The following is a rough transcript which has not been revised by The Jim Rutt Show or J. Doyne Farmer. Please check with us before using any quotations from this transcript. Thank you.
Jim: Today’s guest is Doyne Farmer. Doyne is a complexity systems scientist and entrepreneur with interests in chaos theory, complexity, and econophysics. He’s a professor of complex system science at Oxford University, where he’s also director of the Complexity Economics Program at the Institute for New Economic Thinking. He’s also an external professor at the Santa Fe Institute. Welcome, Doyne.
Doyne: Nice to be here.
Jim: Yeah. Good to chat. Not only is Doyne a very distinguished complexity scientist, but he also has an entrepreneurial business background. He co-founded The Prediction Company, one of the first companies to do fully automated quantitative trading, made a pile of money at that, I believe, and is today chief scientist at Macrocosm Inc., a company aiming to scale up complexity economic methods and reducing them to practice. And one of my favorite Doyne stories is one of his earliest business enterprises was he and some friends built one of the first wearable computers, maybe the first wearable computer – computer in a shoe – and they developed some associated theory to help them beat the roulette table in Las Vegas and managed to escape without any broken kneecaps. That’s a hell of a good story, actually. I remember first time I came out to Santa Fe Institute, I was reading your book about the whole deal. It was “Eudaemonic Pie” or some sort of thing, right? And I was like, whoa. This guy’s definitely a dude. Anyway, today, we’re going to talk about Doyne’s book. I think it came out last year called “Making Sense of Chaos: A Better Economics for a Better World.” Let’s start with defining some terms that you use in the book, which we’ll probably be referring to lots of times. The first is chaos. It’s a core idea of the book. You used it 139 times. One of the fun things I do is try to pull out some of the key terms and then see how many times they’re used. So let’s start and talk about this very important idea, and it’s one that’s kind of interesting. I did not ever get exposed to it in my education, probably because my education happened before you guys formalized it. And I didn’t really get exposed to the idea till about 1999. And I must say it’s one of the most important things I’ve added to my intellectual toolkit as I’ve grown. So for the audience, if you could, describe what we mean when we’re talking about deterministic chaos.
Doyne: Deterministic chaos is when a dynamical system, one of the rules that we use to predict the future, like Newton’s laws, when the dynamics are such that nearby points diverge exponentially away from each other. This happens because in a sense the action of the system is to tangle everything up. It’s like it folds it over and over again. And so you can make a measurement at one time, you know, at some future time get one answer, and at a different time you get a different answer because of tiny variations in what the measurement actually is. The other thing about chaos that’s very important in the book is that it generates what we call endogenous motion – motion from within the system. So even though the equations may not say anything explicitly about it, the solutions move, and they move by themselves in a significant amount. The great example of chaos being weather, where as we all know, it can be difficult to predict, and it’s always changing, and it changes endogenously. There’s nobody up in the skies telling it what to do other than seasonality and a few other things.
Jim: Yeah. Of course, the original discoverer, at least the one that’s given credit, is Lorenz and his very simple system of equations, which generated a pattern with what are called strange attractors, which don’t always turn up in chaotic systems. But could you talk a little bit about attractors?
Doyne: Well, an attractor is, as its name implies, something that everything gets attracted to. Where does stuff end up in the long run? The simplest kind of attractor is just a fixed point. Something moves and it comes to rest, and if you knock it, it will go back to that same rest point. So that would be a fixed point attractor. Next most complicated kind of attractor is called a limit cycle, meaning that there’s a periodic oscillation, like a grandfather clock. You set it in motion, and it just makes the same periodic motion again and again and again. So as long as you keep the clock wound, it’s always gonna go to that same motion. And then chaotic attractors are more interesting because chaotic attractors go all over the place in ways that are difficult to understand and predict, and are not periodic. They’re irregular. And many things in the world, like I argue business cycles, can be described by chaotic attractors or strange attractors. Those are synonyms really.
Jim: And as I said, once I got that concept, my whole view of reality actually changed. I started to realize that almost everything that’s actually interesting about the world has at least an aspect of strange attractor to it. And I found that to be an amazingly powerful lens. And I’d contrast that with what I would call what most of us 14-year-old nerds adopted, which is naive Newtonism. Right? That you could take Newton’s equations of motion, which any good nerd knew by the time they were 14, and apply it forever. Laplace famously said, if an intellect knew the positions of velocities of all particles in the universe, it could predict the future with certainty. Right? Well, in theory, that is correct. When you add in deterministic chaos, even with no stochasticity at the base level, let’s assume a totally deterministic universe, you’re still fucked. You still can’t actually do the calculation practically. And that is a subtle idea, but it’s hugely important. I want the audience to really focus in on how central this is to the complexity lens of understanding reality that we often talk about on the show.
Doyne: Yeah. Maybe just to add a side remark. I just finished writing up some stuff for the program of a museum exhibit on time in Barcelona that will run starting next year. And one of the main points I make, starting with the quote you just made by Laplace, is that chaos actually makes time possible. Laplace is saying if we could predict everything, we could predict the future, we could see into the past, the present would be just like the past and the future just like the present. And the reason we can’t do that is because of chaos. So in a sense, chaos makes time possible.
Jim: I have to read that. It’s interesting. I’ve been talking to people about theories of time recently, including Lee Cronin and Sarah Walker. That’s an interesting conversation. Might have you back. We’ll talk about your theory of time.
Doyne: Sure. No. And I also wrote an essay on the arrow of time that I’ll send you.
Jim: Yeah. Do that. That’d be good. That’s a topic I’m very interested in. Very timely. Alright. Let’s go on to our next topic, which is bounded rationality, which appeared 62 times in the book, which is pretty high for a bigram. Another hugely important concept and one that turns the economics that we were taught in school more or less on its head, which we’ll get to later, so don’t do that. But just say, what do you mean by bounded rationality? Maybe give you a little history of the idea.
Doyne: Yeah. No. It’s a very simple idea, which is that we’re not infinitely smart. And so since we’re not infinitely smart, there’s a lot of things we can’t calculate or can’t easily understand and certainly can’t predict. And so the argument is that an economic theory needs to take that into account. The first person to really push this idea was Herb Simon, who came out with it back in the sixties and talked about whether he felt that rather than couching economic theories in terms of optimality, we should couch them in terms of satisfiability. We come up with ways to do things that are good enough and are as good as we can do.
Bounded rationality, though, it’s interesting – if we’re playing tic-tac-toe, boundedly rationally, I mean, as a 10-year-old, I figured out that there was a strategy O could play and still always get a draw. Actually, I was being rational as a child when I figured that out because I understood the game well enough to see through it and come up with an optimal strategy. So if you could do that, that’s what economists mean by rationality. But if you can’t do that, it’s bounded rationality. When I give talks, I sometimes show a picture of Bobby Fischer and rationality, and then later on I say, well, actually, when he’s playing chess, Bobby Fischer is not rational. He’s a good chess player, maybe a little more rational, certainly more rational than I am in that respect, but he’s not fully rational because he can’t calculate the optimal solution. So it just means you’re stupid enough that you can’t do the task that you’re given at hand perfectly, and so you use some other ways of getting reasonable answers.
