. So me, [17:49] I can give you a building. [17:51] Yes. [17:51] I'll give you... Gaudis. [17:55] beautiful and cerebral familiar. [17:57] And, [17:58] You look at it, you dissect it, [18:01] you destroy it in your mind, not in reality, in your mind. And in order to reconstruct for yourself an edifice, [18:10] And for me, [18:14] Thoughtful people treat knowledge that way. [18:17] they treat it as [18:19] a composite [18:21] to be understood in terms of its parts and its rules of construction. [18:27] People who are not so thoughtful just take on the structure wholesale. [18:32] And that has the character of fact, not of knowledge. This is what [18:39] Such and such says. It often manifests as an appeal to authority. [18:45] And I feel as if the right way to use knowledge is to disrespect it.

[18:51] and respected, right? Respected by disrespecting it in the most profound way, by using it as a means of constructing your own. And I think one of, again, one of the problems of the current moment actually with LLMs and so on, is that it's really knowledge as fact, [19:09] And not knowledge as system, which is somehow a different way to think. [19:14] Yes, in a way, to disrespect something properly, you at least have to understand it. Exactly. Rather than just repeat it. Well, I definitely want to talk more about LLMs, but I'd love to sort of discuss... [19:29] your specific discipline first a bit more, which is complexity and complexity science. And in trying to ground myself at least a little bit in understanding what that is, I sort of went through your colleague, Melanie Mitchell's, has sort of an e-learning experience about the introductions to complexity science. And I was gratified to discover that [19:49] It's very hard to explain even to people who are, you know, even from practitioners, you know, many of the different experts she talks to offer different explanations of what this discipline really is. And so how do you describe your work? So it depends who I'm talking to. But I would say, most of all, a little background, because. [20:16] You know, [20:17] Think about it. [20:19] What is a work of art? [20:21] What is a joke? What is love? What is physics?

[20:26] There is really, if we're honest, no easy way to... [20:30] explain or describe any of those things. And what's happened is we're so habituated to their existence [20:38] that we take them for granted. So you think, oh, I know what physics is. But in fact, if you spent a few years studying it, you'd very quickly feel less confident about what physics is. Yes. And one of the characteristics of complexity is it's so young that we haven't habituated to it. So the question remains out there, as it should always remain out there for everything. Yes. Okay. So I like this kind of unstable nature of everything, quite frankly. So that's one thing [21:08] level at a very average level [21:10] one way to say it is that [21:13] If you think about physics and chemistry, we are a kind of natural science, we are a mathematical science, natural social science. We've been very successful at studying ordinary matter. [21:24] You have the standard model, we know about quarks and gluons and fermions and all the rest. [21:29] And in fact, many of our founders worked on those. But what we're not very good at is understanding what... [21:38] we call problem solving matter. [21:40] And problem solving matter is not just computers and it's not just brains, it's also markets. [21:46] It's also civilizations. It's also cities. And in one way to think about complexity, just think about all those principles... [21:55] that give us insights

[21:57] into how problem solving matter works, [22:02] and how different problem-solving machines... [22:05] and I have a very capacious understanding of what that is, connect. And in a way, if you think about it, when you're building... [22:12] through an evolutionary process, [22:14] ultra evolutionary process a city, you're kind of connecting [22:20] various machines together in the hope that it's going to operate. And it often doesn't, sometimes it does. And so that's what we study. We study problem-solving matter and their connections. [22:31] And our belief is that we have to understand adaptation to understand that. We have to understand computation to understand that and so on. And the history of complexity is, in some sense, a whole series of new models, new principles, new frameworks that... [22:47] are better and better at allowing us to understand that domain. A question I wondered as I was learning about [22:54] this discipline at the very, very basic level, is how does one as a complexity scientist decide at what altitude to study this problem? Because you could, it seems, very easily, you know, look at this at a city level or at a molecular level or at an entomological level or, you know, any number of different spaces where you could decide to focus your attention. Well, here's the beauty of it. One of the things that we study, as you know, is fractals. [23:22] And the complex world is a fractal. [23:24] which means it has a kind of scale invariance.

[23:28] So it doesn't matter is the answer. So in other words, if you think about it, [23:36] It wouldn't surprise you if I said to you, look, [23:39] a cell [23:41] has to adapt to its environment [23:43] It has to... [23:45] process signals [23:47] It has to enact some kind of policy. [23:51] and so on. [23:52] I could have used exactly the same language for you, or affirm. [23:56] And you would have said, "Yes, of course." [23:58] Thank you. [23:58] And so now an interesting question is, is it truly scale invariate at all scales? Presumably, right, as you get very large, you move back into ordinary matter. [24:09] stars, planets, and then it's no longer relevant to use our vocabulary. Or if you get very, very small in the other limit, get down to atoms. So there is this range of scales, [24:20] at which a series of questions [24:22] The same questions can be asked of objects that appear to be superficially totally different. [24:29] But I think it actually doesn't matter, is my view, as long as you understand your scale. [24:35] you can ask all those questions. Then I suppose my question becomes, [24:40] Do you find you have a preference as the investigator that you start to think, "For some reason, this collection, this type of system, I get better data out of or speaks to me in some way that it sparks my mind in an interesting way?" Or even deeper than that, perhaps? Yeah, there are. I mean, it gets to the sort of Aristophanes-Dyson

[25:03] you know, frogs and birds question, right? I think there are people who [25:08] really, really like studying, you know, the order book, you know, in the market. And that's the [25:17] They're like natural historians of that object, but they also apply these principles, right? So they're also theorists. [25:24] Um, where there are others that my colleague, Jeff West, for example, who really is not one of those. And he really loves, uh, [25:33] super generality that his aesthetic, I think this is actually aesthetics. Is it? Yeah. It's not science. It's not science anymore. Right. Because I think the science takes care of itself. It's aesthetics. And I, you know, as you know, I mean, we all have, there are people who just love natural history and love studying butterflies and flowers and all the rest. And there are other people like studying stars. I view that all as a natural history. And there are others who have an interest in the underlying nature, [26:00] mathematical structure, you know, like the [26:02] the Li algebra that describes certain things. It's a different kind of aesthetic preference. In an interview, I think it was 10 plus years ago at this point, you were asked what you would do perhaps if you weren't a complexity scientist. And I believe you answered, you might do a film directing, which has some of the flavor of complexity to it. And it made me wonder, speaking about analogizing, if there were any films that you felt particularly well captured [26:29] the spirit or sensations or thoughts of complexity science to them. In a way, you just look for those directors who have this

