DOES TECHNOLOGY EVOLVE?
[BRIAN ARTHUR:] In my career I have looked at very disparate subjects or areas of interest. I got a chair at Stanford very early on by being an expert in human fertility in the Third World, believe it or not, as a demographer. I've been interested in the economy; I've been interested in certain types of mathematics; I've been interested lately in technology. I'm getting to an age now where I can start to look back and think, what on earth was all that about? What was the common thread? I realize there is a common thread and it's very deep inside me.
One of the things I was interested in also was what became known later as complexity and all the stuff that the Santa Fe Institute became known for. It turned out I got to Stanford at the right time in the mid-'80s when a lot of that was unfolding. So I thought: demography, complexity, economics, technology. I began to think in a rather despondent way that maybe I was a dilettante. I dabbled here and dabbled there and there wasn't really a common theme. But in the last year or two I have realized there was a common theme. This is really what motivates me. I'm interested in systems, the unfolding or the development of systems or of patterns.
When I studied at Berkeley — this was early '70s/late '60s — I wasn't interested in what became known as equilibrium economics where everything is static and unmoving; I specialized in the economics of development. Then, when I got interested in complexity, I realized it was about systems, that complexity is basically about systems that are reacting to the overall situation they create, individual elements reacting to their overall pattern. As a result the composite unfolds and develops. My fascination throughout my entire career has been with the unfolding, emergence of patterns, or what you might call evolution in a rather narrow sense. Lately this has become in turn an obsession with the evolution of technology.
For me all this happened in the early '80s. Something was in the air and I think for me this was just pure instinct and I would say it goes an awful lot deeper. This is a matter of temperament or personality. Some people like to freeze what they are looking at. It's as if you can fast freeze. You might be interested in how butterflies fly and, so you catch them and chloroform them, nail them to a board and stare at their wings. But I was always interested in what makes the dynamics actually work.
My Ph.D. was in operations research, essentially applied mathematics, in control theory. But how do you control systems that are unfolding over time? Most of my training has been in either engineering or mathematics, but I became fascinated with the economy. I don't know what was in the air. Sometime around 1980, we all got computers. By the mid-'80s these were what you would call workstations. I had a Next Computer as soon as those things came out. Stuart Kauffman had a Sun Sparc Station. I remember a Sparc 2. But we all had workstations. You could simulate actual systems unfolding, so you could watch things unfold on your computer. You could write programs for each of these elements. They might be B cells or T cells in an immune system. They could be species in Stuart's world or agents in the economy, investors and so on.
We began to extract parts of the world that we were interested in, then recreated them on our computers and then hit the return button for them to evolve. We watched what happened. We tried to get a mathematical description, because it wasn't sufficient to just say, well, here is how it unfolds on the computer. Here are nice pictures and we can freeze it here and we can look at these simulations or evolutions happening. It was better to see if we could get a mathematical version. Usually it was stochastic, probabilistic, and could go in different directions.
But there is some deeper part of my personality that became completely fascinated with the idea that the world is always unfolding. If somebody said to me, "Here is a picture of the universe as it is. Isn't it amazing? Isn't it incredible? Look at all the incredible structures that are there and they are frozen in time in this picture," I would say, "Yeah, it's interesting." But if somebody showed me how something was actually changing and unfolding, how new structures arise and fall away and further ones arise, I found that fascinating. I never knew how deep that went in my personality.
To give some personal history here where all this fits in: I was born in Northern Ireland, which might seem like a kind of fractured culture, and it is, but it was also a very stable culture. People knew who they were. The generation I'm from in Northern Ireland was known for writing and poetry. All the top poets were from the north of Ireland, not the Republic. I grew up in a very stable atmosphere but it was like a pressure cooker. It was fundamentalist Catholic. On the other side it was the same milieu, we had fundamentalist Protestantism. Something inside me just couldn't stand that. It was as if there was a lid on. When I was 12 or 15 or 20, before I left, I felt I could just simply scream. I felt that everything about the religious tensions in North Ireland hadn't changed in hundreds of years. Something was screaming at me. To cut a long story short, if that is a highly, highly ordered system in complexity terms, then there is the chaotic system.
I went to university in Belfast and got a superb education in engineering and then went to Berkeley where I was essentially in Applied Mathematics and then Economics. But it was as if I had gone from an ordered state into pure chaos in Berkeley. I couldn't handle that either. That was far too disordered for me. The element I want to stress, and this is where I talk about people's personality, is that until you get older, into your 30s and 40s and mature a bit — until I had a family and was teaching at Stanford — you really don't know what your temperament is all about. I knew I liked certain types of science and I knew there were certain areas that appealed to me. But, as I said, a lot of this had to do with how populations grew and unfolded, human population, how economies developed and more lately how technology develops and evolves.