Jim: And, of course, that is the reality of essentially everything we do. Right? If we had to figure out how to move our legs while we climb the stairs, we’d fall on our face every time. Right? We have a whole series of hierarchical heuristics that essentially allow us to do that almost all the time. But every once in a while, we do fall on our face because we aren’t actually doing it by calculation. And, again, another hugely important lens that our educational system has done a disservice in some ways by not making clear to people – making again, this is very closely related to this naive Newtonism that we can figure everything out. Well, one, you can’t in principle because of deterministic – well, in principle, practically, you can’t because there’s the chaos. And then practically, you can’t do it because we’re too stupid. Right? And that explains an awful lot, particularly of our contemporary politics. Let’s go on to the third one, which is equilibrium. Well, and we mean in that sense in the economic sense, which appeared 123 times in the book. So, again, a central idea.
Doyne: Well, equilibrium’s a slippery one to pin down ’cause economists mean different things by that term in different places. But I would say there’s two basic meanings. One is just supply equals demand, which allows you to write down an equation that often is essential for getting the solution. But the other is relating back to what I said a moment ago, it’s a strategic notion of equilibrium. If we’re playing a game like tic-tac-toe, have we really thought it through? Do we really understand it well enough to play it like a rational player would? Equilibrium in that sense means that we both settled on strategies that we aren’t gonna change. We found the best possible strategies given the situation, and we’ve settled in to a fixed point to use the attractor nomenclature that I said before, because everything’s settled down. It’s not moving anymore. We’ve settled into that fixed point attractor in the strategy space. And so oftentimes economists really mean that when they talk about equilibrium.
Jim: And closely related to the concept of
Doyne: No. Nash equilibrium is when you’ve found a solution that you cannot find a better solution. As long as your opponent or opponents don’t change their strategies, there is no better solution than the one you’re playing.
Jim: And I would also add to bind our three defined terms together – chaos, bounded rationality, and equilibrium – that deterministic chaos plus bounded rationality make it seem damn unlikely that you’d actually get equilibrium from actual humans playing the game of economics.
Doyne: I talk about a paper we did – I had a series of, I think, five papers and talk about them in, I think, chapter seven, but I could be misremembering the chapter numbers. We actually have really kind of boiled that down. We took a class of games called normal form games. That’s a pretty big class, and we actually were able to show that boundedly rational players will give rise to chaotic behavior in the strategy space if the game is both competitive and complicated. Where competitive means that if I win, you lose, or some version of that – if I win, you tend to lose. And complicated means just that there’s a lot of possible moves in the game, and so the strategies can be very potentially complicated. And so we’ve studied that shitload and shown that chaos is the norm when that occurs, which means if you’re talking about something like the stock market or competing firms in the world or most of economics, we shouldn’t expect that we’re gonna find good equilibrium solutions.
Jim: Yeah. That, again, when I started to learn about this stuff in the late nineties and then when I came out to SFI, powerful learning. Right? That this is really the right way to think about social systems in general. The other kind of Laplace on one side, Harry Seldon on the other from Asimov’s Foundation theories. There’s some kind of – again, every 14-year-old nerd who read Foundation says, “Oh yeah, psychohistory. We can figure out history even at a higher grain level than Laplace.” But nope. Sorry, people. It’s computational incompressibility. The only way we can find out what happens in the real world is to watch it happen. Though we can make predictions to a degree, which we’ll talk about later. Now let’s talk about your perception of the need for a new lens in economics. What are some examples of failures of the traditional approach to economics?
Doyne: Yeah. Well, one of the classic examples was the crisis of 2008, where everybody was way overleveraged and mortgage-backed securities were all over the world, and economists were in denial about the possibility of a housing crash. A few people got that one right, Bob Shiller. But what nobody really got right was that the consequences for the global economy were gonna be enormous. The reason they were enormous was because banks throughout the world held mortgage-backed securities, which turned into garbage, so suddenly everybody’s balance sheet was hit, which meant banks stopped lending money, which meant the whole economy ground to a halt. I open with an anecdote about somebody who was high up at the New York Fed saying in February 2006, ran the best model that the Fed has, and they asked it what happens if the housing market crashes, and it came back and said, “Oh, not much. A little blip.” They were off by a factor of 20 with that. So they were doing their job. They perceptively realized we were at a dangerous point. When they asked the model they had, it came back and gave them completely the wrong answer.
Jim: Yeah. That was quite astounding actually, and there were also many moving parts that they didn’t get right. One of my own pet theories is that in reality, the size of the economic loss to the banks was manageable over a long period of time. It was under a trillion dollars. If you scaled it up after inflation, not much bigger than the S&L crisis, actually, maybe 2x, something like that. But there were other parts of the system, and my own favorite is that the rehypothecation of collateral had produced a pseudo-liquidity of huge scale in the banking system – somewhere between 6 and 10 trillion dollars where Treasury bills and other high-quality securities could be lent out by the people who had them. It’s complicated. You can read about it. But anyway, after the initial Lehman event, the pseudo-liquidity from 6 to 10 trillion dollars of rehypothecation of securities overnight and up to thirty days, imploded, disappeared within two weeks. And so there was a giant shock wave that came from the inner dynamics of the system, which my guess at least is what turned a medium-sized economic mistake into a major financial crisis. Does that make any sense to you?
Doyne: Yeah. Totally. Very interesting too. I mean, I was part of the group that worked to form what now became the Office of Financial Research, whose original intent was to keep track of what’s going on in the economy so we understand stuff like that. Unfortunately, the original vision was never acted on, so we still don’t have that, but we could. We could have a tool.
Jim: Yeah. It’s funny. I was actually proposed as director of that office, and that was turned down because I didn’t have a PhD in economics. I laid out a plan of how to build the data and how to tax the institutions that created the data to get the data and how to do some of the things like you’re talking about later in the book. And apparently, I also scared them because they knew I was not a very tractable person. I thought that would have been funny, but I had an inside advocate there, which was kind of fun. But anyway, another story for another day. Okay. Now let’s talk a little bit about standard economic theory so we can contrast it with your vision. Can you give the audience the three-minute version of Econ 101 and 102?
Doyne: Well, so this is standard for how theories work. Theories in economics were all built on the same template. You assign all the agents, anybody who’s making decisions, you give them a utility function, and then you give them a model of the world. Rational expectations is the most common one, but it doesn’t need to be that. It could be some more cloudy lens. Then you assume equilibrium. Supply equals demand. You write those equations down, you solve them, and you deduce the optimal decisions that each agent should make in order to maximize their utility, subject to the fact that they know that all the other agents are doing the same thing. So very powerful, very hard to do, and hence models are kept pretty simple.
Jim: I took some economics, I actually took macro from Samuelson himself in a class of 15 students, a little seminar. You know, my take was this sounds like fairly obscure bullshit actually. Decided not to pursue a career in economics based on that course. And one of the things I would now say looking back is the sociology of economics is very bound up in closed-form solutions, i.e. equations, systems of differential equations, etcetera, that can in theory be solved or at least with Wolfram Mathematica or something. And yet that’s probably a peculiar lens to restrict the exploration of a complex system like our economy. Maybe talk about that a little bit and then transition to the complexity view.