[26:39] world systems thinking. And I mean, just to give some examples, I mean, obviously Bergman, obvious, you know, so in his The Seventh Seal, which is his epic meditation on games, life and death. I think, you know, this is this famous scene where a knight plays chess against death and the winner of that, if he wins, he gets to remain alive and he loses, he dies. [27:05] I think that is what we're talking about. Those are the stakes. And understanding the rule, the tactics and the strategies [27:13] of some model of existence is what we're doing. So I love that. I think, you know, Kurosawa's film Rashomon is also fascinating because that film is about [27:26] an event that takes place, a murder, and it's seen through multiple different perspectives, each of which is true to some extent. And fractal, perhaps, somewhat. And maybe even fractal. No, I think so. I mean, absolutely. Different levels, different depths of field. [27:45] you're trying to reconstruct reality from all of these multiple perspectives and the only way to reconstruct reality is from multiple perspectives if if you can at all right which is the ultimate aporia in that film so and then you get you know just masterpieces like Tarkovsky's two sort of sci-fi um movies Stalker uh and Solaris which are [28:08] They're really the boundary of physics and metaphysics. That is...

[28:14] on the one hand, they're scientific subjects, right? One, a post-apocalyptic landscape populated at some point, probably by some extraterrestrial civilization. Um, [28:24] making sense of abandoned artifacts that you can't really work out. And Solaris, the concept of a sentient planet and your [28:34] absolutely futile efforts to communicate with it. Right. And, and yet in the process, making outrageous discoveries that are shockingly amazing. So, but anyway, so I think a lot of directors, and I could go on Bunuel's The Extemporary Angel, I think has this character. There are many great directors [28:53] who take on [28:55] Projects. [28:56] the [28:57] integrate [28:59] the big questions which would be familiar to us in complexity science, but what they do that we do not do, [29:05] is they bind them to the meaning of life. [29:08] and I feel that's something that [29:11] novels and films and art does much better. Science is not good at that, right? That's not science's superpower. We do the other side well. And I love film and novels because many, not all, try to do it, try to connect them. And in a way that it feels subjectively meaningful. Hmm. Yes. Did you say the director's name was Tarkovsky? [29:34] Tchaikovsky, yes. Okay, I haven't seen those, so those I'm excited to. [29:40] It's a must. They really are. Just honestly, I mean, Solaris, I love 2001, but I think I would make a claim that

[29:49] Solaris is the better of the two. And they're both masterpieces, so it's almost impossible to make that judgment. [29:56] Wow. Incredible. Well, [29:58] We're talking about complexity and knowledge and intelligence, and you have been investigating and studying this new form of, I don't know whether we should call it intelligence or not, but artificial intelligence and large language models in particular. [30:28] wouldn't have been able to articulate, but [30:30] perhaps observed. To sort of bring folks along with us, how would you sort of describe that work and these sort of arguments within it? [30:40] So the, the, the, um, [30:43] impetus for that, Deeper. [30:45] was the claim... [30:48] that has been made by now in many papers, that intelligence [30:54] in LLMs is just a matter of scale. [30:57] that if you... [31:00] add GPUs or add data, something you can scale that multiplies up. [31:06] then you go from [31:08] no function or minimal function to some kind of miraculous discontinuity [31:13] a kind of singularity where [31:17] intelligence [31:18] general intelligence, super intelligent, whatever language people prefer emerges.

[31:23] emerges. [31:24] And this paper was a sort of... [31:28] Reckoning. [31:29] with those overblown claims, starting to root this discussion in [31:36] what is a very rigorous science of emergence, which they are completely unaware of. [31:43] And so I think it was partly a little bit of us [31:46] saying you should be familiar with these ideas because [31:51] maybe they'll be useful to you, so it wasn't entirely critical, but also questioning whether or not what they were saying was true at all. [31:59] And I think the conclusion that we reach there is that there's nothing intelligent about Nellalem. [32:04] And that shocks a lot of people. And we can talk about why. But that was the impulse behind that. And just to explain to people, why is emergence sort of the wrong way to talk about [32:17] the capabilities that we see here and that we should expect to see [32:21] as LLMs and this paradigm scale, let's say. [32:26] Yeah. [32:28] Emergence is another one of those ideas, right? [32:31] people don't agree on. [32:33] But that's different from saying that there isn't a very rigorous literature out there. [32:37] You don't have to reach and say, no one agrees on quantum gravity, but we have great theories of gravity and great theories of quantum mechanics. So we haven't put it all together, but the parts are pretty impressive. [32:48] I think just a little bit on emergence. There are two ways to think about emergence. [32:52] One is ontological, about the nature of reality itself, and one is epistemological, which is how we model reality or think about it.

[33:00] The ontological observation is, if you look out into the world, [33:04] There are lots of little quarks or lots of subatomic particles, but somehow they come together to make planets. [33:10] And, um... [33:12] How does that work? How do you get large bodies of stable matter made out of elementary constituents, where these large bodies have properties which aren't obviously [33:22] rooted, not obviously rooted, in the properties of the elementary particles. [33:27] And hence we have [33:31] fields that we call high energy physics or particle physics [33:34] Condensed magic physics and astrophysics. And they reflect the fact that we describe each of these levels differently. [33:41] Now that second point about describing these levels differently is about epistemology, because it's saying, coincident with these different forms of organisation, [33:51] are different theories. [33:54] And... [33:55] Each of these theories is an efficient set of propositions, [33:59] about its respective scale, right? You don't use particle physics to do astrophysics. They overlap, okay. So emergence of that ontology on the one hand, part-whole relations, which have nothing to do with us, [34:13] but also about epistemology, which is what the best theory for the human mind is. [34:17] to describe each of those levels. [34:20] Now, all of them have both the theories and the reality, a very important characteristic, which is, [34:27] Thinking about the mapping from the micro

