James Pethokoukis in his own Substack: One of my favorite big-think papers of recent years tries to explain humanity’s societal evolution, “The computational power of a human society: a new model of social evolution” by David H. Wolpert (Santa Fe Institute,) and Kyle Harper (University of Oklahoma, Santa Fe Institute). The authors suggest that we should think of societies as machines that process information and harvest energy. They argue that as societies have gotten better at both gathering energy and processing information, they’ve become larger and more complex. As I wrote back in August in “A thermodynamic miracle: Compute and energy are key to humanity’s continued evolution”:
The formula: Better energy harvesting (allowing societies to do more work and support more people) + improved information processing (allowing societies to use that energy more effectively and to coordinate more complex social structures) = greater complexity (which manifests in various ways such as larger populations, more diverse occupations, more advanced technologies, and more intricate social organizations).
Wolpert’s research focuses on information theory, computer science theory, and mathematical logic, though his interdisciplinary approach has carried his work into a variety of diverse fields. In addition to his work at the Santa Fe Institute, Wolpert is an adjunct professor at the Arizona State University Center for Biosocial Complex Systems and a visiting professor in the MIT Astronautics and Aeronautics Department.
1/ How does thinking of societies like computers help us understand how and why societies grow in scale and complexity over time?
So the easiest answer to that, which actually was the inspiration for Kyle Harper, my co-author, for his reaching out to me — because he’s a classicist by training and my reaction was, “What have your research interests got to do with mine?” . . . The reason he reached out to me was because, a couple of things: One, there’s a paper that I published in Nature Communications with Tim Kohler. Second, there’s a great book . . . called “Social Development” by Ian Morris, also a classicist at Stanford . . .
Ian found . . . that if you look at all civilizations across the past 5,000 years, you look at a real dumb measure of their “complexity,” energy usage per capita — all kinds of reasons why that’s a stupid thing to look at. It’s absurdly crude. But let’s go with it, see where using it to look at civilizations gets us.
Energy usage per capita is the total amount of energy harvested by the civilization divided by the number of people in that civilization. And Ian identified “civilizations” with polities, so there’s all kinds of squirrely issues about how you define a polity, especially when you go back to 4,000 years ago and so on; but ignore all these concerns.
Then, suppose you look at that curve of energy usage per capita over, for example, for China, or Rome, and you also look at various proxies for the computational capabilities of the society: things like, how sophisticated is the writing system, how sophisticated is the monetary system, how sophisticated is the transport network, how sophisticated would be things like the libraries? All of these are proxies for the “computational capabilities” of the society. Even so, you can put those proxies all together and spin it up as the “information-processing capabilities” of that society.
What Ian found, loosely speaking, is that energy usage per capita and computational capabilities go in lockstep. It’s not that one enables the other, it’s that they go together. So the more computational power that society has, the more energy extraction per capita, and vice versa. So certainly, if you are interested in the energy usage per capita, which captures so many attributes of a society, then viewing it as a computer and quantifying its computational chops, so to speak, is going to be a very, very productive thing to do. So that’s Ian’s book — very, very easy to read, by the way, no matter how much training you have.
In this Nature Communications paper by me, and Tim Kohler, Michael Price, a bunch of other people, we actually looked at a different dataset; it’s called Seshat. Peter Turchin . . . put together a dataset . . . And this was 35 different polities across all six inhabited continents for the past 10,000 years. He got like 50 co-authors on this paper, it’s an amazing thing, archeologists who had expertise in all of these different polities, and he put together this great dataset and did some stunning analysis of it.
The paper that Tim Kohler and I did followed up on that . . . through this lens of Ian Morris’s insights, and what we found was that all human societies go through a process where first you grow in size without much change to your computational capabilities, measured in these proxy ways, until you get to a certain sufficient size, measured by population, population of the capital, and so on, at which point, suddenly bang, your computational capabilities go off and you don’t grow much until they reach a threshold. And going past that, all of a sudden, no more increase of my computational capabilities. But now I start increasing in size again until another threshold, and we go through a sawtooth.
So, putting this all together, what Kyle perceived was that there’s a lot here to be learned about all these things that we’re concerned about in human societies: their size, energy usage per capita, and so on and so forth, by viewing those societies as computational systems. He saw that the computational power of that society gives you a lot of insight into all these other attributes which we are so interested in.
Over and above the scientific interest of viewing a society this way, as a computer, doing that provides insight into much more brass tacks, economic attributes . . . We are making no claim that such insights are going to allow us predict what’s going to be happening in human societies over the next decade or so. But they can tells about how societies evolve over much longer timescales.
. . . what we found was that all human societies go through a process where first you grow in size without much change to your computational capabilities, measured in these proxy ways, until you get to a certain sufficient size, measured by population, population of the capital, and so on, at which point, suddenly bang, your computational capabilities go off and you don’t grow much until they reach a threshold. And going past that, all of a sudden, no more increase of my computational capabilities. But now I start increasing in size again until another threshold, and we go through a sawtooth.
2/ How does the role of job specialization help us understand this complexity issue?
It’s great to look at things where the human society is a computer, in the sense that all that we’re quantifying is these proxies: how sophisticated is your writing system, and so on and so forth. That’s not going to tell you anything, frankly, past the point that you’ve got a phonetic alphabet, and that you can write well, and you’ve got a decent postal system. To go beyond that, if we are going to actually explore what’s going on, why are the computational capabilities so important at a deeper level, we need to quantify: What do we mean? Not just what are these proxies for, but what is a human society as a computer.
There’s this huge field of computer science theory, and it’s got all these different models of what computers are. To be able to make any progress of understanding the deeper scientific issues here, we need to actually say, “What is that human society as a computer? What computer is it?” So one of the very, very simplest ways to go after this — and it’s only one of a bunch we’re considering — is to say that a human society is a bunch of occupations and technologies that are communicating with one another. Those occupations implement like production functions in the economic sense. The technologies — this goes back to Schumpeter — the innovations, you get new technologies, new capabilities that way. And so that is one of the lenses with which we are trying to actually get more specific about what we mean when we say a human society is a computer, getting beyond these crude proxies…
More here.
