The following is a rough transcript which has not been revised by The Jim Rutt Show or by Eric Smith. Please check with us before using any quotations from this transcript. Thank you.
Jim: Howdy, this is Jim Rutt and this is the Jim Rutt Show.
Jim: Listeners have asked us to provide pointers, some of the resources we talk about on the show. We now have links to books and articles referenced in recent podcasts that are available on our website. We also offer full transcripts, go to jimruttshow.com. That’s jimruttshow.com.
Jim: Today’s guest is Eric Smith. He’s a researcher at the Earth Life Science Institute in Tokyo and is in the biology department of Georgia Tech. He’s also external faculty at the Santa Fe Institute.
Eric: Thanks Jim. It’s great to be here.
Jim: It’s great to reconnect with you Eric. You were one of my favorite people from my time at SFI, when I was a full-time researcher at SFI. Rather often, if I just had a conundrum, I would go talk to Eric and it was amazing. No matter what the area was, Eric could provide some interesting insight. Eric’s by background and training a statistical physicist, Caltech and then, his PhD from the University of Texas, but in fine SFI fashion, he’s used those tools to invade many other fields, including at least origins of life, the nature and origins of language, money and finance as institutions, evolutionary game theory and ecosystem sustainability. He is the author of two deep and important books, The Guidance of an Enterprise Economy with Martin Shubik and The Origin and Nature of Life on Earth, the Emergence of the Fourth Geosphere with my good friend, recently departed, Harold Morowitz.
Jim: Before we jump into these scientific topics, Eric has also spent a considerable amount of time over the years in China and has recently refreshed his knowledge of China with a trip over there. He’s got some interesting perspectives on that whole China thing. Eric, what do you think of China, what do you think about it?
Eric: Well, anybody who’s been there will tell you that the pace of change is striking from people who were born there to outsiders. That certainly is right. The thing that’s interesting is what’s changed and what hasn’t changed. Of course, I mean China’s not one thing, a country of 1.6 billion people is many things to many. In the business sector, you can see the concentration of wealth and the confidence that it brings. You can see that China wants to be back in what it considers its natural place as kind of the central front country in the world and everything they’re doing is aware of the wish to get back there. There are certain things from the old days that have not changed. My contacts are mostly academics and there’s a cheerfulness and a sincerity and a sense of goodwill in trying to address deep questions and just enjoying community and enjoying collaboration. That’s the same as it always was and a friendliness and cheerfulness and welcoming people there. It’s been a good reconnection after a while away.
Jim: The nature of the Chinese system is something I’m very interested in. It’s not really exactly like anything we’ve ever seen before. Do you have any insights into the nature of what we might call the Chinese social operating system?
Eric: Well, it’s intriguing, right? The Chinese idea of good governance since time immemorial has been benevolent dictatorship and that shapes a lot of what they think can work on a lot of what they think doesn’t need to be tried. They see the innovations in the manipulation of opinion that are happening worldwide. I think they don’t trust civil society as a machine to control that but they think that it can be centrally controlled. This whole engineering project is an attempt to make that work.
Jim: Yeah, it’s interesting. If you look at the history of China, even though obviously it has a very deep history in time across a geographic space, it’s also a classic case of punctuated equilibrium, where highly centralized command and control systems work until they don’t. When they break down, they’ve often broken down very rapidly. This would be an interesting experiment to see if with modern tools, they can make centralized control actually being at least metastable.
Eric: Yeah, this is the thing to ask the experts, who really spend a lot of time there and know the society and the people in it well because I think they will say that there are combinations of centralized control and an attempt at innovation that have always been believed in the past to be impossible. The question is were they really impossible or was that context dependent? I don’t think anybody knows.
Jim: Yeah that’s the question. It is a competing model. There’s no doubt about it. There’s obviously hypotheses that China is still benefitting very substantially from follower syndrome, where they don’t have to invent nearly as much as say the West did, so they can rise rapidly, basically absorbing the technology and expertise of the West. They’ve done it better and faster than anybody I guess, even better than the Japanese. In some areas, they are really at or near the state of the art, like an AI. Have you gotten any exposure to what’s going on in AI in China?
Eric: No, I’m a viewer of that information, like most people who are not specialists in the area. The interesting thing about AI is whether it’s going to be a classic example of Philip Anderson’s More is Different. We understand for instance from deep learning that the ideas that are now dominating the field were not created in recent time. They’re Carnegie Mellon Computer Science Department from the 1970s and 80s, it’s just that those ideas were not viable for the scale of computing and data, particularly training set data and the ability to use it had reached the modern level. China is going to be the world’s largest repository of integrated data in the decades going forward. The question is whether there will be new conceptual horizons that show up with that increase in scale that we didn’t foresee.
Jim: Chinese certainly understand the criticality of data, at least with deep learning though and people listening to this show know, well, I think deep learning is immensely interesting. I suspect at least, it is not the answer to artificial general intelligence, at least not the sole answer. It is clear that the Chinese are going to wire up their society to produce lots more data than say the US is able to do, if only because they have five times as many people.
Eric: You know what I like on this though, I think back to the, I guess it was AlphaGo competition with Lee Sedol in the computer human contest for Go playing. I really loved Lee’s comment at the end of it, where he was saying that of course those had been the most difficult games he had had to play, but that he had never enjoyed playing Go more than in those games because before, he was the best in the world in a style of play that was essentially established and playing the machine, it was opening [inaudible 00:06:46] of play that no human would have opened against him. It was giving him an insight into the game that had not been available to him from anyone before. Apart from the superb character that that demonstrates in the man, I think that’s a good way to look at human-computer interactions that we have all of these big branching structures. The question is when will computational solutions open [inaudible 00:07:11] of play that human conventions were not exploring.
Jim: That’s actually a perfect transition to our first scientific topic because that would obviously be relevant here and that’s origins of life, the transition from the abiotic world to the biotic world. How did you happen to get interested in this field and why don’t you provide a little framing for our audience about what are the questions about the origins of life?
Eric: Yeah, very good. I of course came out of high energy physics and in a period after I had left that field, I was looking for something to do. I was looking at phase transitions, which was sort of the big idea of the 20th century in modern physics, in some ways more transmissible across the sciences even than relativity or quantum mechanics, which were domain discoveries. I was looking at dynamical phase transitions and the sense in which sort of tautologically their engines and thought, “Well, this seems like a good way to understand how biology works,” but biology was not studied from that perspective. It was really hard to get domain information to get up to speed and solve problems. When I came to the Santa Fe Institute and met Harold Morowitz, I met kind of my perfect complement because he had been working in topics in biophysics and physiology for 40 years at the time and said, “I have all these signatures of the biosphere that looked to law like to have been accident, but I don’t know what kind of laws should be used to explain them.”
Eric: What Harold had done was cut to the very simple core of biology, where there were few enough boundary conditions and few enough components that one could think about trying to bring laws of self-organization to bear on them. That was the beginning of a wonderful 20-year collaboration. That has continued to be a foundation of my work to today.
Jim: Yep, I’ve chatted with Harold quite a lot about origin of life, enough for my simple mind at least to get its head around a little bit. The thing he always told me about, he had this amazing picture on his walls. He was what I would describe as a metabolism first person. Could you tell us what that means?
Eric: Yeah, I can tell you what it means without endorsing it in the classical sense because in some sense by now, these things are tropes. They’re oversimplified positions and in the oversimplified form, all of them are somewhat absurd. Metabolism first was in a sense a response to a different absurd position, which you could call a control first. Starting I guess in the 1950s, Francis Crick recognized that a lot of control flow in modern biology is kind of top-down. You have the phenotype, the kind of organism that lives or dies and that life or death selection preserves what genes are in the population. Then, through a very complicated control machinery, the genes are expressed in proteins and RNAs and other components, which then make up the organism, which will have a phenotype in the next generation.