[34:31] to the macro. So any emergence theory has both. [34:34] What is the microstructure? What is the macro property? Now, if you look at LLMs, [34:39] and read all these papers on emergence, they do not even talk about the mitra macro, except in the most superficial sense. And in fact, most of the [34:50] scaling emergence literature doesn't even look at the micro it'll say [34:54] Look, we added data and now it can do three-digit addition. [34:58] right um big deal okay now [35:03] How does it do it? Well, that we don't look at. So that's part of our argument was, how should we in a principled way think about what should you be looking for? [35:12] such that you are entitled to say that that property is surprising. [35:18] or emergent, right? And the paper sort of goes into different ways to do that. So I think that's one thing. And the other point, of course, is that [35:29] The whole point of the epistemological approach to emergence [35:33] is coming up with parsimonious theories. [35:36] So if you look at LLMs, what is the parsimonious theory of a trillion parameter model? [35:42] There isn't one. It's a trillion parameters. Yes. And whereas my theory of you... [35:48] would be possible as I say, no, well, you know, [35:51] Tell me a bit about your family. Tell me a bit about... [35:54] your interest, tell me a bit about where you went to school, tell me a bit. In other words, it's all very high level, right? It's very psychological, it's all sort of folk psychological.

[36:03] And it would give me a deeper understanding [36:06] not a perfect, but a deeper understanding of you. I would not really get much of an understanding of you if I put you in an fMRI. [36:15] experiment. Yes. And so the deepest... [36:19] point here is that the LLM community is not committed yet [36:24] to a theory of LLM psychology. [36:26] or LLM cognitive science, at that point, they're going to have to reckon with neuro to cognitive to psychological, which they're not. And that can be turned into a kind of complexity speed. [36:39] Fascinating. You've spoken elsewhere, or perhaps you've written about it. I can't remember where I found it, but the sort of three qualities of intelligence being inference, representation, and strategy. Do you see sort of... [36:55] Any of those embodied in an LLM at this point? [36:59] and more broadly, what are your questions about whether this is really intelligent or just [37:05] strictly knowledgeable. Yeah, right, so I should explain why. [37:11] There's a huge literature, it's 100 years old, if not more, on what intelligence is. [37:17] and [37:19] The problem with 99% of it [37:21] is it's entirely human. [37:23] And... [37:24] if you're an evolutionary person like I am. [37:28] then you have to worry about where that came from. It's not then a miracle happened, right?

[37:34] There are people who believe that, which is fine. It's just not my particular church. [37:39] And so the... [37:41] So then I wondered what would be a general framing of the concept that would allow for the possibility of talking about intelligence in viruses? [37:49] or intelligence in bats and intelligence in humans, and intelligence in companies. [37:56] And hence this three-dimensional approach, which is, if you like, one way to say this is, [38:00] Any intelligent system, [38:02] as to perform calculations on representations of the problem [38:08] Thank you. [38:09] To achieve a strategic objective. Yes. [38:13] And it turns out that we have different kinds of theories for all three. The strategic bit is all about game theory. That's what game theory is for. It's about strategy. It doesn't worry too much about it. It does now in algorithmic game theory, but that's quite new. [38:26] Certainly not in the world of von Neumann and in Nash. So strategies are what it's for. [38:34] What's the utility? What's the benefit? And, um, [38:38] Inference is what LLMs do well. [38:41] They do calculations well. That's what they were built for by McCullough and Pitts in 1943. They basically perform logical operations on... [38:50] quasi-binary valued units. [38:52] And what they're also not good at is representation. And that's the bit, by the way, that we would require for emergence to look at the internal wiring, so to speak, to see what they're actually how they're encoding reality. And that's something that artists do beautifully.

[39:10] So, [39:11] You know, art... [39:13] or the novel is not particularly powerful inferentially. It doesn't look like fontan mechanics, but it's representationally extraordinarily deep. It's not particularly strategic either. I mean, it doesn't really have an obvious utility. So I think what I wanted to do is have a space that was very capacious. [39:32] in which you could put all things that have elements of all three, but to different degrees. And so I think that you can see what's happening with LLMs now. The current obsession with LLMs, as you know, is... [39:44] is agentic. Agentic is just code for strategy. [39:48] They're just saying, we need to have a strategic, game-theoretic dimension to what an LLM is. But the representational one... [39:57] has been somewhat neglected except in the trivial sense of "this thing can make". [40:02] kitschy artworks. But that will get better. But even when it is making kitschy artwork, let's say, it doesn't seem to me that it is actually doing sort of representational work in the way that you're really referring to it. Is that a fair appraisal? I think that it's absolutely right. I think that, which requires for me again, looking inside to see what is the... [40:30] elementary principle of representation. What are the atoms? How are you composing things? And I think, if you think about music, when we learn music theory, we learn notes, which are representations of tones. Then we learn phrase structure, right? Then we learn forms, the sonata form, the symphonic form, what have you. These are representations. They're clearly not

[40:53] strategic in a utilitarian sense. They're kind of weird. I mean, what is [40:59] the album kind of blue strategically trying to do other than delight you. So we need a space to allow for the possibility of forms of experience, which are not about utility. [41:12] which is a big part of intelligence. [41:14] This episode is brought to you by Persona, the B2B identity platform helping businesses verify users, fight fraud, and build trust. Fraudsters are already using AI to spoof faces, voices, and documents, so your defenses need to adapt just as fast. [41:30] Persona helps secure some of the Internet's largest and most trusted platforms with identity verification. If you're building a product where trust matters, identity should be a priority. [41:40] you've probably already experienced Persona without realizing it. [41:43] verifying your LinkedIn profile, signing up for Etsy, [41:46] or renting a scooter with lime. [41:48] trusted by leading companies like Square, Brex, and Twilio, [41:52] Persona gives you the building blocks to create identity flows that adapt to your customers, risk tolerance, [41:58] and locales you operate in. [41:59] whether you're verifying age, [42:01] onboarding businesses, [42:02] or automating KYC. [42:04] It's fully configurable, so you can launch in days, not quarters. [42:08] Want to see for yourself? Generalist listeners get a free year of the starter plan. Head to withpersona.com slash generalist and check it out. [42:18] This makes me think of a book I think you must have written about at some point, which is The Glass Bead Game, because it strikes me that actually what Hermann Hesse does in that book then is turning representational into strategy in that there are sort of warring or dueling compositions of music and literature and art. Does that sort of sound roughly right to you? Am I understanding some of what you're saying there? Yes.