Doyne: Yeah. Well, I totally agree with what you just said. It’s losing track of the real purpose of what you’re doing. My view is if I’m given a problem, I go, what tool should I use to solve this problem? And I look around and try and find the best tool, whatever it is. And the problem is most many problems at least, if not most, in economics are complicated, and the formalism I just said, this you can’t get the solution. If you can find it, it would take a computer, you know, an infinite time, and so you have to do something more practical, which is what we’re focusing on, namely to simulate the economy.
Jim: Yeah. I liked, I think you used it in the book, a description of comparing “as if” versus “as is.” Maybe you could expand on that a little bit.
Doyne: Sure. Economists frequently justify what they’re doing. They recognize that real people don’t say, “Oh, this is my utility function. Ah, I’m gonna do a calculation to see what I should do that would maximize utility. Okay, it’s this, I’ll do that.” People don’t really do that. But economists will argue, it’s “as if” people did it because if you could get the right answer, that’s all you need. And the problem is it’s hard to tell generally in economics whether you’re getting the right answers, and the data is always very limited for testing things. And in any case, I’m much more comfortable with an answer that matches the way things really work, which I call the principle of verisimilitude. You want your model to reflect the things in the real world that you care about in a way that makes sense and reflects how the real world works. So I call that “as is” rather than “as if.”
Jim: Yeah, I like that. Quite powerful way of thinking about it, and it allows us to examine behaviors that we know exist in the economic world. Like for instance, why do 50% of people who have credit cards maintain balances near their limit and pay 22% to some thieving bank company? Very few of the standard economic theories would explain that behavior. But when you bring in some behavioral lenses and particularly very bounded rationality, then it starts to make sense. And to the fact that we know that to be an empirical fact, why wouldn’t we build that into our models of reality?
Doyne: Yeah. Well, it’s hard to write down an equation that describes that.
Jim: Yep. And, again, that’s the sociology of economics. They – and I think you actually quoted the old story, which I use all the time for various scientific failures – the guy looking for his keys and he’s under the streetlight and someone comes by. “What are you doing?” “Looking for my keys.” “Did you lose them there?” “No. I lost them over there, but the light’s better here.” Right? So the idea that you become so in love with your methodology rather than trying to get to the truth.
Doyne: I heard that story originally in the context of nonlinear dynamics, nonlinear behavior from Stan Ulam, who was saying, “We collectively have spent our entire scientific careers up until the advent of the computer looking under the lamppost, and now finally the computer frees us to go look where we want to look.”
Jim: Which then gets us to some of the principles of your complexity economics. You mentioned one, verisimilitude. I had to look up how to pronounce that. I’m not sure I’ve ever said that word before, but I knew what it meant, but it’s not one you see in a book too often, right? Two of the other ideas, core principles from complexity economics that seem to pop out of the book – first and foremost, heterogeneity and dynamics. Maybe we could talk about that a little bit. Heterogeneity is something I often point out when we talk about social systems. Some people want to collapse everything to, you know, ping pong balls or the famous stereotype of a physicist: “Let’s model a cow. Let’s assume it’s a sphere.” Right? So talk about heterogeneity and dynamics, then we’ll go on to the next piece.
Doyne: I mean, heterogeneity is just the fact that we’re all different, and people live in different houses. They make different incomes. They have different ages. And that means it’s very dangerous to do what economists often do, which is to reduce things to what they call representative agents. Some of the most famous models that are actually used in economics reduce all the households of the world, all 8 billion of them, down to one household. Sometimes that can work if everything’s linear, but when things are nonlinear, that’s dramatically different. I give some examples in the book. Now it’s true that economists are working very hard to tackle heterogeneity, but they haven’t been very successful. And it really ties up this formalism into knots because it’s just too intractable to put it in there. They have some solutions, but their hands are really tied ultimately. Whereas with what we do, we just simulate the world. We simulate as many individuals as we need to simulate. They can be arbitrarily different.
Jim: Now let’s get to the third part of complexity economics, the core tool of agent-based modeling.
Doyne: So agent-based model is just a simulation, a digital twin, making a replica of something inside the computer and trying to render it as faithfully as you need to to understand the system you’re trying to understand. If it’s the economy and you’re trying to build a model of the whole economy, put in a software object called a household that has individuals inside it. They have ages. You can list their ages, and you know something about – you ideally want to have that module understand their consumption behavior and their savings behavior, and what availability they have to work in what kind of jobs. That would be your housing sector. You have a firm sector. We have firms that make goods that need workers, that need capital. You have banks that lend out the capital that could fail. You have the government that’s setting short-term interest rates – or the central bank, I should say. The government’s setting fiscal policy, how much you’re gonna spend. And so you just make a digital twin that has all the basic elements in it, and you endow them with as much verisimilitude as you need in order to get a good model.
Jim: And, of course, the heterogeneity. Right? There’s no reason that they’re all the same. Though they can be in some cases. Sometimes it’s very simple stupid model as we learned way back yonder. Zero intelligence model could actually share some interesting results, but you basically layer them in. Now on the flip side, there are some critiques of agent-based models with respect to – I remember when I showed up at the Santa Fe Institute as a naive, baby researcher and showed somebody my wonderful social simulation with 47 parameters. You know, they slapped me on the hand and said, “Boy, 47 parameters, you can produce any result you want. See if you can cook that down to three, two would be better, one would be best.” What about this critique that agent-based modeling is kind of like a startup business spreadsheeting that you can get any answer that you want by tweaking a zillion parameters and adding new parameters when necessary?
Doyne: Well, that’s a danger that needs to be taken seriously. It’s a valid criticism of much of the work that’s been done. But agent-based modeling is going through a transition right now, at least in economics, actually in other fields as well, to becoming a more quantitative science or a tool that you can use for quantitative purposes, which means that you have to have datasets that tell you how the world works that you can match your model up to, and most importantly, you have to have a way of testing the model to see whether it’s producing reliable predictions. So that’s something we’ve worked on quite hard over the last decade. We have better algorithms for setting parameters of agent-based models from data. We have ways of understanding – well, okay, you have 40 parameters in a model, but what’s really going on? It turns out that usually there’s only about three independent directions in the parameter space that control 99% of what happens. So if you understand those three directions, then you’re in good shape. We’ve sort of come up with ways of constructing models that have verisimilitude and have it in a quantitative way where we could really – I mean, rule of thumb is test everything in sight. Anything you can measure and match your model up to, test it and see whether it’s right. So that you can look down inside the model, ideally, and if it’s not doing the right thing, see what’s wrong and make it do that part better. It’s a big enterprise to build a large-scale quantitative model, not something you’re going to do in your spare time in your kitchen, but it’s quite doable. Just requires a lot of work and good judgment.
Jim: It’s been a while since I’ve dipped into the literature of agent-based models. I have to refresh my knowledge on these newer theories and how to build really quantifiably verified models and probably something like machine learning approaches where you can use large datasets in an intelligent unsupervised fashion, I suspect, would be relevant also.