Eric: One approach to origin of life was just to adopt Crick’s central dogma for control flow, the idea that control flows from DNA to RNA to proteins and try to map that all the way back to the beginning of life. If you need to do that then you have to say, “Who are the genes, what’s the phenotype and what are the Darwinian individuals that can be selected?” There was an attempt to put all of this heavy burden on the narrow shoulders of RNA molecules, making them the individuals that are both genotype and phenotype. The thing is chemically that’s very hard to do because RNA is complicated and it’s not clear what kind of a system it should first appear in, but also when you look at the small molecule chemistry of life, it’s very law like, it’s very economical and it’s economical in ways that would not be explained just by an arbitrary control flow from RNA downward. What the metabolism first point of view was an attempt to do was to say that a lot of those laws of deep small molecule biochemistry were actually inherited from regularities in geochemistry that were there in place before the biosphere formed. The biosphere basically inherited those and formed itself around them.
Eric: I would say the modern version of metabolism first is an idea that life is not one thing and its emergence was not one event, but rather it’s a kind of a hierarchy and a cascade of events. In the events before you have something recognizable as fully integrated life, you have increasingly organized states of geochemistry. They’re organized around [inaudible 00:11:43] of least resistance. What we want in understanding the origin of life is how those [inaudible 00:11:48] of least resistance sort of set a template for biochemistry and then that template was strong enough and simple enough that it allowed control structures, like genes and the central dogma to come into existence.
Jim: Subtitle of the book is The Emergence of the Fourth Geosphere. Could you unpack that for us, what were the first three geo spheres and how is biology the fourth geosphere?
Eric: This is the elegance of Harold’s insight on how to break these problems down. He was quoting Vladimir Vysotsky in the use of the term geo spheres. It’s a little bit of an antique usage. Well, the word geosphere is used today, if it’s used at all, it’s not usually used in Vysotsky sense, but when Vysotsky first brought the term into usage, he was interested in trying to get some traction over the great complexity of geology. He said, “Look, if you have a planet, you can’t ask what one molecule does because one molecule goes all around or one atom goes all around and there are fluxes and currents and exchanges.” How are you going to make sense of this big complicated system? The way you start is to realize that at the very least, you have three big central systems. You have the solid core of a planet.
Eric: If it’s an Earth-like planet, you have a distinct subsystem, which is its liquid ocean and then, for most planets, you have an atmosphere. While each of these is complicated in itself, there are certain things that tie them all together. The solid part of the planet is solid state. Its chemistry and its physics are those of solid materials. The ocean is liquid state. Its chemistry is the chemistry of what can happen in water, which is permissive in some things and restrictive in others. Then, the atmosphere is gas phase and directly coupled to the stellar irradiance and the chemistry and the state of matter is different in the atmosphere than it is in the ocean or the solid.
Eric: Vysotsky referred to these as the atmosphere, the hydrosphere and the lithosphere. Nowadays, of course, lithosphere is the top of a rocky part of the solid part of planet, so the usage has changed in that way. Harold’s insight was to say that we should view the biosphere the same way we view any of the other geo spheres. It’s a new state of matter. It’s dynamically defined. The rules in it are different and distinctive from those of the atmosphere or the hydrosphere or the lithosphere. It is more tied together by the integration of its processes than the exchanges that it makes with any of the other components of the planet.
Jim: How does the fourth geosphere emerge and what is the mixture between the water and the land and the air as inputs into this fourth geosphere? I know in talking to Harold, he wouldn’t even rule out that maybe life originated deep in a mile below the Earth surface, in the rock.
Eric: Yeah, the subsurface water alteration zones are a very, very good place to do planetary chemistry. In fact, since life is driven by electron transfer under sort of voltage differences, it’s very difficult to find any other place within the planet, where you have so many conditions that are good for providing those electron transfer potentials and also, the catalytic environments to make use of them. I think that idea is still very much in play. There’s a quote and I can’t remember the source, I should, he’s a member of the astrobiology community, but it’s pithy and elegant. The quote was that the origin of life is not something that happens on a planet, it’s something that happens to a planet. That was really Harold’s insight that the origin of life should be understood through the emergence of a biosphere. It’s not something that’s contained within individuals. It’s rather a transition of systems, which means that it’s multi-component, it’s robust and it changes the dynamics of everything around it. You can say that a biosphere came out of chemical organization, probably from rock water interfaces in the chemistry that they do, the disequilibrium of the atmosphere against the deep bulk of the earth. As it formed, it became its own system with its own sort of robust central tendencies. That’s what he means by the emergence of a fourth geosphere from a planet that formerly had had only three.
Jim: Okay, how about your thoughts on how that happened, more to the specifics when we have the black smoker theories, we have deep rock theories. What are your insights into where and how it might have happened?
Eric: Yeah, my first insight is that we shouldn’t try to spin scenarios at this point. When most people work on the origin of life or have worked on it traditionally, they’ll have really an insight about one process that occurred in one domain of chemistry at a particular stage. That’s the part where they a really good idea, but they say, “Well how do I ground this idea, how do I know if it’s a good idea? I should try to connect it to the things that happened before and the things that happened after and argue that the things that happened before plausibly set up the case I want to make and that its outcome plausibly set up the things that should have happened later.” Problem is they don’t have good ideas about what happened before and what happened later. They spin these long stories that are mostly meaningless. A meaningless story doesn’t actually improve the strength of the one place, where they do have deep insight. I think for the origin of life, we have to start a new project that we’ve not known how to do before, which is how do we reason about the good sense of a claim that’s very incomplete when that claim is going to stay very incomplete for a long time. This is a new problem in science and processes that are very, very complicated and have a long history.
Eric: All that said, I do think there are some general things that we can recognize. A planet like the Earth is a battery. All of the materials from which it formed, the iron and the hydrogen of the early stellar atmosphere, those are very electron donating materials, but once you separate into an atmosphere and ocean and a rocky bulk, the water tends to boil off the top of the atmosphere. That’s not quite the right way to say it. The right way to say it is that the ultraviolet and the x-rays from the Sun split the water and the hydrogen escapes, leaving oxygen behind. That drives the atmosphere away from electrical equilibrium with the deep bulk and you start to get an electron transfer out of the bulk into the atmosphere from which the electrons are then transported off the top as we lose our ocean to stellar breakdown. There’s a net electron flow in this battery out of the deep bulk into the atmosphere and then, out into space. If you look at where that electron flow occurs, it dominantly occurs at the spreading centers that are underneath the ocean basins at this time. This is where new crust is being formed from rock that came out of the bulk and was not recently in contact with the oxidizing atmosphere.
Eric: I still think if you want to look for the organization of organic chemistry, the places where new crust is being formed or the best candidates to look for that. It’s a controlled chemistry, it’s a mild chemistry. It’s not the violent chemistry of ultraviolet irradiation and in being mild and relatively low energy, it looks much closer to biochemistry than the chemistry you would do with UV light. The important thing about low energy chemistry is that if you don’t have catalysts, nothing happens at all. If we’re looking for an origin of life at rock water interfaces, where new crust is formed, we really need to understand the catalytic environment to be able to predict anything at all.
Eric: Now, if you don’t mind that as a general context, then you could say, “Are there organizing centers in biology that look like the foundation for everything?” If there were not, we’d have no story to tell, but it turns out that there are. Everything in biochemistry that’s formed comes through a small system of only 11 organic acids. The biggest of them only has six carbon atoms. This is called the citric acid cycle.
Eric: On Earth today, it exists as a carbon fixation cycle, but even in the organisms that don’t use it to fix carbon, it still is the starting point to synthesize everything. If you were looking for a foundation for all of biochemistry, I think there’s no better evidence of a place to look than the evidence in modern biology for the citric acid cycle. It’s really funny, if you want to ask, “In all of this big planet, is there one molecule that is the center of metabolism and the center of carbon fixation?” It’s acetic acid, it’s vinegar. That is the central molecule for the deepest part of the biosphere. I think it’s really fascinating is that I’ve been looking more recently at the organization and the amino acids sort of continuing our interest in the genetic code. You could say, “Okay, if acetic acid is the core molecule for fixing carbon and the start of biochemistry, is there a central molecule for the beginning of everything complicated?” The answer to that looks like yes too. That central molecule is pyruvic acid. Biology is full of these little core linchpins. In the way Harold and I and others who are of similar mind, looked at the origin of life. We look at these as the circumstantial evidence for where to put the search light to do small molecule chemistry.