[42:45] No, it is absolutely right. And I think that, I mean, that's to me a very important book because the whole world that Herman Hesse evokes in the Glass-Bee game is kind of the Santa Fe Institute. And the game they're playing, the Glass-Bee game, is a little bit, as far as I'm concerned, a precursor to complexity science. I believe that quite strongly. In fact, John Holland, who invented genetic algorithms, that was his favourite novel for a reason. [43:14] And interestingly, Marie Galman's favorite novel was Frankenstein. Oh, wow. No way. Wow. Yeah, that's very interesting too. So yes, I think you're correct. I think in the end, all three come together, right? Yes, exactly. In different orders. In the Glass B game, it was different. It started with the aesthetic impulse. [43:33] and then it moved into the strategic. I think we're living in a world now which is a little bit too led by the strategic inferential and I think it's going to be felt. [43:46] I think we're already feeling it because there's a certain sterility to my experience with AIs. And I think, for want of a better word, it doesn't have the art. [43:59] in it. [43:59] Definitely. I think that's definitely true. It makes me wonder how you would view reinforcement learning through this lens, because in some respect, it seems like that maybe has more of that strategy piece, certainly. But I don't know, maybe that's too simplistic. Maybe it's playing strategy in the same way that

[44:17] generated art is [44:19] playing representation, so to speak. Yeah, the thing about agentic matter is it has [44:27] Intentions. [44:29] It has purpose, it has function, it has desire, if you like, in the book. [44:33] And [44:35] there's you don't require that for rl right all you're saying is [44:38] No. [44:39] here's an input, here's an output, that's not the output I want. [44:43] Don't do that again. [44:44] Try that, right? In other words, some form of Pavlovian conditioning or what have you. And I think that that's very simple and it works for pigeons and it works for ants. And I think [44:55] But what happens, I think, as... [44:57] a system becomes more [44:59] truly strategic and representational is that this [45:03] forward-looking [45:05] intentional dimension starts to expand. [45:09] And I mean, I think that is the, [45:12] if you like, in the space of this three-dimensional space, [45:16] The human... [45:18] prognostication, the human planning, the human counterfactual future, is our special characteristic. And so I'm not sure R.L. [45:30] I think it's kind of agnostic. I think it's a slightly, it's like an orthogonal dimension, I think, to that discussion. [45:36] Very interesting. [45:39] You've spoken before about your sort of concern, which I think is very well founded, that [45:45] LLM's act is sort of such a tempting mental prosthetic for people that

[45:50] you know, if we're not careful, and perhaps even if we are careful, we kind of can't help but allow it to, [45:58] atrophy our own [45:59] cognition. What is the state of your concern about that these days? And do you think there's a way that we sort of managed to navigate a path where that [46:08] That doesn't happen in a profound sense. [46:11] I think it's really much more dangerous than people realize. You know, I've been listening to people at Geoff Hinton talk about the existential threat of AI. [46:20] I think this is the greater threat. [46:22] - Yeah. [46:23] Because we have, you see, with his... [46:26] concerns which are very real we don't have good precedent but my concerns we have great precedent and you just have to look at the [46:37] opioid crisis [46:39] or the sugar crisis, processed food, whatever, the obesity crisis. These are all essentially about [46:49] easy access to either long-term non-nutritive sources, right, of foods of some kind or another, or materials that are addictive. [47:00] Thank you. [47:01] And... [47:03] I'm concerned that... [47:06] LLMs in their current incarnation are late sugar cocaine. [47:11] It's a [47:13] really low quality diet, [47:16] of pre-processed factual knowledge [47:19] that flatters you into addiction.

[47:23] And the consequences are not insignificant. I mean, people have talked a lot about what social media has done to the world. [47:30] And there are good things it's done too, I don't deny it, but there are some dreadful things it's done. And the dreadful things it's done relate to this problem, I think, of low-quality addictive substances. [47:41] And so in an attempt to formalize that, I've been trying to develop a theory of [47:47] Intelligent artifacts. [47:49] When they work for you, they make you smarter, and when they don't. [47:54] I'm happy to talk about it, but briefly speaking, [47:57] I basically divide them into [48:00] right who and [48:01] competitive cognitive artifacts [48:05] And these are things [48:06] like the GPS in your car, that will not make you a better navigator. [48:11] without it. In fact, you're going to become a worse navigator because you have no idea what space even means when you've become a car subject, right? And so, as opposed to complementary artifacts like a map, where if you have the map in your mind, you don't need the map to be physically present to navigate through the world. And I think we can start thinking about all of technology, which is what I've been doing, and developing a mathematical theory for this, which [48:40] You plus the technology works, [48:42] Thank you. [48:43] You plus the technology for a while, minus it technology works. Yes. In other words, like the map case. And I think we have to think very carefully about this because the consequence of the human desire to minimize its energy expenditure are natural tendency towards slothfulness.

[49:03] Our natural tendency to have excessive belief in authority [49:07] And here I am waffling on. Don't believe me. I think, you know, I think, I think those [49:14] Those two [49:15] predilections in combination with this technology are deeply troubling to me. And I think [49:22] It won't be long, right, before... [49:24] you outsource all of your judgment. [49:27] I mean, you know, what should I eat today? Who should I talk to today? Who should I vote for? [49:33] I mean, at that point, [49:34] What are we? Yes. Not only who should I talk to, but what should I say to them? They've just responded. What should I say back? I think I mean that I already see. [49:44] to sort of a startling extent that you can see it on social media, certainly, but you can see just in people's correspondences with one another, you're like, ah, what's actually happening here functionally is your AI is talking to this person's AI. And, you know, the human is actually sort of more or less just a delivery mechanism at this point. [50:05] It's a kind of enslavement. [50:10] Yes. And I think we should understand this. [50:15] And again, we do understand this at some level we know, right? Because everyone is making this judgment call, which is, [50:21] I know that this company just wants to profit from me. [50:24] I mean, that's just a fact. Yes. There's nothing wrong with that. That's great. You know, that will hopefully be a source of innovation. At the same time, you need to develop resistance. But it's so hard to do that. I wonder, you know, if...