Doyne: Very relevant. These days I usually don’t start designing a model until I understand the data that I’m gonna use the model with.
Jim: Yeah. One of the big insights of recent movement towards a more data-based approach to science, which some people disagree with, right? Say, “Oh, whoop, we can’t forget about theory. Theory is also important,” but we’ve probably been over-indexed on theory in some domains, particularly social science where theory is just some stuff some dudes made up, mostly. All right. Well, now let’s move on to what new perspectives does complexity economics give us. I think let’s start with what is the actual nature of complexity? What is it about complexity that you, I, lots of other people find so interesting? What is the core driver of it?
Doyne: Yeah. So I opt for a definition that really is not a full definition, which is that complex systems exhibit emergent phenomena. So, okay, what’s an emergent phenomena? An emergent phenomena is something that you couldn’t have easily anticipated no matter how smart you are. So chaos is a nice example of a simple emergent phenomenon. You show me some equations – general, even the best mathematicians can’t just tell you whether it’s gonna be chaotic or not. To find out, you really have to simulate it and see what it does, even if those equations are very simple to write down. That’s in a way when complexity becomes most dramatic when you see something that’s simple generate very complex behavior.
The classic example of a complex system I like to quote is the human brain. You take 85 billion neurons, wire them up in some complicated way, and how would you have predicted you would be able to get that if all you were doing was looking at a single neuron? From the understanding of a single neuron, it would be hard to predict that you could get a person.
The economy is maybe a simpler complex system to understand, but it very much has emergent phenomena. I do an estimate in the book, and I think I calculate that if we didn’t have the economy, we’d have to have about a factor of 100,000 less people. Because if we were all just Robinson Crusoes out there operating independently of everybody else, plowing our fields and shooting animals or whatever we were doing – we probably wouldn’t be shooting because how would we manufacture guns? The capacity to support human beings on Earth would be way, way smaller. That’s because the economy is a remarkable amplifier of our effort. I find it shocking as I thought about this more, how just specializing and sharing ideas and commerce, what a powerful thing that is. What a huge amplification it gives us.
Jim: I had that similar insight when my daughter bought her first car. I helped her negotiate the deals. I love to negotiate car deals. I’m one of those perverse people that actually likes to screw with car salesmen. Her first car was about $25,000 – it was a Nissan Juke. And then as I stepped back a bit, it’s amazing. $25,000, which was about six months’ salary for a starting graphic artist. Six months of a person fresh out of college.
How long would an artisan take to build a car?
Doyne: And I-
Jim: It had an automatic transmission, AWD, air conditioning. And I looked at it and said, you know, a skilled artisan with a metal shop and a forge in six months might be able to make the driver’s seat if he was good and knew other things like how to stretch hides and how to sew and all this stuff. And I go, damn. This thing has a motor and transmission and tires and complicated electronics, and it didn’t yet have a computer screen and stuff. But how in the world did our work get so optimized that things from all over the world could come together to be able to produce this artifact that could be paid for in six months’ worth of work by an entry-level person? Quite remarkable. And, you know, Adam Smith hits on this with his early examples of the pin factory, for instance. Right? Which is kind of interesting. Alright. So let’s move on a little deeper into complexity, and that is the distinction between endogenous and exogenous change and how that relates to the economy.
Doyne: So endogenous change is when it comes from within. The 2008 crisis was a good example. It didn’t happen because a comet hit the Earth or something like that. The opposite would be the COVID recession, which happened due to a disease that’s exogenous in the sense that no economist would have been able to foresee that was gonna happen outside of that domain. And both are important for the economy, but mainstream economics is really set up to deal with exogenous changes and not set up to deal with endogenous changes.
Jim: Yeah. And indeed, a lot of traditional economics oddly assumes that everything is exogenous because otherwise you’d have equilibrium. When I was a young entrepreneur, I would sometimes say, wait a minute now. If the economists are correct, why would anybody be an entrepreneur? Because nothing ever changes. Right? And yet we know it does. So go ahead and be an entrepreneur. Alright. One final, sort of basic building block is one of the things that differentiates the economy and other social sciences from, say, the physics of chaotic systems and complex systems is the agentic nature of the players. Talk about that a little bit.
Doyne: Well, people make decisions, we think. And one of the biggest debates we already touched on with bounded rationality is how do we model thinking? Classic economics way to model thinking is assume everybody’s perfectly rational, and problem is that’s too powerful a model and too hard to solve. Then you’re back to understanding how do you put thinking into the agents that you have in your simulation, in your model, and that’s maybe the most challenging problem of all, other than finding data.
Jim: Yeah, fortunately, we now have some actual help from the behavioral economics folks.
Doyne: Yeah, and, you know, one of the things you find is, well, first of all, people make a lot of decisions using heuristics. Rules of thumb, imitate the best. You know, look around, see who’s doing it. Oh, they seem to be good at it. I’m just going to mimic what they’re doing. Or trial and error. Try something. If it works, keep doing it. If it doesn’t work, try something else. If you’re a financial investor, buy undervalued assets. I mean, in a model, that’s a heuristic that works great for many purposes. You don’t need to know exactly what they’re doing, in most cases, to get good ideas about how agents are behaving in financial markets, because there’s always a substantial body of agents following that rule. You also make strategic decisions trying to think ahead. If you’re playing chess, for example, you start with some heuristics like control the middle. You wanna swap a queen for a rook if you can, or vice versa. I mean, you wanna get the queen. You’d rather kill if you can kill somebody’s queen with your rook and they kill your rook, that’s a great trade. You have heuristics, but at the same time, you’re trying to look ahead, but you can only look ahead a few moves. You have bounded rationality. Very often it’s easy to put these things into models where you use just that kind of logic.
Jim: Yep. And adds a whole another level to the problem essentially, right, as opposed to particles bouncing off each other, which can also produce complex adaptive systems. It’s a different class of system. Alright. Well, now let’s go on. Why does this matter? You have actually quite interesting section in the book on understanding the metabolism of a civilization. Maybe just pop up one higher level on, you know, why is this important? What is it that the economy is actually doing for us?
Doyne: As a complex system scientist, we’re always fascinated by analogies that you see. And if you make an analogy to a living system, the economy is the part of society that makes our food. It produces stuff that allows things to run. It takes and I think more to state a more concise definition, it combines natural resources with human labor to make goods and services that we need or at least think we need or at least want. Whereas if you look at what is a metabolism, metabolism’s the thing that breaks down food and produces energy and allows an organism to reproduce. A metabolism takes in raw material in the form of food of some kind, breaks it down, and reformulates it to power the organism. So it’s very, very analogous function.
Jim: Basically, it has resulted in this amazing stack of things that we have out here. Some of them useful, some of them not so useful. You know, that’s the problem of the endogenous system that optimizes for short-term money-on-money return. It doesn’t necessarily maximize for real human well-being. Yeah, we’ll get back to that conversation another day. So now let’s take the next step. We have this metabolism, which is essentially the engine that has created the 100,000x intensification of humanity. How does agent-based modeling help us understand this?