Jim: That is really interesting. Now, when we talk about catalytic cycles, it brings to mind another guest we had on the show earlier Stu Kauffman and his autocatalytic networks. Can you maybe compare your ideas to his and how they’re different, how they’re similar?
Eric: Yeah, so there’s a truism that one has to start with. If we were just doing equilibrium physics, you can have organization that just sits there, a diamond will just sit there for a very, very long time because it’s a stable configuration under some circumstances and metastable under others. If you’re looking at dynamic stability, like we have in the biosphere, everything falls apart. The only kinds of patterns that can persist are the ones that continually renew themselves. Now, renewal can happen by a different process than the process that led to the patterns origin in the first place, but the simplest kind of renewal would be one in which the mechanisms that generated the process at the beginning are those that continue to rebuild it as we go.
Eric: Now, if you ask what this looks like in material, well, organization in material means, you’re putting a lot of your material into a few forms, a few biomolecules, a few genes, a few regulatory circuits. Falling apart means that that material drifts into the places that are not part of the organization. You have to ask, “What then moves material back into the core of the organization?” That’s what autocatalysis does. Autocatalysis says, “We’ll take things from a general and diverse environment and we’ll concentrate them into a few outputs,” but those outputs will be like a kind of flypaper. They’re sticky for the general environment and when the environmental components stick to them, they get converted into the same material that the flypaper is made of. That’s what Stu’s insight is and that’s also what we recognize to be true.
Eric: If you look at biochemistry, it’s full of autocatalytic cycles and those are the things that allow biochemistry to remain integrated and purified in an environment, where things are constantly falling apart. The difference between Stu’s approach and ours is that Stu starts from a very unstructured flat kind of a combinatorial medium. It can be boolean networks. It can be abstracted models for gene regulatory networks. He wants the laws of large numbers to give him feedbacks in those large networks. That can be a good thing to do if you’re looking at for instance gene regulation. If we look at biochemistry, the important thing about biochemistry that’s new in science is that chemistry itself is a very structured medium. In physics, we have only understood collective and cooperative effects and phase transitions in systems that have a lot of symmetry. When you get down into the dynamics of chemistry, you have systems that have a lot of structure. The leading edge is going to be to understand what collective and cooperative effects look like and how one searches for the natural feedbacks in autocatalyses in a highly structured system like organic chemistry is.
Jim: It actually has its own structure. One of the topics that comes up in complex, adaptive and emergent systems is the fact that they’re always far from equilibrium. As you said in static physics, a diamond just sitting there that’s one thing, but the systems we’re talking about need energy flowing through them. I know some of the SFI people don’t like his formalism, but [inaudible 00:24:34] idea of dissipative systems for instance. Talk a little bit about the energy fluxes and how they push the system far from equilibrium and how that is involved in the evolution from the abiotic to the biotic realm?
Eric: Yeah, very good. I don’t want to downplay the importance of energy flow because in practical problems, it often is the thing that determines, but I don’t want to overplay the importance of energy flow because it’s only incidental and the fact that we’re talking about material systems that gives energy such a large role. If we really open our minds and look at the statistics of self-organization in the big picture, energy becomes only one of the flows that can become important. There may be other things that are not the flows of something that are important. Just put a flag on that that we can talk about the particular, but we want to keep our minds open that the concept is more general. In thermodynamics, energy is one of the quantities that tells you how much of your state space is available to you. Other quantities are what are you made of, how much of each chemical do you have, how much of each element, but also volume, are you partitioned, how much space can you take up?
Eric: In equilibrium, the internal energy you have in your system is essentially set. The volume you have, your material composition is set. The part of your state space you can explore is fixed. As you go away from equilibrium, you can couple one system to boundary conditions that would not be in equilibrium with each other, which allows energy to then flow in through one boundary and flow out through another. If we’re talking about the origin of life, you can have energy flow in in form of one molecule and flow out in the form of some other molecule. Energy flow in the Prigogine sense remains important because the more information you provide about the boundary conditions, the less freedom your system has to explore, another way to say that is that the more ordered the system can be. A driven system always has more information in the boundary conditions than one that’s not driven because if I’m at equilibrium, all of my energetic boundaries have to be in equilibrium with each other. I can go away from equilibrium in many ways. It’s kind of like Tolstoy, all equilibrium systems are more or less alike. Every disequilibrium system is disequilibrium in its own fashion.
Jim: I like that.
Eric: I can move the energetic boundary conditions out of equilibrium with each other in a lot of ways. I can put energy in different molecules. I can take it out in different molecules. I can have temperature gradients. Once I do that the restriction on the system’s freedom can take many more forms than the restrictions I had available in equilibrium. This is what Harold meant when he said, “The flow of energy through a system tends to organize the system.” Now, the interesting thing is I would say Prigogine does not go far enough from equilibrium because if you look at the work that he’s doing and the work that has continued in the hands of many very fine workers, who followed him, they still want the equilibrium state variables to be the main thing that gives order to the system. They want pressure and temperature and chemical potentials in the [inaudible 00:28:12] to re-equilibrate quickly, so that then they can study the information in a system that moves slowly and episodically in that fast equilibrating [inaudible 00:28:22].
Eric: We can do something even more radical than that which is that we can ask what is the native thermodynamics of any stochastic process driven in any way at all? Then, you contain the Prigogine class as a subsystem, but you can also go much more general. You can go into what I would call the thermodynamics of rule-based systems, the beautiful things that Walter Fontana and his collaborators work on for systems biology or the theory of chemical reaction networks regarded as systems in their own right rather than just expressions of the surrounding physics. I agree, but I want us to be more radical and more exploratory even than Prigogine’s generation could be.
Jim: Excellent, even stronger perspective. You said earlier that this origin of life work is interesting and kind of different in that we only know a little and the box is big. Could you relatively briefly say what you know or what you think you know about what’s inside the box?
Eric: Yeah, great. It’s a bunch of little islands, where we have repositories of information and about those, we can hope to say a lot. Then, there are these big areas, where you don’t even hear good ideas being floated. I mentioned the core biochemistry, the carbon fixation pathways, the chemistry surrounding the citric acid cycle, the reactions that tend to get used and how they’re used. We have lots of recorded information there and probably, it’s not even recorded. Probably, this is actually an extension of the periodic table that is being expressed by the rigidity of our biosphere. You go up a level and look at the ribosome and all of a sudden, you have an extraordinary repository of history about RNA and early peptides and the emergence of translation. We don’t know what process that history froze into place and yet, we know an amazing amount about what it froze into place. We’ve been given the answer to a final exam, but we don’t know what the questions on that exam were. That’s another place, where we have a streetlamp giving us tremendous light. A place where these two intersect is the genetic code. Traditionally, the genetic code has been viewed as a problem of assignment or shuffling, where people take the amino acids as somehow given and they take RNA and the translation system as somehow given. Then, they look for how many ways you could assign amino acids to codons in the translation system.
Eric: I think that’s a wrong view. I think it’s not nearly dynamic enough because I don’t think the amino acids were given. I think the process of becoming biotic was integrated with the process of becoming energetically active and discovering how to fold. I think that when we understand the origin of the translation system in the genetic code, we’re going to realize that regularities of deep biochemistry, fundamental physical regularities of the problem of folding and how easy or hard that problem was different for peptides than for RNA kind of bootstrapped themselves up. What we see recorded in the genetic code is the way each system got as far as it could and then, hit a wall and a discovery in one of the other systems that right now looks like it’s a different system opened a door for the next innovations. They jointly went through that door. I think we have these little islands that have information, but where our own disciplinary boundaries have kept us from seeing what that information is about.
Jim: Interesting. Now, this comes right next to one of my favorite topics, one of the topics we talked a lot about on the show is the Fermi paradox. Like all good 14 year old nerds, I said, “Oh yeah, got to be 100,000 intelligent species in the galaxy.” Well, there’s one possible pruning rule in kind of Harold sense that I wonder about. This is something I had like a six-hour conversation with Stu Kauffman about and we both walked away saying, “Hmm.” That’s the following. My home field is evolutionary computation and one of the things we know about evolution is if the information fidelity between generations isn’t above X, depending on the nature of the system, we have what’s called the error catastrophe, where the ability for evolution to produce very much is not very strong. We have at the core of the genetic code, DNA, not just a highly complex structure, but most importantly, a rather complicated machinery of error correction. The two together give us a high fidelity information substrate for evolution that’s well above the error catastrophe and allows evolution to produce things like ribosomes or maybe they co-evolve, but how one can imagine chemical evolution prior to this high fidelity information transfer system existing, have pumping up through evolution high enough to create a machinery of this level complexity is quite perplexing. Does your work have any thoughts at all about this problem?