[50:38] if that expects too much? I don't think it does. I don't think it does. I feel as if [50:44] You know, it's interesting. Here's why I'm optimistic about that. It's a battle. [50:50] It's a battle between our addictions. [50:53] And, uh... [50:54] Greed. [50:56] against our desire to be stimulated, right? Human beings, it's very interesting. You know, at a certain point in your life, tiddlywinks or... [51:05] what we call in England noughts and crosses. I can't remember what you Americans are. [51:10] It's not interesting anymore. It's like, that's not interesting. I want to play chess now. I'm going to go. I want to play a video game that challenges. I don't mind. But that's not enough for me. So there's something in us that craves challenge. Most people enjoy going out on the weekends and engaging in sport. They're challenging their body, right? It's not effortless. I don't want someone else, as I always say, I don't want someone else to go to the gym for me. That sort [51:35] So there's that side in us which wants to be challenged, wants to be surprised and be thoughtful. And it's fighting. [51:43] this other thing. And what happens in life is that [51:47] Some people win that battle and others lose it. [51:50] And I would like more people to win it. You know, and I feel that it is a battle. A battle that I fear, the part that makes me maybe more pessimistic is that [52:01] Are we as a species becoming... [52:04] more resilient towards that, because certainly the other side is becoming, you know, the temptation is only becoming more and more strongly weaponized. So unless we're sort of at least commensurately raising ourselves, are we sort of railing against a more and more powerful foe? But

[52:21] I suppose we'll have to be hopeful. [52:23] You know, it's a... I don't know. And, you know, it was interesting. I... [52:30] I remember. [52:31] when Christopher Nolan, whose brother I know, Jonah, made his Batman films. [52:38] And... [52:39] I think everyone who went to see those films thought, [52:42] Those were great. Yes. Those were good Batman films, right? And then we go to all these other Marvel DC movies, and they're shit. Yes. Not all of them, but basically they are. But yeah, they're in a very different category, right? They're in a different category. You go to the Batman films and you think, wait a minute, that's a good one. And these are, they're fun. It's a roller coaster ride. No problem. [53:02] And I think... [53:03] To me, what he did there is proof that the human being knows the difference and can appreciate it. And it's our responsibility. [53:13] to provide those, right? [53:16] I mean, you've talked about Cormat. You know, if you've read Cormat McCarthy, let's say you read Blood Meridian, which is his masterpiece, and you've read some kind of... [53:25] Whatever, I'm not going to mention anyone because I'll get in trouble. You know the difference. Of course. But you have to have experienced that to know the difference. If you haven't, you'll say, oh, don't be snobbish, don't be elitist. Yes. There is no hierarchy in the arts. Yes. [53:40] But there is. And I think it takes really having spent time with it. And that's the thing I think we need to do, right, is we need, [53:49] to give people enough time

[53:52] with high quality ideas. [53:56] and they will know the difference. I don't think we should tell anyone what's better or worse, just [54:01] have the opportunity. Yes. I like this idea very much of complementary technologies and the optimistic part of me hopes that we're going to see people find ways to use [54:14] LLMs and AI in this sort of more complementary way. Alan Kay was one of the guests of the podcast, and he had such a brilliant way of just condensing the question, which is how much help is too much help? And I think that was a really good way of sort of putting it succinctly of what we really want these technologies to do. You perhaps want them to educate you, but not give you the answer and guide you and so on and so forth. I know, Alan, I think that's a very nice way to put it. [54:44] Let me add to that, um... [54:46] One of the areas that we work on at SFI is emergent engineering. [54:50] which is [54:51] engineering emergent systems that you don't fully control. And the key concepts in emergent engineering are twofold. [54:59] One is improving signal to noise, and the other one is improving gradient. [55:05] So, [55:06] When Alan's next to that point, [55:08] What he's really saying is, I want to give you a gradient [55:11] I don't want to give you the answer. Yes. But I want to give you a path towards the answer. [55:17] And I think they're radically different. Yes. And, I mean, that's what good...

[55:22] pedagogy, education is. It's not giving the answer, it's saying this is how you can arrive at one. And of course they're an infinite number. Yes. That will give you that. Then you'll have navigational tools for any problem. So in this emerging engineering framework, I think it makes explicit that point. Make the signal clear and [55:40] provide mechanisms that induce in the problem space [55:44] gradients. What is the right way to protect one's mind against these temptations? Like, do you do you just choose to not use this technology at all in your day to day life? Do you sort of treat it more as if, you know, a brief investigation with it that you sort of figure out its capabilities and then try and [56:02] really almost blockade yourself from it. How do you, what is a thinking person's best course of action? Yeah. I mean, I don't know. I think that [56:15] I'm very interested in this concept of autonomy. [56:18] which is... [56:19] You know, when you learn a mathematical science, you solve problem sets. [56:26] Right? That's how you do it, right? [56:29] Here's a problem, solve it. And you solve them on your own. [56:32] and [56:34] They're ruthlessly [56:37] Soothful. [56:39] benchmarks of your capability. [56:42] Yeah. [56:43] And I'm interested in ruthlessly truthful benchmarks of capability. [56:48] and [56:49] we could [56:50] Think carefully about that, what that means for each of us.

[56:53] And I feel as if it gets to the earlier point, the Alan Kay point. So I'm always testing myself, which is, [57:01] I think I know the answer. Do I know the answer? [57:04] "Can I do this on my own or do I need to ask someone else?" And it is actually a part of my practice. It's like going for a run in the morning, right? Or meditating, whatever it is you do. I think... [57:16] You have to develop this [57:18] sit down with a pencil and paper and say, "Do I understand it?" And it's interesting, by the way, because the current sort of renaissance in the Zettelkasten and Nicholas Luhmann's way of organizing his knowledge is actually an effort, I believe, [57:34] to encourage people [57:36] to distill for themselves, right, the essence of an idea. Don't copy and paste it from Wikipedia. Yes. [57:44] write it yourself in your language and I think that's that is actually a very nice example of what I'm talking about in terms of autonomy so I think people should be out there developing truthful benchmark benchmarks of their abilities I think it depends on the field but in the end right if I give you a violin and I say can you play this beautiful piece by Bach and you I mean you either can or you can't and weirdly enough that's why um [58:09] AI now is so effective when the when ImageNet was introduced. [58:14] as a test, all of a sudden [58:18] there was a common standard for what it meant to do well. Everyone agreed. It was like, well, can you do it or not? Is that an elephant or is it a bicycle? And I think we need our own.