Doyne: Well, it helps us understand it because it allows us to create a replica and use it to make predictions or analyze policies or strategies and see where they lead us. I mean, a good example – I start the book off with describing the work we did to build an economic model of the impact of COVID on the United Kingdom. We were able to lay out several different scenarios for how the government could come out of the first lockdown, analyze their economic consequences, and find a least-bad policy, which was for them to keep upstream industries open, like mining, forestry, etc., but close the downstream customer-facing industries where the infections were really happening most, but where you can actually get by pretty well without crippling the whole economy. So maybe that’s sort of intuitively obvious, and it may be a part of why we tested it, but we could test it and say, “Look, this is what the economic impact will be if you do that.” If you do these other strategies, you’ll have these other impacts. Well, they took the choice we favored. We were, as you can imagine, tickled as we could be when a year later, when the government finally gathered the statistics, it turns out we’d hit everything bang on the head. And as we did postmortem on our model, we could see that we were a little bit lucky in places, but mostly we got a good answer because our model had a lot of verisimilitude. It really did mimic the way the economy was shutting down, and we had things in there like realistic production functions that actually gave the correct formula for what a given industry absolutely needs to keep operating. So we were able to show that was essential to getting a good result.
Jim: We talked earlier about the fact that the nature of the complex is emergence. Do you have some good examples of emergent phenomena that have come out of your agent-based models?
Doyne: Yeah. I do. I talk in the finance part of the book about a model we made for leverage cycles. And how did we get that model? Well, we were building a model of the economy because we’re the generic model where we were very focused on the financial system in this model. We had lots of banks and we had firms and households, we initially didn’t impose any risk control on the banks. So we said, “Well, obviously we need risk control,” and so we imposed the Basel II method of risk control, which is called value at risk. You try and anticipate how much risk you have and adjust your leverage, that is how much you’re borrowing, to be appropriate. If things are more volatile in the future, more uncertain, you use less leverage. If things are fairly certain, you can use more leverage. So it’s a strategy that makes plenty of sense to an individual.
But anyway, what we did, we imposed it, and we saw the whole economy just started oscillating. There would be a fifteen-year slow run up like the Great Moderation and then a big crash, and then it would happen again. And so we took the model apart, simplified it down to the absolute bare bones to see what was causing this, and we discovered that it was really built into that situation because when large numbers of people use the Basel rule, value at risk, to manage their risk, then the problem is that they can make matters worse. Because if you’re leveraged and your bank asks you if there’s a downward price movement for whatever reason, then your leverage goes up. So now you’re over your leverage threshold. So the bank that lent you the money says, “Oh, you need to pay some of that money back cause we’ve got to hold you at your leverage threshold.” So how do you do that? You sell assets. And if it’s triggered by the market falling and now people start to sell assets to get their leverage back in balance, then they make even bigger price moves, and you can set up a cascade and cause the whole economy to come down. That’s what was happening in our model, and it just took about a fifteen-year buildup to reach the point where the economy got so highly leveraged that it went to that place. That was an example of an emergent phenomenon. It certainly surprised us when we saw it, and I think it surprised everybody in 2008 when we experienced it.
Jim: Absolutely. And one of the things that people describe 2008 as the mother of all correlations. Right? Even my part of my portfolio that I designed to be bulletproof took a small hit in 2008 because the correlations that were so unexpected by traditional statistical measures occurred. And you make a very good point that some of the policy prescriptions may well have been forcing these correlations. Right? Behaviors that otherwise would have been more heterogeneous were all forced into a single behavior, everybody trying to go through the same door, and we know what happens in a live market when everybody’s trying to sell. Boom. Down it goes. Yeah. A little aside here. I didn’t really intend to cover this, but we’ll talk about it anyway, which is the flash crash of 1987, I believe it was. You do talk about that a little bit as another example.
Doyne: I mentioned it. You know, my colleague, Jean-Philippe Bouchaud, has been studying this extensively. It turns out there’s flash crashes going on all the time. That just happened to be a particularly big one. The origin of those flash crashes is not completely clear. He’s been working on a theory for them. We contributed something to that, the work I did at the Santa Fe Institute when you were around, Jim. Basically, you have a combination of a lack of liquidity and feedback mechanism that causes things to temporarily spiral out of control, and that’s inherent to market making. Those a combination of simple things like that, it seems, can just cause an instability to pop up, but we’re still trying to properly understand that.
Jim: Yeah. I saw that in my own stuff when I was working with your group where I built this little simple ecosystem of noise traders plus book-shaped traders, market makers, and trend followers. You would see that. You’d see this then boom, boom, boom for no good reason. It’s endogenous interaction of strategies. And again, another interesting example of emergent results. So now let’s kick up one level higher, and that is the thinking that you laid out in the book about the nature of business cycles. First, maybe define for the audience what a business cycle is and why they’re important and what your emerging insights are on business cycles with respect to complexity economics.
Doyne: So business cycles just reflect the fact that the economy is changing all the time. GDP goes up, GDP goes down. We see fluctuations, we have recessions, we recover from the recessions. So any of the lines you would draw are all wiggly, and they’re wiggly. The question is why? And in traditional models, the assumption is there’s noise from things like us becoming more impatient or less impatient as a group, and various external causes that get amplified and turn into business cycles. Whereas many of us, including many traditional economists, think many business cycles are endogenous. The 2008 crisis certainly seems endogenous, and yet so we need a way to model it. And hence, chaos, which is a conceptual mechanism that generates business cycles. We also have extensively studied examples of this in game theory, and one of our main motivations for building an accurate agent-based model of the whole economy is to understand the drivers of business cycles. Big problem. Surprising that it’s still not answered in economics. It’s one of the big questions that’s sitting there for macroeconomists, probably, certainly one of the top five questions in macroeconomics, and we don’t know the answer.
Jim: I remember when I was a student, one of the theories that was put forth was due to imperfect information flows in the hard economy, inventories would tend to build up, and that they were significant – thought to be back in the seventies, when I was a lad, thought to be a major driver of business cycles. But then with better information systems and better logistics systems and the computerization of everything, companies got way better at controlling their inventories, and that may have had something to do with the great moderation. But business cycles, as we discovered, did not disappear.
Doyne: So there are many other things going on. But I argue in the book that the key element is that we’re boundedly rational. So we can’t look ahead and see what everything implies. I was once giving a talk at the OECD in Paris, and there was a discussant who was from the Bank of International Settlements, people that bring us the Basel rules, and we presented the model that I just mentioned a bit ago about leverage cycles. He said, “Well, but a rational investor would see that these were going to occur and behave in a way to keep them from happening.” And I said, “Yes. But where were those rational investors when this happened?” I viewed that as a strength of the model rather than a flaw of the model. I think many economists are coming to concede this point now. Let me just go back to the bigger point. The bigger point is that if we collectively can’t accurately see the future, we’re gonna overshoot and undershoot, and our expectations go backwards and forwards, we’re collectively hurting. All kinds of things are happening that are gonna make the economy oscillate around the right answer.