Eric: Yeah, this is one of those places, where you stand on the shoulders of giants and you inherit insights that are very good from many sources. My mind’s been out of this space for a little bit, so it’s going to take me a minute to get my sources back in line. A really great early thinker on this was Herb Simon, one of the early systems thinkers across fields. Simon has a set of classic arguments about the architecture of complex systems, where he says, “Kind of like politics, all error correction is local.” That’s not an incidental actually. This is true for any distributed system and this is where decentralized banking came from in England. If you have a complex control problem, the higher order systems that are your controllers are often made by assembling building blocks at a smaller scale, but that means that they’re coarser and the information bandwidth through the interfaces is too small to keep up with all the things that can go wrong internally to the building blocks. Herb Simon used to argue that if you see a complex system in the world, the only plausible mechanism to have created that system is one that has integral error correction in the subsystems along the way that creates sort of stable intermediate states of assembly, which then provide platforms to explore from for later assembly.
Eric: Herb Simon uses the analogy of watch assembly. You don’t put a thousand pieces in a watch together, holding them all in place and then, put a screw on the top. Instead, you assemble components of 10, put each on the shelf. You put 10 of those together to make assemblies of 100. Put them on the shelf. Put 10 of those together and that’s your assembled watch. If you were never interrupted, it wouldn’t matter that you do those things, but in a world where you really are interrupted, it does matter. He used also the example of the Alexandrian Empire, where Alexander’s Empire was only possible in a relatively sophisticated state system, where the states could be left to manage their own internal affairs. In looking at the origin of life, we are looking for where in chemistry the [inaudible 00:35:35] of least resistance tend to regress toward their own central tendencies, so that your control systems don’t have to constantly herd them back in those directions. The control systems can provide low bandwidth signals to sort of make the major switches between the stable or metastable domains of the traction.
Jim: Do we have any ideas about what was on the pre-side of DNA and it’s error-correction architecture?
Eric: Yes. Let me give only one example, the one could go on for a long time, think about protein catalysis. Catalysis by proteins is obtained by first making a fold architecture and then, putting the side chains of a bunch of amino acids in very particular configuration. That’s a hard thing to do because you don’t get catalysis until you have a lot of supporting machinery. If you want to understand how you could ever have had the catalysis the proteins enable, you would first say, “Is there any catalysis by small molecules that’s doing much of the job within a small assembly that doesn’t require this sequence coordination to happen?” The fascinating thing is that if you look at the reactions that involve RNA and DNA, they often involve what’s called substrate assisted catalysis. What that means from the technical side is that the RNA molecule or the DNA molecule itself is defining most of the catalytic contexts. The surrounding catalysts only needs to kind of orient the components. It doesn’t usually need to have its fingers in the transition state of the reaction, which means that a much simpler surrounding environment can be catalytic for those kinds of reactions.
Eric: It’s fascinating that RNA and DNA would be the place, where those substrate assisted catalytic processes are most common because those are exactly the sorts of chemistry that would need to have been in place to enable the memory systems to get more complicated protein catalysis.
Jim: Very good. What do you see as right on the frontier that we need to learn next in the origin of life?
Eric: We need to learn how to explore chemistry.
Jim: Could you unpack that a little bit?
Eric: Yeah, chemistry is a combinatorial system that is so big that we don’t know most of what’s in it. We don’t know what happens. We don’t have a system for searching systematically and as a consequence, we don’t have the ability to reason about it the way we reason about any mature science, which is in most mature sciences, we start with a hypothesis. We get up to a point, where we get stuck and then, we have the ability to figure out what we don’t know and to backtrack to the last place where we needed something and take a different path. In spaces that are so big, we don’t know how to search them. What we do instead is we sort of ad-hoc based on the expertise that any given person or community has, we explore some region. Then, if it doesn’t have what we want, we just kind of go back to exploring, so we’re taking potshots. This is not to denigrate people’s expertise. The expertise of chemists is everything we have to go from and it’s extraordinary what they know and how they can reason. The space of everything dynamical that can happen in chemistry is so much bigger than anything we have understood before in science that it just requires new methods that have not existed before and need to be created.
Jim: Is there anything in the emerging deep learning, machine learning space that looks like it might be applicable?
Eric: There is work in that area. I’m not involved in it. I know there’s some work being done by German chemists, the discovery of Hamiltonians. There’s work from our wonderful colleagues at the University of Southern Denmark and the University of Vienna, doing what they call pathway detection I think. This is given a set of generating rules and a set of inputs and products you would like, can you discover a pathway that will get you to those products with the generating rules from the inputs you have and then, can you optimize that pathway by one or another criterion?
Jim: Well, thank you Eric for that amazing deep dive into the origin of life. We’ll keep our antenna up for what comes next. Let’s move on to our next topic. It’s amazing that Eric has deep work in all these areas, money and finance and institutions. My own interest is most strongly in monetary systems, so if you want to tell the story from that perspective first that would be fine, but why don’t you start off by giving a general framing about what you were trying to get at in your work in money and finance?
Eric: Yes, so this was very much a merger of two minds. The leader on this was certainly Martin Shubik, whose interest for his entire career had been the institutional foundations of money and the nature of money, the role of money within the polity and within the society. Martin was a RAND era game theorist and he was an operations research guy, which means he didn’t pursue the economic view of institutions [inaudible 00:40:36] phenomenal. If you look at neoclassical economics, it’s very strongly rooted in preferences and sort of the idea of the decentralized actor. In much of the foundations of that theory, institutions don’t really need to exist. They’re sort of external expressions of the preferences people have and what you really care about is people’s ability to optimize their preferences against the conflicting preferences of other people in some convex possibility set. Martin looked at that and said “That doesn’t make any sense at all in the world I actually live in. The institutions of the world I live in determine what I can know, they determine what I have the agency to do, they determine what communication I have against every other actor. Sometimes, I don’t even act against other individuals, who are peers, I act entirely against institutions such as the government, which is a coalitional structure.” The institutions are the entirely the conduits for what’s possible.
Eric: If we want to understand the nature of money or in fact, the nature of most of our institutionalized interaction with each other, we need to understand the mechanisms that are possible for institutions and also, the historical processes by which there arrived at, which can lead to lock-in. Martin’s interest was money seems like such a sort of a primitive concept and a thing that’s at the center of everything we do. How can we understand why such a primitive concept can come out of all the complexity of the institutional foundations of our interaction? I kind of joined into that thread because that’s very congenial to my sort of physics and mechanical approach to the world. We would look at things like problems of bank runs or problems of self-fulfilling prophecies or self-violating prophecies. From a mechanical nature, how do the permissions and the limitations of institutions enable people or limit people in the collective states they can achieve?
Jim: Good introduction. Let’s start with … I actually did read long ago Martin’s Theory of Money and Financial Institutions, which is a three fat volumes and it’s very interestingly structured from the very simple to the very complex by a whole series of steps. I’m familiar with that general approach. What could you say about what you put in this book about the early stages, pre-institutional society and what were those pre-institutions that may have given rise to the institutions that did arise?
Eric: I don’t know how much you could say we did that was pre-institutional. In some ways, Martin viewed our book as kind of volume four the Theory of Money and Financial Institutions, which was meant to get at the domain of dynamics. This was the place where maybe my presence was felt in terms that will be familiar from our conversation about origin of life. The origin of life problem was largely the question, if you’ve understood a lot about organization and equilibrium, what do you have to learn that’s conceptually new about organization away from equilibrium? Martin wanted to ask the same kinds of questions in economics. He wanted to say, “If you’ve understood a lot about the stable functioning of institutions, what do you need to learn that’s new about dynamics when you’re trying to deal with transients or adaptations to changing environments or the origin of institutions themselves?” One of Martin’s favorite rants was about how people had refused to understand Keynes until they were able to misunderstand him. There’s this famous passage from Keynes about how economists set themselves too easy, too useless a task if they only tell us that after the storm has passed, the sea is flat again. This long-run is a misleading guide to current affairs. In the long run, we’re all dead.