[58:28] Build your own benchmarks. [58:30] That's a good practice. I'm going to do that for myself, I think. We've talked a little bit about Cormac McCarthy, and I'd love to talk a little bit more. I think in his last two books, Stella Maris and the Passenger, you see so much of the influence of SFI on his thinking and how he was really spending his days at the Institute to a certain extent. What do you think his legacy [59:00] I think there are many. [59:03] What does the single-minded pursuit of excellence look like? [59:08] that you can never get enough of that. [59:11] I really believe that. I'm extremely... [59:15] Gung-ho for... [59:18] that deep sense of mentorship [59:20] or apprenticeship or whatever you want to call it, which is, I think the only way to really know is if you're lucky enough to spend time in the company of, and it doesn't matter what it is, it could be a swimming coach, I think it, but you know, [59:32] oh, that's what it means to really strive to be excellent. And I think he, regardless of how different his [59:39] medium was, it wasn't writing science papers, that was evident to everybody. [59:44] And that's true for all of our Miller scholars. We have a whole program of other writers as well, of course. He was a reader before he was a writer, right? [59:53] People don't realize this. [59:57] and at some point

[59:59] I might write something on this. He read everything. And he read a lot of stuff he told the public he didn't like. And so I remember when I first met him and we were talking about Henry James and Vladimir Nabokov. [1:00:16] He said, "I don't like that." [1:00:17] Wow, no kidding. Yeah, but years later, he recited to me from memory the first page of Nabokov's novel Pale Fire. [1:00:25] And it was like, [1:00:27] Okay, I see. I see what you mean by not liking. Not liking means... [1:00:31] reading it and not liking it. [1:00:33] Or my earlier remark about how you deal with knowledge, break it down. [1:00:39] as opposed to being ignorant of it. [1:00:41] which is a completely different game. So I thought that his engagement with all of this world [1:00:47] that didn't necessarily show up in the work that he produced that we know. That was important. And also the thing I really interested in him was... [1:00:56] How he... [1:00:58] carved out time for himself or space for himself to think where he was apparently non-productive. I mean, essentially, where would you find Cormac McCarthy in Santa Fe? [1:01:10] At the Centife Institute, [1:01:12] at Baskin-Robbins having an ice cream. [1:01:17] Coffee ice cream. Or watching a movie. He understood that the number of productive hours in a day is you could count on the fingers of one hand, optimistically, and that he needed to gather himself together. And so he was not the slightest bit interested in...

[1:01:38] Posturing - Productivity [1:01:40] Thank you. [1:01:41] If he was bored, he'd get up and go get nice. He didn't hang around and do nothing. And I think part of the reason he was able to do that is because he never used a computer. [1:01:50] Hmm, typewriter, right? [1:01:52] Only a typewriter, but more importantly, no internet. There was no such thing for him. And the rest of us, unfortunately, when we get bored, we all know what we do, right? What email have I got? What things should I look up on the web? And I think so our time gets filled with a kind of low quality busyness. [1:02:13] Whereas his time is filled with sort of high quality hedonism, whatever that was for him. And I think to see that in practice and to see how beautiful his work was and how productive he was, [1:02:26] I think that's important to know. It's like, oh, I see. You don't have to do all this nonsense. [1:02:31] It's okay. It's okay that when you're working, you're working with an extraordinary focus. But when you're not, it's fine. You don't have to do all sorts of rubbish that actually isn't one. [1:02:41] Wow. There's so many lovely tidbits there that I think about. One of them, well, one, I love that he sort of disdained Pale Fire, but clearly had memorized it so much. And to a certain extent, it makes sense because there is such a stylistic difference between them. But on the other hand, [1:03:00] I think if you had to pick maybe the two great stylists of the 20th century, you would arguably pick those two, or at least I would. So anyway, fascinating. But it also strikes me that it sort of gets at the heart of...

[1:03:13] You'll have to remind me exactly the pronunciation. Is it the KELULELE problem? [1:03:17] the... I'm going to give you what? Kekulele? Yes, that is. Kekule, Kekule. Kekule. The idea that, which is basically the idea that the subconscious arrived before language that Cormac wrote about, in a way, I wonder if that sort of [1:03:36] hedonistic time was him sort of allowing the subconscious to do its work. [1:03:42] Absolutely. Absolutely. That's actually a, that is the correct conclusion. I think he understood better than the rest of us. [1:03:51] that [1:03:54] the way you cultivate the unconscious is different to the way you cultivate the conscious mind. [1:04:00] It has different... [1:04:03] gyms, right? It has different exercise regimens, and they're not the same. And it turns out, and we know this actually from the writings of the great mathematician Quangare and Hadamard [1:04:15] where they studied the unconscious processes of mathematicians, including Einstein, actually, where they were very interested in this idea of, [1:04:23] What is the kind of leisure time that [1:04:26] contributes to solving hard problems. [1:04:29] And it had nothing to do with solving the hard problem. It really was about the obvious stuff, like spend time in the woods or whatever it is. Things that we're told to do all the time. Walk. Yeah, all thinking is walking, as far as I'm concerned. Yes. And so, right? And I...