Jim: Gotcha. And that’s what we actually see. It’s amazing that the economists – you know, well, here’s this elephant in the room. This stuff happens, right? And their normal set of tools don’t really predict it. So let’s move on to the next thing you talk about in the book. An important part of the economy in some sense is the financial system, and maybe we can start out with your personal story as a prediction company, which actually gave you some important insights into all this.
Doyne: Yeah, it certainly did. In a sense, this goes all the way back to doing roulette, because what roulette made me realize is it’s really fun to try to find something that people say you can’t predict and then figure out how to predict it. And so that’s become a theme in my life. In the eighties, we developed some methods for making good forecasts. Somebody would hand us data – it might be ice ages or sunspots – and we would run our procedure and make predictions, and we could beat the standard models that people were using at the time.
Whenever I would talk about this, somebody in the crowd would raise their hand and say, “Have you tried applying this to the stock market?” I would have to say no. Finally, I was approaching ten years at Los Alamos, and they were gonna give me a ceremonial nut dish, so I decided I had to get out of there. And so we just spun off and started a company to predict markets, not knowing much about what was going on. We realized we needed to find a partner who really understood finance and could help us raise capital and so on. So we did manage to do that. We found a very good partner. And then we got to put our heads down and really work and figure out how to beat the market. We had a team of about 20 people. We built a very good computing system, languages. We wrote languages for processing the data and built a system that allowed us to make fully automated trades, and finally came up with a pretty good method that made steady profits.
Jim: One thing that I found very interesting in your description here was the days when you put on a suit and tie – I wanna see pictures of that, Doyne Farmer in a suit and tie – and made the rounds of trading houses. And I thought the fact that you had discovered heuristics even at these top-of-the-food-chain trading companies was hugely interesting. Maybe you could tell that story a little bit.
Doyne: Yeah. Well, as you said, we actually had a debate in the company. We realized you needed to go to New York. They expect you to wear a suit. So I didn’t own a suit. And so we had a company debate about whether the company should pay for it or I should pay for it. Decided I would pay for it. I actually went to this little shop in downtown Santa Fe where kind of a gay guy helped me get the right clothes to go to New York. I had the right kind of tie and the right kind of shoes and everything.
The thing that fascinated me was the diversity even within the financial business of the anthropology. There are different kinds of financial firms. People in different firms speak differently, dress differently, and follow different strategies. And even in the same building – remember, in one floor of Bank of America, this was actually in San Francisco, they said, “Oh, we just follow technical trading rules. That’s the only thing that makes any sense.” And the floor below them said, “We only follow the fundamentals. Those technical trader guys are crazy.”
I was just surprised at the diversity of beliefs that vast sums of money were being traded on in the market, and also became aware that a lot of this was fairly bogus back then. I think it’s much tighter now. The quants are more in control of the system. I had actually edited a few things out of the book where I named specific companies and specific clowns I met in them and those ridiculous strategies they were following.
Jim: Yeah. It was quite remarkable to me when I did business on the other side selling tools to Wall Street through a lot of my career, and I’d scratch my head. God, these suckers are making millions of dollars doing technical analysis. What the fuck? Right? I think I would be just as good off to go talk to an astrologer, tell you the truth. They might even give you a better reading. And the people on Wall Street believe this stuff. And, of course, one of the things that we discovered when we were doing the work in your group was that having these beliefs actually has impact in the market. Right? Which you describe in the book, which I think is a very – again, another very useful lens is thinking about finance as an ecology and ongoing evolution. Maybe talk about that a little bit.
Doyne: Yeah. During my time at Prediction Company, we started reading the financial economics literature because we’d get a little thing from our partner O’Connor and Associates. You’d check off the ones you wanted. This is before the internet. And they would just send us a pile of papers to read every month. And so I became familiar with the financial economics literature. It was useful because there were a lot of papers along the lines of what were called stock anomalies, weird things that appeared to make money. So we implemented and tested every single one we could find and found that half of them we couldn’t reproduce, period. The other half, we could reproduce the result on the data they were using, but when we used other data, we couldn’t reproduce it. And then the remaining quarter actually was useful. So we built a lot of our portfolio out of those.
But as I was thinking about it, I realized that in this financial economics literature, they typically assume you have these rational investors, and you have some tiny perturbation away from perfection. Whereas what I had deduced is people don’t have any real idea what’s going on in the market. Everybody’s strategy is highly inaccurate. And, you know, one of the calculations you can do is if you can beat the market, if you can make the right decision 55% of the time and you do daily trades, you make a ton of money. So even a very noisy strategy can work extremely well as long as it’s actually true.
What he realizes is, no, the market has all these different specialists. You’ve got the guys on one floor who are doing technical trading, the guys on the next floor that follow fundamentals, the guys on the floor below that are market makers, the guys on the floor below that who are option traders, and they’re all doing very different things. And the market efficiency just doesn’t come automatically. Insofar as the market is remotely efficient, it’s because all of these different strategies combine together to make some profits, even though they’re all quite inaccurate and nudge the market a little bit towards efficiency. I’m a fan of Fischer Black’s statement: “Yes, I believe in market efficiency. The market’s within a factor of two of fundamental values 90% of the time.” I think he actually probably has the numbers about right.
Jim: Yeah. It is amazing how much this even in S&P – any given S&P 500 stock oscillates in a given year. It’s a big number, way bigger than the economic value could actually be changing.
Doyne: Exactly. There’s no way all that financial volatility is coming from fundamentals, as Bob Shiller pointed out. But if you have a market ecology where each agent is implementing a strategy that has some useful bounded rationality in it – typically, though there are people who trade with what really is equivalent to astrology. Because we found some of the technical trading rules actually work, not super well, and, you know, they may have years of not working, and they’ll have a few years where they work. But you have this big ecology of different strategies groping its way towards setting prices at some reasonable values, but fluctuations up and down are huge and often well off the mark.
Jim: In your work at Prediction Company, did you explicitly model ecosystems of traders, or were you more brute force statistics?
Doyne: We were brute force statistics. It was only later that I started making models with different kinds of agents to see how they would work and so on.
Jim: Yeah. In retrospect, I wish I’d taken my work that I did with your group. Well, actually, made a conscious decision not to trade it. Right? I was so burned out in the business world after busting my ass for twenty-five years. The last thing I wanted to do was trade. I actually had a guy come and offer me half a million dollars to put it to work, and I just said no. I think it would have worked at the time, but it was, you know, specifically the book shape stuff, back testing, and it seemed pretty powerful, but no doubt that’s all.
Doyne: Have made money doing that, Jim. I know people, these high-frequency firms, they do that in a major way nowadays and have been for a while.
Jim: And, of course, we know that the opportunity keeps shrinking. You do talk about that over time that inevitably the sharks find the food, right, and eat it all, and then they eat too much to start losing money. And then somebody figures out how to take advantage of that behavior, which is the evolutionary part.
Doyne: Totally right. And I think you’re right that if you’d done what you said back in the nineties, that would have been a big moneymaker.
Jim: Or in my case, it was early 2000s. But even then, it would have been a big moneymaker.
Doyne: There was still time to do it back then.