Eric: Martin felt that economists had missed Keynes’ main point that it’s all in the dynamics, until Hicks came along and reformulated Keynes in what’s called ISLM, which is framing Keynes in terms of equilibrium of investment and savings. Martin felt that that had annihilated the whole spirit of Keynes’ point. What he wanted us to do with what he called them strategic market games was to re-instill the dynamical spirit of Keynes. If you have a system that doesn’t have stationary structure, how does a game create the solutions that are available to you?
Jim: Perfect. Give an example, a real-world example, if you have one?
Eric: Sure, so the Hahn paradox for the value of money. If we only use money in trade, which is certainly the case [inaudible 00:45:19] approximation for fiat money, but to some extent, it’s a decent approximation even for gold. Most people don’t have direct use for gold, but gold is a store of value and it’s a medium of exchange. What determines the exchange price of gold against things you actually need, like food and cows and houses and shoes and stuff like that that number should be arbitrary. If gold is infinitely durable and there’s never a time when [inaudible 00:45:46] nothing anchors that arbitrariness. Now, if you have consumption value for gold, but it’s in the far future, so maybe you can use it for jewelry, but that’s kind of a luxury. Maybe you can use it to pay taxes, but that’s a long horizon. Then, you can say that that consumption value will anchor the exchange value of gold. Now, we have a problem in the order in which we take infinities. If I take the future infinitely far off, then it seems that the value of gold should become arbitrary. If I take the future close in, then you’re placing too much of a regulatory role on things like the payment of taxes and fiat money.
Eric: Then, the question was how do we understand the actual fluctuations in the trade value of things like fiat money or durables in the real world, where we live? My contribution to that topic was that in the real world, fluctuations have a way of creating stabilizing effects because they put temporary uncertainties into the system, where they bring actors into the system, who would not be there if the sea were always flat. In fact, the right way to look at the long run is that in the long run, it’s not the policing by very weak regulations that stabilize you, it’s the fact that the spectrum of fluctuations on the set of uncertainties that they introduce become the stabilizing force that take over from very weak policing actions in the long run.
Eric: One of the chapters in our book is a model of the Hahn paradox that’s along those lines.
Jim: To try to chip into this a little bit, is that essentially saying that “We need uncertainty to keep the game going and if the game became too transparent, it would stop working?”
Eric: I guess that’s possible. If you don’t mind, let me give you an analogy that captures the math. It’s not obviously about money, but it comes from evolutionary game theory. I should tell you there’s a third book in addition to the two you mentioned, which is called Symmetry and Collective Fluctuations in Evolutionary Games. That was written with my co-author Supriya Krishnamurthy.
Jim: Well, now I miss that one goddamnit, but I’m going to have to order it.
Eric: It’s an institutive physics book. It’s sort of library only, which means it’s too expensive. They know it was never going to sell on the popular market, but we were interested in things like what stabilizes the prisoner’s dilemma, one of these famous games, where you can always cooperate or you can always defect or you can be a conditional cooperator, like a tit for tat type player. In the traditional formulations of the prisoner’s dilemma, the conditional cooperator in an environment of full-time co-operators becomes invisible. There’s nothing to protect his self-protecting strategy because everybody’s nice all the time. You don’t have a way to anchor the composition of a society between the always nice and the conditionally nice. If you put fluctuations into that, so that occasionally some nasty guys come into the system, their presence will distinguish the conditionally nice from the always nice. They’ll keep enough conditionally nice people in the population to stabilize it against unraveling to the invasion of defectors. When you formulate this problem correctly in its fully dynamical sense, you find that the things that stabilize the system are not anything like near equilibrium fluctuations. They look almost more like solar flares. They’re these big loops of excursion through the interior dynamics of the system. There’s a new kind of a dynamical stabilization that you need to explain how a population has a non-arbitrary stable, long-term configuration of the always nice and the conditionally nice, who protect it.
Jim: Yeah, so we call that at least in evolutionary computation, the search for evolutionary stable strategies. In this case, it’s an ensemble approach, right, so that the society can be resistant to predators by not being too nice all the time.
Eric: Exactly and this is sort of the dynamically maintained composition that is stable, so it’s one step beyond Maynard Smith’s original idea of the evolutionary stable strategy as a quasi-static thing.
Jim: In terms of the emergence of monetary systems, to what degree are they endogenous, are they exogenous, are they some combination of both, are they frozen accidents, are they manifest destiny of teleology, where do our monetary systems come from?
Eric: Yeah, this is great and this is a dimension, where I think I can add something that Martin and I never discussed in our work together and yet, it would be the most natural topic in the world and I think he would agree with this point of view. Coming back to origin of life and to Herb Simon, once you arrive at something complicated, you should ask how you got there through [inaudible 00:50:29] of least resistance and things that were simpler. If you have something complicated like a governmentally mediated fiat money that has a heavy scaffolding of institutional supports and also social norms to make it possible. Martin was of course very interested in the role of commodity monies, barley, salt, gold. I think he has a book by that title Barley, Gold and Fiat with Tom Quint. He was very interested in the role of these things that have high use value in the short term and they’re only semi durable as the entries to money systems.
Eric: The thing that I think we could add to Martin’s body of work that’s almost more biological is how do commodity money support the emergence of complex institutions, which then make more complex money systems possible. Part of this was in Martin’s work and was integral since before I was born. Gold as a commodity money brings in the problem of cheating and dilution of the gold. That brought in the printing of coins and the idea of defacement. When the king’s faces on the coin and if you scrape gold off the face in order to try to steal some of it, the king can kill you for this, which is where defacement of coinage came from. Now, a functioning monetary system can increase the strength of the central state and its ability to enforce laws, but once the central state is stronger, then you can start to have base metals like aluminum, which are the foundation for coinage that don’t have intrinsic value and are supported entirely by the punishment against defacement. That bootstrapping between the easier forms of money, but the less flexible forms and the support of more complex institutions that give money systems more flexibility, there’s still more work to do in that beyond even the considerable work that Martin already did.
Jim: You may know I’ve done a fair amount of work on monetary systems and in my demonology, I focus a lot on the emergence of private commercial, fractional reserve banking with the emergence of the Bank of England in 1694.
Eric: Wonderful.
Jim: Any of your work that touches on that particular phenomenon?
Eric: Yes, absolutely. In the wake of the 2007-2008 banking meltdown, there sprung up a very good literature from insiders to outsiders about how to understand bank runs, leveraged instruments, fractional reserve lending and the instabilities in those systems. That took us all the way back to [inaudible 00:53:07] and the rise of fractional reserve banking in London and the fragility against bank runs that that brought with it and then, the emergence of lender insurance as the stabilization institution against bank runs. The only paper Martin and I ever wrote that was accepted for publication without a terrible fight was our paper on self-fulfilling prophecies and bank runs. That was because I could make a simple statistical model of how you can enter a banking contract for fractional reserves, believing that the payoff is better than you can get by staying out of that market. Yet, once you’re in there, you can be subject to the instability of a bank run if people wind up essentially escaping from the basin of attraction of trust.
Eric: Now, the interesting thing is I always wanted to follow that up with a model of the mortgage-backed securities and the collateralized loans, the repo markets that were the source of instability that set off the 2007-2008 problem.
Jim: Absolutely. Let me insert here just for the audience, I spent a lot of time researching this in the 2008 to 2012 range. I found one book that did the best job, called Slapped by The Invisible Hand yes by Gary Gorton at Yale. I believe his forensics is the correct one, just as a call out to folks that might want to do their own reading in this.