[1:04:46] But again, it doesn't come to us from a... [1:04:50] education at school, that sort of insight. It comes to me from spending time with extraordinary minds. [1:04:57] And that seemed to violate so much that I was told to do. [1:05:01] Yes. We mentioned Wittgenstein earlier and seems to be an obsession both for you and for for Cormac. What do you think it was about him that sort of managed to unite, you know, both of you coming from very different disciplines, but in this sort of deep fascination? First of all. [1:05:21] I remember when I was seventeen. [1:05:23] in a bookshop in [1:05:26] Chelsea and in London [1:05:30] And I found, I saw this slim little volume with the title Tractatus. [1:05:35] To me, that was like... [1:05:36] What? So like this occult knowledge, what is that thing? I didn't know what it meant, but I have no idea what it meant. [1:05:42] And so I pulled it out of the shelves and I was not disappointed because when I opened it, [1:05:48] in hand, you know, that beautiful enumeration of propositions, each of which was a kind of little bit of poetry in its own right. I mean, it's like, is this, what is this book? Is this, I don't know what this book is. And so me, of course, I bought it and I [1:06:00] spent the rest of my time obsessing over it. And then of course, [1:06:04] You learn about the origins of the project and his time with Bertrand Russell and G.E. Moore at Cambridge. And the problems he was wrestling is his influence on the positivists. [1:06:15] Um, [1:06:15] And then his own life, his interest in architecture. By the way, that's very important because Cormac started as an architect. Wittgenstein spent time doing architecture, a house that he made for his sister, which he never lived in because he thought it was so brutal. But nevertheless, nowadays we love it because it's very modern.

[1:06:45] and lived in a hut in Norway. So he's a kind of... He fulfills many latent desires, doesn't he? Because he created a work... [1:06:57] sort of out of whole cloth. He was quite disdainful of, talk about your earlier question about knowledge, and of course he was much more knowledgeable than he let on, like Cormac. I mean, so well educated because of his wealth, right? Exactly. And his... [1:07:10] the access to extraordinary figures at the time when Vienna was sort of, you know, the New York, it was an extraordinary sort of... So it's his life, it's his work, of course, the content of his philosophy as well, because his, I mean, it's worth remembering, people remember, of course, [1:07:29] the two translations: [1:07:31] in English at least, Ramsey's translation of the Tractatus and Elizabeth Anscombe's translation of philosophical investigations, both whose graves I've gone to visit in Cambridge. But of course he wrote all these other books on the philosophy of mathematics, on the philosophy of psychology. [1:07:50] that weren't published in his lifetime. And that's the other thing. He was a polymath, like a novelist. He covered so much territory. So, [1:07:59] Interestingly, if you were to visit our campus, here we have two campuses, one of them is the Miller campus. When you enter, there is the Wittgenstein mural. And on the Wittgenstein mural is proposition 1.13, not one, not six, which everyone knows.

[1:08:16] And Proposition 1.13, which is... [1:08:21] the FAPs in logical space for the world. [1:08:23] And that's a homage to Cormac. It's a homage... [1:08:27] to Bill Miller, who supported the campus. And of course, I can do it because it's my interest as well. And so he's very important. And Wittgenstein's biographer, Ray Monk, spent time with us [1:08:39] And that book was extremely important for Cormat. He considered it one of... [1:08:46] I think it's fair to say [1:08:48] one of maybe the two greatest biographies ever written. [1:08:51] certainly Remax Vickersay and the other one would be Elman's [1:08:55] Joyce and it was Elman's Joyce that really convinced him that he had to be a novelist. [1:09:01] huh i never do that wow how fascinating [1:09:05] I think you've said before that there was sort of nothing... [1:09:08] less interesting to him than than discussing his own work within the the the world of literature. What was worth his energy to him? You know, beyond his work, were there [1:09:21] You know, maybe Nabokov was not up to snuff. No. What did he what did he actually like talking about? [1:09:28] Well, he liked to talk about everything, except his own writing. Yes. Right? And so that's important. I don't think there's anything he wasn't interested in talking about. That's something to bear in mind. [1:09:39] And I mean, I really mean that quite sincerely. [1:09:43] And I think in that article I wrote on him, I give a list of some of the things we talked about. Yes. Pretty much everything, right? Yes. In terms of reading, though...

[1:09:51] I mean, I think the way it worked with him is that [1:09:54] In his youth, he read everything. [1:09:56] in the world of fiction, not everything, but a lot. [1:09:59] more than you might imagine. And very carefully, I mean, I lent him books over the years and they always came back with scribbles all over them. And, uh, yeah, his books are highly annotated. Um, [1:10:09] So he was a careful reader. [1:10:12] Nietzsche once said he defined a philosopher of someone who reads with a pencil. And I think to that extent, he was a philosopher. As he aged. [1:10:22] He settled on a few writers who he loved and, of course, [1:10:27] preeminent among them was Melville. [1:10:30] I'm thinking he considered Moby Dick just without peer. And I'm not sure I agree with that, but there's no doubt that it's kind of a divinely inspired work of literature. But then huge numbers of biographies and then books on everything. If you go to his library... [1:10:46] Very interested in philosophy, very interested in the philosophy of mathematics, philosophy of physics. [1:10:51] And, um, [1:10:53] And everything else, I mean, you know, six volumes on rattlesnakes of Texas or, you know, it just didn't matter. He was, if you think about Judge Holden from Blood Meridian, he was, [1:11:05] the model of his voraciousness is Cormac himself. Wow, fascinating. And in some respects, [1:11:12] In the kid, I see some of that wise infant energy that we talked about earlier, perhaps not genius in that sort of sense. But there's something about that having almost being two sides of a spectrum, both old and young at the same time. You know, that is very true. I mean, that is interest in Frank Lloyd Wright.

[1:11:37] is very similar. I think he's very interested in these people [1:11:41] who are [1:11:43] or at least constellations that do both well. He's not really interested in the middle ground very much. [1:11:50] Right, so you either get to be a kind of Wundeken naïve, [1:11:54] where you're sort of effortlessly brilliant and you're kind of, [1:11:58] by practice an iconoclast. [1:12:01] not because you want to be, but because you are. And then on the other end, as you say, these kinds of polymathic sensibility of wanting to devour the 9th and 11th edition of the Britannica. And I think he... [1:12:16] I think he felt himself over the course of his life... [1:12:20] occupying those two roles. In an interview from a long time ago, you were asked if you could ask an omniscient higher being scientific questions, what they might be. And I really liked your answer and I wanted to follow up on it. Your answer was, does physics plus enough time imply life? Does life plus enough time imply intelligence? And does intelligence plus enough time imply consciousness? Does consciousness plus enough time imply catastrophe? [1:12:49] I followed you, and then to the last sentence, I wondered why in your mind... [1:12:54] Consciousness plus time might imply catastrophe. [1:12:57] Yeah, I mean, I think about that in sort of through three lenses, I think. One is, let's call it... [1:13:07] existential, one is strategic,