Jim: But in retrospect, I’m glad I stayed away from that because it would have been a tar pit. Then we all know what some of those finance guys are like in terms of a moral ecosystem. Not all of you guys out there in finance are bad, but there’s a fair number of sick fucking sociopaths to not put too sharp an edge on it. Yeah. Alright. Let’s move on again at a higher level to another area where Doyne has made some serious contributions, and it actually ties together in an interesting way, and that is climate economics. And let’s actually start with the interesting analogy to where you’re going with complexity economics, which is the development of weather forecasting.
Doyne: Well, I just found fascinating. I read a really good book called “A Vast Machine” that chronicles how weather forecasting got developed. And the obvious question you ask is why not do the same in economics? And the answer to that is a sociological one. But you know, just to review how they forecast weather – you make measurements all over the world. You throw them into a big computer that simulates the entire atmosphere of the planet, and you try and do that with as much fidelity as you can.
On one hand, it’s an easier problem because laws of physics are known and agreed upon. On the other hand, there’s a lot of things like clouds and heat transfer between land masses that are actually not well understood. So even there, they have to do a lot of heuristic reasoning to make things work right. And plus, it’s a good story because it involves von Neumann, his wife, and Jule Charney, and people working around the clock to build the first forecast, which took twenty-four hours to make, and it was a twenty-four hour ahead forecast.
Anyway, very interesting history. But the point is we could do that with economics, but we don’t. Why? Because the dominant paradigm that we discussed is not well suited. You could give them all that data. They wouldn’t have any idea what to do with it. Whereas if you’re taking our approach of simulating the economy, we’re doing pretty much what the weather forecasters are doing. We build a simulator, and we want the best data we can to feed into it, and I’m pretty convinced it can do a lot better than what we have.
You know, key fact, weather forecasts maintained almost the same accuracy for a century. And then finally in the 1980s, with the help of Ed Lorenz and many others, the physical weather forecasting broke through. But it took them thirty years from 1950 to 1980 before they could do that. Now we have the advantage that computers are a billion times more powerful. Actually, going back to the fifties, probably 10 or 100 billion times more powerful than then, and we have vastly more data. It’s much more possible to do this now than it was then. I think we have a big jump on weather forecasting because the economy doesn’t have as many moving parts as the weather, many, many orders of magnitude less. Even if we don’t really have laws of economics like we have laws of physics, we still know enough about the way people behave and the way institutions function to build a simulator that should be able to make predictions that are a lot better than the ones we make now.
Jim: Yeah. You quote in the book, I don’t remember the number off the top of my head, the amount of money that’s spent on weather forecasting. It’s significant.
Doyne: Yeah. My recollection is that we spend $5 billion a year on weather forecasting, and people have estimated it has $30 billion a year in financial value.
Jim: Kinda surprising it’s only $30 billion, actually. You think about things like agriculture and things. It might be more, but yeah. But it’s certainly significant.
Doyne: Yeah. Well, I guess the question is how much value can they extract from knowing – if you’re growing crops, well, because you can’t look ahead a year. That’s one of the problems. They can’t tell you this is gonna be a dry year.
Jim: A good year to grow corn.
Doyne: They can tell you three days from now, it’s likely to rain.
Jim: I have a farm, and we lease out the land to a local farmer to take care of it. And the current guy is good, but his father was even better. In all the years they made hay in our place, he never had his hay rained on even once. He knew when to cut hay. He knew when he had three days of good weather. I do believe he, in the later days, was using the NOAA satellite data feed that was coming in with the agriculturally oriented weather forecast. Quite interesting. Let’s now take that analogy that society gets benefit from spending $5 billion a year from building a detailed model, always more detailed and always more processing every year. I think I believe it was in your book you quoted that our weather forecasts are getting better at about the rate of one day per decade.
Doyne: That’s right. Your forecast after ten years for two days is as good as your forecast was for one day ahead ten years earlier.
Jim: Yeah. And that’s pretty impressive, right? But that’s a significant rate of change in systems as inherently complex as weather. So if we took this analogy and applied it to building economic models, what would be your prescription? How what should we be thinking about? What should we be doing as a society? And maybe give us some guesstimate on how valuable this could be.
Doyne: Well, I mean, first of all, we’re actually starting to do this now. I’d moved forward since the book was written. We don’t have nearly the resources we would like to have. How one would approach this is to really say, “Okay, what does an economy consist of? Households? How do households make decisions?” You really plunge in and look at how different kinds of people make the decisions they make about what to consume or how much education to get or where to work. Similarly, you do the same for firms. You do the same for banks. You do the same for all the major institutions, and there aren’t that many. Then you put all your knowledge into the computer and simulate it.
Now we’re doing this on a couple of levels. We have a simulator that can simulate a country. We’re getting close to being able to simulate the whole world, but we can at least simulate countries. We’ve simulated 38 different countries, and we’re making decent forecasts now. And we’re just simulating all the details. We have a couple of million individuals organized into the order of 800,000 houses and 50,000 representative companies, and we do a simulation.
Now, you asked what value could come from it, so there I could do a little back-of-the-envelope calculation. We lost the order of $10 trillion from the 2008 crisis, maybe 20 or 30, but at least 10. So if we could avert that crisis, obviously that would be hugely valuable. If you had a $100 million project that could avert a $10 trillion crisis, I think if I’m doing the numbers right, that’s a factor of 10 to the 5 payoff. Now, of course, that’s over-optimistic because we might not work. So let’s say it only has one in 100 chance of working. Right. There’s still a factor of a thousand payoff. And let’s suppose that we don’t actually get rid of the crisis, but we ameliorate it so that it’s 99% of what it was. We knock off 1% of the harm. We’re still giving a factor of 10 payoff.
Seems valuable to me. It seems like given that nobody’s done this or even come close to doing it, it’s worth doing. I’m pushing that, spending a lot of my time these days trying to convince people to put some money behind doing that. In the meantime, we’re doing something else that’s even more concrete and more feasible. We’re building a model of energy investment, and we’re doing it by building a model at one-to-one scale, a literal model.
So there’s 30,000 energy generating companies in the world. All 30,000 of those companies are in our model. Those companies own 160,000 assets. All of those assets are in our model, at least in the initial condition. Every year we look at, we simulate the energy system. We figure out how much energy each company supplies and at what price, and therefore, we can estimate their revenues. We can estimate by other means their costs. Therefore, we can estimate their profits.
So that’s already pretty extreme, potentially extremely valuable. But then we can also simulate their investment strategy because companies make decisions in a fairly transparent way. Not always, but they look around at possible projects. They look at the returns on the investments, and that will determine whether it is decided to build a solar farm or another gas-fired plant. So we think we could also simulate that and then use that as a guide to help us go through the energy transition.
The thing I love about this project is we have all the micro data we need. We have twenty-five years of data where we’ve got to see each firm adding each asset as it goes along. We have the data about each asset to figure out how much it’s gonna participate in the electricity system, and we can see all the investment decisions that got made. We have more than 100,000 investment decisions in the dataset, and we have a kazillion decisions about electricity and oil and gas in the dataset in terms of the day-to-day operation. So I’m quite optimistic this is going to work at the end of the day. If you think about the macro model I mentioned earlier, we can build out from this in that direction because for this model, we’re really putting in a lot of domain details about how industries work. What’s the difference between a gas industry and the oil industry and the hydropower and so on? And so there’s just a lot more embedded knowledge in this model by orders of magnitude than any economic model.