Eric: Gary Gorton and also Michael Woodford, those two were two we read heavily. In the Journal of Economic Literature, they have some excellent articles, some of which were collected in Gary’s book. Yeah, super sources on both of those. If you look at the way markets for repurchase agreements work, it’s clear that if the securities behind them are priced using mark-to-market methods that system is inherently unstable. It’s almost like a classical electrical engineering design of amplifiers. You could write it down and ask when the poles go outside the unit circle? I always wanted to do that and never got around to it because it would be a perfect application of Martin’s strategic market game formulation. After Gary and Michael had written, I didn’t feel like there were any new problems left for us to solve. It was never a priority to do.
Eric: The interesting thing you can ask is if everybody knows it’s unstable, why do they go into it or if they know that it’s unstable because mark-to-market intrinsically replaces the diversity that is the assumption for representative agents and stabilizing the economy with a sort of a mechanized, everybody steps on the same foot at the same time. Why would they do that? Well, they do it because it gives you legal defensibility. You can claim you’re using best practices when you get sued. This was one of Martin’s favorite things. When you look at something pathological in the economy, look for where it comes from the law as its origin. The interesting question about all of this is how does the legal system create homogenization in people’s actions by replacing their intrinsic diversity with algorithmic homogeneity and then, how do these institutions like mark-to-market interact with the adoption of repurchase agreements on very fast turnover markets to create a system that has these overnight instabilities of huge scale?
Jim: Yeah, not only the first-order effect, but I thought what Gorton was particularly brilliant at was he looked at the phenomena of rehypothecation of collateral in the repo market, which is a very peculiar thing, but turned I believe and I think Gorton does too was the actual spring that imploded and caused what was a big economic error, about a trillion dollars’ worth of bad mortgages into a near collapse of monetary system, which is where a bank [inaudible 00:56:58] as an example. They were one of the biggest players in the repo market and they get a pile of collateralized mortgage obligations in the form of a security as collateral for their repo. They are allowed under US law to take 40% of that and re-lend it to somebody else. That person can do 40% etc. In London, it’s a 100%. You can essentially have an infinite expansion of the near money supply through this overnight rehypothecation of collateral phenomenon.
Jim: When Lehman way down and panic ensued, everybody refused to take rehypothesized collateral and this gigantic six trillion dollar pseudo money, near money phenomenon imploded down to about 550 billions. More than 90% of it disappeared in two weeks and that was the shockwave that almost brought the house down.
Eric: Yeah, this is wonderful. It brings up two other people that I think we want to call out in this conversation. One is John [inaudible 00:57:57], who was interested in haircut revaluation, but your real point is that the magnitude of the bad debt was absorbable. The fact that people didn’t know where it was what pulled down the system. For this, there’s a wonderful book Hyman Minsky, Stabilizing an Unstable Economy. I can use your podcast to get a term out in circulation that I never got to use in a paper. I want to call that argument Minsky’s ratchet, which is that during stable times when you need regulation, you can’t muster the political will to get regulation in the necessary forms. The quality of credit degrades and degrades through innovations like rehypothecation. Then, when things go bad, there’s only one actor who can act on the necessary scale and on the necessary short time frame and that’s the central bank and the government. There’s no government that will choose not to act unless it chooses to collapse its own economic system. It almost doesn’t matter who’s in power. If you have anyone who’s not willing to implode the country that actor is going to bail the system out, everybody’s going to howl about it, but the horse is out of the gate.
Eric: The time when a needed regulation is passed, you can put the regulation on ex-post, but there’s too much entropy in the system. That money has been dispersed. There’s too much information gone and you can’t recover it, which is what we saw in the mortgage-backed securities case. There’s a meta question here, how do we overcome Minsky’s ratchet, how do we maintain political and legal vigilance, so that it’s active during the times when it’s needed, but people are comfortable and they’re not good at assessing the fragility in the system.
Jim: Ideas?
Eric: Well, this is what the systemic risk people ought to have good ideas in. I have not done original work in this area, apart from sort of thinking that this is the problem. No, I can’t offer much that I think has substance in this.
Jim: Too bad. That would have been fun. Now, let’s talk a little bit about money versus its substitutes, various near monies. Did you guys look into that, any?
Eric: Yes, a lot. One of Martin’s ideas was that money is only made effective when it is surrounded by suitable forms of credit that give the degrees of flexibility that the central money system can’t. He was always interested in things like the monetization of personal credit, concepts like the real bills doctrine that supported much of mercantile English traffic. What did we call it when England was the merchant marine of the world? Mercantile capitalism I guess.
Jim: Yep, sounds reasonable.
Eric: Yeah, so he was interested in how you can take things like promissory notes that are generated in a local distributed fashion and which are therefore sensitive to local conditions and then, what the bank wants is a credit evaluation feature that will allow it to pick and choose among these distributed instruments, monetizing them, stabilizing them by trading them for things like government money or gold. Thus, leveraging their flexibility, but then giving them the stability to function as money’s on a much larger scale. The reason Martin was interested in this is that it highlights the need for trusts and institutions as the foundation of money because when you have these near monies or monetized credits, you can have a many fold expansion of the volume of trade that’s supported. Exactly, as you say when rehypothecation came to not be trusted, you can have the collapse of all of these near monies down to sort of the core of trusted money, which can be a small fraction of what the trade and process had been assuming.
Jim: Very true, when it works, it works. When it doesn’t, it doesn’t. That seems to be the nature of monetary systems. For those who are interested in some of this, I have an hour-and-a-half video, called Dividend Money, might check it out on YouTube. This has been very interesting. Before we go to on money, of course, I have to ask you what are your thoughts about the new alternative currencies, the crypto currencies etc., can they possibly escape their little miniature basins of attraction and become significant?
Eric: Not in the trivializing way that their proponents hope. There’s a kind of a simplified view of the world that one sees a lot in engineering, which comes from knowing one system very well and imagining that that’s a surrogate for the world, instead of just one little component in a world that remains complex. The crypto currencies fill a role in distributed authentication that’s very hard to fill because the problem is always who will guard the guardians, what happens when every institution that provides trust is not ultimately trustworthy? A money is a complicated system. It requires low-friction. The cryptocurrencies have the feature of limited availability, scarcity, which secures the stability value of money, but they don’t have the institutional surroundings that make money predictable. Gold is incompressible and if you have a gold only economy, every shock in that economy, every speculation, every surprise propagates through the entire economy, so you really don’t want your retirement savings in Bitcoin if you’re relatively old today. We will be brought back to these classic questions of how one creates a stable society and a stable set of laws and monetary institutions and I think what we will one is to embed the crypto currencies within that larger conversation about how they can contribute to the stability of institutions and then, benefit from it instead of envisioning them as a replacement for it all.
Eric: If nothing else, look at the energy footprint of cryptocurrencies right now, which is completely irresponsible. You’re trying to be a speculative surrogate for the entire rest of the world, instead of using themselves as a window to better understand the world and know how to make it more robust. That’s the kind of naive mistake that we see being made with everything when it’s first introduced.
Jim: Yeah, I agree a 100%. I own zero Bitcoin and I’m happy to be there. My friend Ben [inaudible 01:04:08] likes to joke that Bitcoin is nothing but a way to accelerate the heat death of the universe, basically stupid computing, burning up some non-trivial amount of all the electricity on Earth. With that let’s exit money and finance, some really interesting insights. Let’s move on to our next topic, which is language, its origins and evolution. Talk a little bit about how you got interested in that idea and sort of the big picture what you’re thinking is there.
Eric: Yeah, so languages have always been a joy. They’re a thing that we’re all embedded in. We have an enormous amount of tacit knowledge, both as speakers of our own language and as participants in whatever languages surround us. Anybody, who’s a young kid and encounters some foreign languages for the first time will have experienced this wonder that things in a foreign language suddenly makes sense of things in your own language that you never understood before. If you grow up as a speaker of English, you should be perplexed and appalled that it’s even a pronounceable language because its rules are so Byzantine. Then, if you encounter Italian, which is a logical language, suddenly you realize that half of English is logical, but it’s the logic of Latin and then, the other half you can recognize as logical, but it’s the logic of German and then, English starts to become comprehensible.