[1:13:10] Game theoretic. [1:13:11] and the other one is technological in reverse order right [1:13:16] technological [1:13:18] is what we're experiencing now. [1:13:20] We build machines, we infuse them with facts, we give them... [1:13:25] agency and they annihilate us. That's the preferred marketing speak of large tech companies. [1:13:34] Um, that's possible. The... [1:13:38] strategic game theoretic one is of course mutually assured destruction right i mean that's the doomsday device yes um that we think [1:13:50] through our consciousness and our own sense of malice that others feel as we do and we preempt them [1:13:58] Yes. And so in the process, we initiate a red queen dynamic whose resolution is our annihilation. [1:14:05] And the existential one is a kind of malaise, [1:14:10] It's... [1:14:11] And [1:14:12] The other outcome of the other two, in a way, I mean, technology that makes us a kind of surrogate proxy for itself, strategic machinations that make us despair, and we kind of fold into ourselves. [1:14:29] and disappear. It's a kind of omega point [1:14:34] of human consciousness that just annihilates itself out of desperation. And I know it all sounds rather miserable, but I think that's what I was thinking when I answered the question. I think that's fascinating.

[1:14:47] Well, as we sort of reach the end here, there's a few questions that I always like to sort of end with that are more abstract and philosophical. One is, if you had unlimited resources and no operational constraints, what's an experiment you would really like to run? So I hate the idea of unlimited resources. [1:15:06] There's nothing more boring to me than unlimited resources. Everything that I've ever seen that's great [1:15:13] works under constraint. [1:15:15] You know, that famous Robert Frost, you know, [1:15:19] Poetry without verse is like tennis without a net. [1:15:22] And I really believe that. And I feel that there's something enormously invigorating and thought-inducing in overcoming obstacles. [1:15:34] So I don't want unlimited resources. Thank you very much. It would be too much. It's a cheat. It's a cheat code. Yeah. Why play the game? Yeah, exactly. No, that makes sense. Actually, no one said that before, but, but I understand that. [1:15:47] Yeah, I mean, it's sort of weird. People think they want more, but more just makes... [1:15:52] So a kind of a [1:15:54] a kind of a maximum entropy distribution over possibilities. I want to be, you know, if you think about it, [1:16:02] constraint forces you [1:16:04] to double down on what you do well. [1:16:07] With his new constraints, you can talk about anything. And it doesn't matter. There are no consequences. There have to be consequences. So I'm a real constraint person.

[1:16:17] Junkie. Fascinating. [1:16:20] What tradition or practice from another culture or era do you think we should more widely adopt today? [1:16:27] I think I love this. I am. Yes. You know, I was born in Hawaii. My father was in the Air Force, US Air Force. And I'm a huge admirer of... [1:16:41] Polynesian navigation. And I think that, um, [1:16:45] and [1:16:46] That appreciation of what Timothy Morton came to call a hyperobject, [1:16:50] This massive object, that's the Pacific Ocean, right? This dynamical system that has these hidden currents and forms of order. [1:17:00] that you have to learn in order to navigate in the open ocean in an extraordinary way. So learning the world's hyperobjects is... I get from Polynesia. I think from Japan, [1:17:15] This... [1:17:18] aesthetic of the worn [1:17:20] you know, that [1:17:22] There's beauty in the old... [1:17:25] and in [1:17:27] the imperfect. [1:17:28] You know, the chi cup that's not perfectly... Exactly. And I think that's just beautiful. And it's very interesting. Where in the world do you find the best preserved old technology? [1:17:40] Well, it's beautiful old cars in Cuba and it's electronics in Japan. And they just have the deepest respect. They give burials to machines and electronic gadgets. So there's something there which is so deeply, presumably rooted in their Shinto origins. But I very much admire that.

[1:17:58] I think... [1:18:00] From Germany and England, certainly England in the 19th century and Germany in the 18th and 19th, the pursuit of these giant synthetic works. [1:18:09] I'm thinking about Burton's Anatomy of Melancholy, James Fraser's The Golden Bough, Immanuel Kant's [1:18:17] Three Critiques, Schopenhauer's The World is Will and Representation. These are these massive, lifelong projects to bring it all together. [1:18:27] and they always fail, but they're always exquisitely beautiful. I love that. I love that. [1:18:33] And maybe Italy... [1:18:37] Not contemporary. So again, well, maybe partly. The way that in Italy, under people [1:18:43] like the palladian architecture or in the 20th century scarper and certainly piranese and [1:18:51] mathematics and geometry into physical form. [1:18:54] So to be... [1:18:56] in a Brunelleschi building is like to be in a theorem. [1:19:00] And I love that. And I think that in the age of disposable environments, I want to move into a timeless, platonic environment. And I feel that more strongly in Italy than anywhere else. I'm sure I could go on, but those are some that I admire. Oh gosh, that was an amazing collection. Okay, sort of last question. If you had the power to assign a book to everyone on earth to read and understand, what would you like to give them? I'd say, choose your own adventure. [1:19:28] Hmm.

[1:19:29] Very good. I don't think we should tell people that. [1:19:33] you'll find meaning in something that you find. So told to read, it's like being at school, ruins everything. Yes. It takes the pleasure out of it. It does. There's nothing more exquisite than believing, like I did with the Tractatus, [1:19:47] that you discovered it for yourself and that no one else knows about it. Amazing. Well, David, this was such a joy. And thank you so, so very much for spending your time with me. [1:19:58] You're welcome. Thank you very much for good questions. Thank you. That's it. Thank you for listening to this episode of The Generalist Podcast. Please subscribe on Apple Podcasts, Spotify, or your preferred podcast app. [1:20:11] Ratings and reviews help others discover these discussions, so if you enjoyed the conversation, [1:20:16] I'd be grateful if you could take a moment to leave one. [1:20:19] For all past episodes and more... [1:20:21] Visit us at thegeneralist.substack.com. [1:20:24] dot com. [1:20:25] See you next time as we continue to explore. [1:20:28] the future. [1:20:30] .