Jim: Let’s drill into this climate thing a bit because if we step back and look at the history of humanity, how we manage the transition to stopping using the atmosphere as a dump for greenhouse gases, it’s probably the question for humanity, assuming we don’t do something really stupid with nuclear weapons or biotechnology or AI. But the one that’s gonna crush us in the end if the other ones don’t is climate change. Right? And finding the right road to climate change that doesn’t bankrupt us on one side or take too long on the other ought to be, though it’s not for despicable reasons, at the very top of our political agenda. How does your work potentially address choosing the trajectory for humanity?
Doyne: So we did another body of work, not necessarily even complexity economics, which is just to collect a lot of data about technological change. How fast do things improve? How fast do costs come down? We’re all familiar with Moore’s Law in computers, but it turns out lots of other technologies satisfy something like that, although rarely at such a rapid rate. Similarly, technologies follow S-curves in their deployment, which have a very universal shape.
We’ve been gathering data to understand both of those, and used that approach back in 2010. I used that approach to predict what the price of solar PV energy would be in 2020 and said, “Well, it’s gonna be as cheap as coal-fired electricity in 2020,” and everybody said I was nuts, and that was right. Why was it right? Because I was just smart enough to plot the data on log scale and draw a line through it. It was not a brilliant act, but they just weren’t doing that.
I think one of the things you see is that it’s the result of all this research, that solar energy in particular is likely to get very cheap in the future. We predict the median cost of solar energy in 2050 will be $3 a megawatt hour, and so it makes it more than a factor of 10 cheaper than it is now, and will make electricity cheaper than it’s ever been. We see batteries coming down quite rapidly, and then there is solid-state energy storage hanging out in the wings. So I think it’s gonna happen faster than most people think, and we’re just damn lucky that we have the technologies that we need coming along right when we desperately need them. I mean, we’re still gonna have at least two degrees of warming, so dramatic change to the earth, but hopefully not runaway outgassing of the tundra. We’ll see.
Jim: Yeah. It is very optimistic because I read that. I mean, it appears that our politics is just utterly broken on climate change. Nobody has the guts to do anything, even though the clear solution is pretty obvious, which I talk about from time to time, which would be an escalating carbon tax or greenhouse gas tax at the source, you know, at the production of the fuels, which is then refundable per capita, which would give everybody an incentive to reduce their own consumption because of the signal at the pump, but at the same time would be net neutral with respect to the flow of economics and actually would have an egalitarian purpose because rich people burn a shitload more fuel than poor people do. So, like, 70% of people would make money off of this deal. But yet even this obvious simple solution, which has been endorsed by hundreds of economists, no politician will even mention it today. Right? But anyway, rant off about the ridiculousness of ignoring climate as a policy, but perhaps your findings would indicate that’s right, that the market will take care of the problem.
Doyne: Well, the market is not gonna take care of the whole thing. Market’s helping a lot. We’ve got the wind at our back finally, but we really need to get better power storage, and most of all, we need better grid, and grid requires governments to build. So you have to have some willingness to support really rolling out the grid. You know, we actually estimate in net present value terms a net savings of money by transitioning quickly of the order of $11 trillion, but we do need government to allow us to really increase the grid and do more things with electricity and change the economy in that regard.
Jim: Of course, a lot of the changes needed to be done on the grid are not actually government spending per se as regulatory reforms. I mean, it’s almost impossible to build a power line today in the United States without a zillion NIMBYs coming out and saying, “Oh, no.”
Doyne: That’s right. And we need to fix that. The US may end up… I think the US is gonna be the big loser on this and many other things that are building these days because, well, as we all know, the Chinese are taking over these technologies, and these are the technologies of the future.
Jim: Yeah. Fortunately, learning curves these days are relatively transportable across boundaries. So even if we’re just a couple of years behind the Chinese in the technology of solar cells, big deal probably. Right? That’s one big difference between now and the past where these artisanal skills would take a generation to propagate. Now, two years, we buy one of theirs, take it apart, build our own. Right? We can be fast followers, and that will…
Doyne: I’m not…
Jim: …be just fine. I’m not too worried about that aspect of it. All right. Well, that’s interesting. So now let’s go on to the kind of conclusion of the book, which is if we take these ideas and extend them further, into creating essentially a huge model of our society and our economy and becoming what you called a conscious civilization. Let’s talk about that a little bit.
Doyne: Well, so the idea of conscious civilization goes back to building, making analogies like if the economy is the metabolism, I argue that the financial system is your gut brain. You know, you actually have neurons the size of the cap in your gut that help you digest. Well, do other things too, like we built scientific models, and we already have used those to help us think about danger, whether it be an asteroid whose trajectory we can calculate, and telescopes allow us to see them, or climate change. Right? It’s a pretty amazing tribute to science that we can look a hundred years ahead and see that we’re headed for this train wreck, even if we have a hard time acting on it as a society. But increasingly, so far that’s better understanding of the physical world. Our collective understanding of the physical world brings us a lot of benefit. But maybe we can even start to collectively understand ourselves, and if we could make better economic predictions that could then guide us to make better decisions about things like which technologies to invest in or how to run the economy, then we are starting to become aware of ourselves. Why do I call it consciousness? Well, I have a very simple view of consciousness. Consciousness is a model of oneself. And so the thing that makes me conscious is I know I exist, and I can model what I do, and how I relate how I alter the behavior of other people. And so similarly, if we can understand ourselves that way, then we can become more conscious as a civilization.
Jim: That would be huge. That would be great. And, you know, to your point earlier, the expense is relatively modest. Right? If we spent $5 billion a year, we’d make some serious progress.
Doyne: Oh, no kidding. I’d settle for a budget much lower than that, and I can tell you the more I do this, the more convinced I am that pumping money into this is gonna pay off big time.
Jim: And as you pointed out earlier, fortunately, there are some probably harvestable things that the market can actually fund the development of, right? I mean, you would think that-
Doyne: My strategy for making this happen now is to do it through the private sector by building a money cow that we can just expand and expand, and hopefully meanwhile give this technology away to the people who really need it. We want to make our tools available for governments at a low price, but for companies at a much higher price. And I don’t wanna get stuck in a situation like I was at Prediction Company where we have to keep everything secret, locked up, and just make profits for some Swiss bankers.
Jim: Well, thank you for this very interesting conversation and this compelling vision for the future. Really a pretty optimistic vision if we can avoid gross stupidity. But talking about Doyne Farmer’s book, “Making Sense of Chaos: A Better Economics for a Better World.” I should add, even though the topics seem pretty abstruse, it’s actually a very nicely written book, very accessible. You don’t need to be a complexity scientist. You don’t need to know calculus. Anybody who’s listening to this podcast could benefit from reading this book, and I really recommend it highly. I’d like to thank Doyne for coming on here and talking about it.
Doyne: Welcome, Jim.
Jim: Alrighty.