Eric: When we went to SFI and I encountered a group that was seriously working on reconstructing the history of all of the world’s languages as a platform from which to understand language as a phenomenon, I thought, “Okay, this is a once-in-a-lifetime opportunity, can’t pass this up.” The thing that’s fascinating is that linguistics is an excessively conservative field. They resisted using modern probability methods much longer than they should have. They’re starting to do it now, but it’s four decades later than it should have been. At SFI, the linguists wanted to engage the biologists, who had been using probability methods for genomics for some decades. That was often a frustrating exchange. It was frustrating because of human difficulty. The biologists were too arrogant, thinking they had solved problems and too uninterested in learning the depth of what linguists knew about the structure of language that’s different from the structure of genome dynamics. The linguists were too stuck in the mud to try to use a polite term to be willing to put in the hard work of being students of probability methods and really taking them on in their modern form.
Eric: I remember witnessing these exchanges and thinking, “This is insane. If either side of this would exercise little humility and really become a student of the other side, we could open up this whole study.” That of course turned out to be a lot harder to do than to complain about and it took us seven years of work to get the manpower to actually do a very simple version of this problem. By the time we got it done and got it published, the rest of the world was already starting to do this at a good level. This system is still really gold on the ground for whoever wants to start to work in this area because we know so much about probability methods and we know so much about the structure of language, but to really work from whole cloth and make probabilistic models of the intrinsic dynamics of language use and language change that is yet to be done. Some young generation with time that wants to pick this up can really create a new field.
Jim: You mentioned language change, I know that was a big interest of the SFI group working around language. What insights did you get about how languages evolve and change?
Eric: I’m going to answer you as a physicist because I love the phenomenon in themselves, in their diversity, but I also like the patterns that we see that unify the phenomena. The interesting thing is that with genome change, we can get a long way by looking at changes that are not context dependent. You can look at the flips of individual bases and how they create new properties in organisms or how they don’t create new properties, so they can stay around for a long time as a signature of population splits. Language only functions because its components are a system. There’s no magic in the fact that I begin the word language with la sound, except that the sound la is used in all the other words in English, where la is used. That’s what gives it its meaning, its contrast against the other things that are different sounds.
Eric: If the sound la is going to change to something else, it’s going to have to do it in all of the places where it appears in English at the same time and it’s going to have to change for all the speakers of English at the same time. That makes the modeling of language change a very different problem than the modeling of genome change in many respects. The sort of the joint change of the system with the tokens that carry the properties of the system and that’s rich because context discovery remains a hard problem. It’s one of the interesting problems in deep learning, I would love to see the deep learning community, which has been so data centric start to take more of an interest in grammar discovery, syntax discovery and things like that to which they have been resistant so far because I think we will learn more about the nature of grammar when we start to ask deep learning systems to unpack what they know about language, like the [inaudible 01:09:36] translation, so that they can start to discover things about grammar and syntax that we have not yet observed.
Jim: Makes sense because it is reasonably close. Unfortunately, it’s not the way that most deep learning people are thinking, though there are beginning to be some people who are willing to think about what’s inside the black box and how to interpret the black box in something closer to symbolic terms because if there’s anything that’s symbolic, its language for sure, right?
Eric: Well, it’s a collaboration I would like to entice somebody in because we know that to have deep learners unpack their knowledge and present it to us in understandable form it is certainly a thing the deep learning community is interested in. Language would be a great place to work because we have the results on [inaudible 01:10:18], so we know we have some working black boxes. We have template for the kinds of things to look for because however incomplete it may be, linguist’s understanding of grammar and typology, syntax, morphology, all these components is very rich. It’s a good type of logical space. That’s a collaboration that we should be pursuing. I hope somebody is successful in sort of grabbing the interest of the people, who are very much in demand on both sides of this problem.
Jim: Very true. One person I know working on [inaudible 01:10:50], very interesting guy, who’s looking at doing, what he calls grammar induction. The theory is you take a big enough corpus of language in the wild and actually induce the grammar in a mechanical way. Now, he’s not all the way there yet, but he’s making some progress. He’s a person to keep an eye on in that space.
Eric: We shouldn’t reinvigorate that conversation and do a workshop at SFI on that question.
Jim: I’ll tell you what, I’ll introduce you to Lennis and see if you guys kind of resonate enough in your approaches to maybe you can be the [inaudible 01:11:20] that then turns into a workshop. I’m happy to do the organizational stuff at SFI.
Eric: Yeah, terrific, let’s put that together and keep [inaudible 01:11:26] in the loop too because I know he continues to want to do this.
Jim: That would be great. Let me ask you a little bit less scientific questions, but more kind of pop questions about language and its origins is do you have a view or do you have any reason to have a view about when something like our full language came into existence? I mean there’s some theories that say, “Hey, we were speaking something pretty close to [inaudible 01:11:47] languages 300,000 years ago.” Then, other people say, “Nah, it was only 40,000 years ago.” Some people even say, “Nah, it was only 10,000 years ago.” You have a dog in that fight?
Eric: No, I’m a tourist of those opinions, kind of the same way that you are. I read the arguments given for each of them, but I don’t have anything to add that would give me a unique take.
Jim: If you had to pick one, what would be your best guess?
Eric: I would go in the 100,000 ballpark in the sense that that seems to be when we get the modern form of Y chromosome and the already in place modern form of the mitochondrion. There may be later changes in human genome that we don’t easily see, but I don’t see what would have changed overall brain development much later than that. The question of what is the cause of cultural modernity, the apparent horizon at 40,000 years remains interesting and compelling.
Jim: I agree. Here’s my contribution to this discussion as I say it has to be more than 65,000 years ago and here’s why, the out of Africa migration that was the one from which almost all non-African people are descended happened about 65,000 years ago. There’s almost no genetic backflow to Africa for a very long time and yet [inaudible 01:13:08] language exists in both Africa and outside Africa. Therefore, the capability to have full language had to have existed 65,000 years ago.
Eric: Yeah that sounds right to me.
Jim: Yeah, I think that seems to be a pretty clean way to look at. Anything else you want to say about language before we move on to sustainability?
Eric: No, nothing too novo.
Jim: Okay that’s cool. Now, this is something that I don’t know to what degree you’ve made an academic study of it, but you and I have talked about it. In fact, I recall when you invited-
Eric: Joel Salatin.
Jim: … Joel Salatin, who as it turns out is my neighbor and we now know real well. Eric invited Joel Salatin to talk at an SFI event. Then, you and I talked quite a bit about that before and after. You clearly have a lot of thoughts about the fact one that we’re [inaudible 01:13:50] if we don’t change our ways and two, at least some thoughts on what that might mean. Why don’t you give us our thoughts on where we’re at with respect to ecosystem sustainability and what are some of the roads we need to think about going down if we’re going to actually get there?
Eric: Yeah, the way I entered this conversation was through a collaboration of agriculture and medicine, people of all kinds that they really are citizens in that some of them are researchers, some of them are practitioners, some are policy people. There are working farmers and doctors from all different contexts. The collaboration was formed around the idea that we know how to do better than we actually do in practice, what would a better forum for the world look like and where our knowledge is insufficient, where should our investigative efforts most go? These were participants like West Jackson, who founded the Land Institute to create the concept of perennial poly culture, but also people interested in soil microbial ecology as the sort of foundation of farming before you worry about what you put in the soil, diversified cropping systems, plant animal, mutualisms and how they affect the long-term viability of soils and terrains. Then, how this integrates back into the sort of large-scale geological processes of [inaudible 01:15:13] weathering, things that Sir Albert Howard was worried about and what he called the law of return.
Eric: What are the natural roles of forests as miners of trace metals and then, grasslands as the inheritors of what forest mine with their deep roots and how these biota on the whole contribute to the sort of mountain to coast process of continental weathering, what’s the big picture? I’ve inherited a lot of that conversation, the luminaries in it were people like West Jackson, Wendell [inaudible 01:15:41]. The question I asked Joel when he came to SFI was how do our solutions on the small scale scale to a world that is now big and interconnected? There are people like Wendell who don’t like that. They don’t like the idea of anything on a big scale because their argument, which has a lot of weight is that the problems, we’re stuck on our complex and the only way you solve complex problems is by having a very deep local understanding. That certainly is true and I think it’s profound, but at the same time, a watershed or a continent or a large society are interconnected entities. We can’t escape the fact that we influence each other. We have to understand on the small scale, but we have to coordinate on the large scale.
Eric: A place where I don’t think anybody has good solutions is how do we leverage that small scale complex understanding, how do we feedback to really support it, how do we keep local farming on the land at a sophisticated level, which it’s not right now by having it supports for it come from the big system? Then, how do we keep the big system constrained by feeding from and using all of that locally acquired knowledge? That’s the big system problem. There are lots of little movements that are trying to do this. Great, support them in any way we can. I don’t see them impinging on the big structures that have lock-in on the industrial farming system and on the terrible waste and inefficiency that the demographics of our society has overall.
Eric: That structural change has to happen or we’re really in trouble.
Jim: Yeah, the way I sometimes phrase it is probably the solutions are local, but the system has to close globally. Either we run into the wall at 120 miles an hour and cook the planet and destroy our soils or we don’t. Unfortunately, we have a real bad collective action problem here. We have to have behaviors locally that are correct, but yet they have to add up and closed algebraically arithmetically even at a global level, a real conundrum.
Eric: This is right, this goes back to where we open the conversation about Herb Simon and systems thinking and Crick in the central dogma. In order to have hierarchical structure, you need control systems and that means that the controller has to be able to make the controlled do what it would not do on its own. Otherwise, you don’t need a controller, but that opens the problem that the controller is not immediately subject to feedbacks. It can be wrong and it can give wrong instructions maybe for a long time. How do we leverage the benefits of having control systems, but keep them responsive enough that when they’re giving the wrong instructions, they can’t continue to do it long enough to crash the system? We see that in every level. We see it in the nature of business and the theory of the firm. We see it in the economy. We see it in the problems of sustainability of social order. We see it in the question of what is democracy and what is required, what is the aspiration of democracy, what form would fulfill it and what do we need to maintain it, which is sort of a problem of major urgency around the world right now.
Eric: I think there’s a meta question at the system’s level, what is the common thing that we are failing to understand in all of these questions and how can we do better in bringing ideas to that problem?
Jim: Yeah, it strikes me that it’s how do we create hierarchical or at least emergent entities at multiple levels, it may not be formally hierarchical that nonetheless cohere together in real time. We have an example of that in the human body. We have billions of autonomous agents, called our cell that we’re at least originally free-floating single cells way back yonder and evolutionary time. Yet, we have this astounding homeostasis that binds them all together in exchange of gases, both bad gas out, good gas in, nutrients and toxins in this astounding modulated quasi hierarchical system. Seems to me that our world system needs to learn from that and be able to hold itself together coherently in approximately real time despite operating on multi scale simultaneously.
Eric: Yeah, transition from single cells to multi cells, who it went through and what innovations were needed is a topic that a lot of people are looking at these days. It’s a wonderful area. It’ll be interesting to see what we learn about the nature of regulation from trying to understand that better.
Jim: Yeah, unfortunately, we don’t have a lot of time. We don’t have time for probably for the answer to come from that kind of basic science, seems to me that if we don’t reform our ways, advanced modern society is going to run into the wall, probably no later than 2100. We’ve got to essentially innovate at higher levels of abstraction I’m afraid rather than wait for answers from lower-level research.
Eric: Yeah that seems right to me too.
Jim: Have you given any thought at all to what kind of signaling modalities, what kind of control systems, what kind of governance might be appropriate in this new world, let’s call it we got to get it in place by 2050 kind of timeframe?
Eric: There’s a book I think by Paul Hawken called Blessed Unrest and he has a data base behind it or did a few years ago, which is an attempt to collect all of the civil society organizations that he could learn about and put them in a place, where they can find each other. We need a better theory of governance. We can see governance being hollowed out by the sort of takeover of concentrated interests that are coming from private organizations at this time. Government has a separate representation and organization in the society, away from business or wealthy individuals or land. That is being weakened and that needs to be brought back, but presumably civil society is the other pillar, which is the smaller and more disaggregated cousin to government. I feel like the right conversation is going back to the conversation from Adam Smith’s time, but in the institutional world of today, rather than the one that he lived in. We have firms, we have markets, we have wealth concentrating mechanisms and we have private ownership. We for good or for ill are going to have massive information gathering and control mechanisms. We have civil society and we have government. What is the most stable ecosystem of those components that we can foresee and try to design? That’s the way I would put the sort of big picture question.
Jim: Yeah, interestingly, I’m putting a lot of my efforts right now on what I call the meso scale. It used to be that a lot of what made our lives both doable and pleasant was for me, either the family or the local face-to-face community, but over the last 200 years, most of that has been ceded to either the governments or the markets. It seems to me we need to find a way to revitalize a meso scale on the order of hundreds and a few thousand people. Some people call that civil society, but it may actually have more structure to it than unstructured civil society. Any thoughts on whether that’s a useful scale to be prospecting on?
Eric: My thoughts are that I agree with your sensibility, but I also think this is where you ought to have a good China scholar in place because if I understand correctly, some of the things that are guiding Chinese governance right now are kind of a distrust in the controllability of that meso scale, which is why they’re trying to weaken that and replace it with centralized control. It would be really interesting to have somebody who’s informed and sophisticated in where that line of thought comes from on this conversation and say, “Why do they distrust it, to what extent do they mistrust it, why do they think they can live without it if that’s what they think, what do they think can replace it and sort of what’s the merit of arguments from the evidence we have about what problems can be solved by institutions at each of these scales?” I would be really interested to listen to informed people have that conversation.
Jim: Do you have any pointers to people who were thinking in this space?
Eric: Well, there was a frontline documentary on AI in the context of Chinese governance today. We could look to people on there and we could start in that as a network of exploration with some of the people who appeared in that show, either they themselves or people they would recommend. That’s where I would start because I’m kind of new to this.
Jim: All right. That’s great. I’m going to go mind map. I’ll let you know what I find out. Well, we’re getting close to the end of our time here. This has been as I knew it would, an unbelievably deep and interesting set of conversations about a wider range of topics and one could ever imagine any one person having in their head. Before we go, could you tell us what you’re working on now and what you might be working on soon?
Eric: Yeah, my work continues to be sort of at the interface of organized geochemistry and the deep origins of biology. I’m getting more into the domain of folding and the origin of biopolymers than I used to because I have collaborators at Georgia Tech, who are some of the world experts in that problem. We are looking again at the genetic code and that’s where some my earlier comments on that came from. I should say that there’s a meta question that to me is the interesting question that wraps many of these others together, we often think about biology in terms of its extraordinary capacity to carry information. That capacity only exists because biology creates big combinatorial state spaces that in physics terms, we would say are relatively flat. There can be big genomes, where any sequence is really available, the substrate of the material gene doesn’t fight you from making one sequence versus another. That means that relatively weak forces, like selection at the individual level can cause you to migrate around in that genome space.
Eric: There is a similar story to be told for protein catalysts, where you can kind of make any sequence and you can fold lots of them and you can catalyze with lots of them. I think if you’re a physicist and you know how heterogeneous chemistry is, you would say that these large combinatorial spaces that are physically neutral are not typical in a chemical system. The typical thing in the chemical system is for configurational heterogeneity to be so big that it just wipes out any kind of combinatorial flatness. If we want to understand why biospheres are possible, we need to understand where there are sort of linchpin steps. They’re not combinatorial in themselves. They’re often small chemical problems but they create narrow bottlenecks. Those bottlenecks reject enough of the configuration entropy of the problem that the stuff that gets through the bottleneck then becomes a combinatorially large and flat space that’s capable of carrying information. That was probably a little bit technically higher level than you want for your podcast, but the sort of soundbite compression of it is that we want to understand when and how systems that are capable of carrying a lot of information become available without being buried in the heterogeneity of the physical material they’re made of.
Eric: Now, I think that’s a question where physics and biology can productively interface.
Jim: Wow, All right that is wonderful. I want to thank you very much for being on here. I think our audience will have learned a heck of a lot if they make it here to the end.
Eric: So pleasure to be here. It’s great to be in contact with you ongoing.
Jim: It really is. Stay in touch.
Production services and audio editing by Jared Janes Consulting. Music by Tom Muller at modernspacemusic.com.