I must have gone through a mid-life crisis sometime in my early 40s or late 30s. I got an endowed chair at Stanford when I was 37. I had striven all my life to get to the next level and then the next level and then the next level and suddenly there weren't any more levels. I didn't quite know where to find my feet. I couldn't find anything I wanted to grab onto in Berkeley. The counterculture didn't appeal to me. This was still not long after the Vietnam War. I had worked in Austria. But there wasn't anything that I felt really formed a basis. I got rid of being Catholic, so what could I put my feet on?
I discovered something that was very much my temperament which surprised me. I got quite deeply involved in Taoism. The amusing thing about all of this is when I first started to get deeply into Taoism, it was at first a philosophy, and now it has mostly been more training, some of it martial arts, the Qi Gong, very much in the context of some of those Chinese movies you see: you wait at the door for three years, maybe I let you in, maybe teach you something, maybe not — this kind of thing. But when I got deeply into the philosophy, I began to realize it was all about things unfolding and things becoming. The deepest part of Taoism philosophy is saying there isn't really anything stable that you can grab onto. The whole world is always changing. The best thing you can do with your life is to go with whatever is happening, allow yourself to flow with it. I began to realize that all my research had the same theme. You could say that I found in a deep philosophy, a several-thousand-year-old philosophy, the counterpart of what interested me in science. Or maybe that's the way I thought all along, that this was deeply inside of me, this philosophy. Then I discovered its counterpart in science.
So I have been interested all my life in the development and unfolding of systems, not so much systems that are fixed in structure that behave mechanically like a car, which are moving through time, but not changing in structure. The way I was trained in economics, the structure of an economy, industries that are already out there, and the various parties involved — banking systems, regulatory systems, government, consumers — all of those structures are first taken for granted, then economists look at how that unfolds. That wasn't enough for me. I'm interested in how structures themselves change. In fact, economics didn't even say that much. Economics was essentially saying given the fixed structure, where is its natural place to arrive at equilibrium? For me that was anathema from my point of view.
The economics of increasing returns
Sometime in the late 1970's I started to read about the dynamics of enzyme chemistry, kind of an obscure thing. Two books I read in 1979 had a huge effect on me. One was the Eighth Day of Creation by Horace Freeland Judson — one of the best science books I ever read in my life. The other one was Richard Rhodes' The Making of the Atomic Bomb. What Judson started to show me was that my view of biology was ignorant. I thought it was just about classification. Like stamp collecting. Here is this species and that species and here's how this works and interacts with that. Judson was describing how molecular biology really worked and how all of that had been founded.
I had previously read James Watson's book on molecular biology. This brought the whole subject alive and made it all dynamic. It was all about interpreting the genetic code and the structure of hemoglobin. It was real science and I got tremendously excited by it. They turned me on to biology. Then I read Jacques Monod's book, Chance and Necessity, and began to realize that what he was describing was chemical reactions that were poised to go one way or another. These were autocatalytic reactions. There might be two end products possible in one of those reactions and if the end product was A and A catalyzed more A or B catalyzed more B, then depending on which got ahead first, the whole thing could fall into more and more A, catalyzing A, and B would be shut out. Or more and more B, shutting A out.
I also read some essays by Ilya Prigogine at the time. He was controversial because there was a lot of self-promotion. He was kind to me. In 1980 I made a pilgrimage to see Prigogine and he took me out to the Royal Academy in Brussels: "As I was saying to the Pope on Thursday" and so on. Anyway, he was very kind to me. I find some people can have good ideas and very deep ideas but they maybe don't dot the "I"s and cross the "T"s quite as well as left gifted people. I began to realize that what Prigogine and his group and Monod and Jacob and others were talking about in France and Belgium were self-reinforcing systems but not ones that would run away like a snowball down a hill but were more like bandwagons, any of which could gain momentum if they got off to a good start.
When you see a pattern like that, you begin to see it everywhere. The English language was like that. In the 1700s if you wanted to make a bet, you would say, well, if we are sufficiently connected as a world, maybe a world language will appear. That had happened in the 1500's and 1600's. The language was Latin. But if you wanted to make your bet in the 1700's you would have bet on French. The Russian court was speaking French, and French was fashionable in high circles. Or you might have said 50 years later, by the 1850's, that it might be German because of the Austrian-Hungarian Empire and a fantastic culture of intellectual sweep right across Europe except in France but east of that. Then English emerged in the 20th century. It has really taken over in this new century. It's the same thing. There is a lot of autocatalytic feedback and I'm sure the rise of America did an a lot to bring English in.
I began to realize that there was a huge set of questions that I had asked myself in graduate school while I was studying economics, not for a Ph.D. but more as a hobby. It was my doctrinal minor and I had been taught very, very well by theorists in Berkeley — equilibrium theorists mostly. I had asked this question: Economics basically said that everything arrives at an equilibrium providing there are diminishing returns on the margin. So if something gets harder and harder to do, like you mine deeper and deeper for copper and it becomes more expensive to find copper because all the easy seams have been mined out, then you might start to substitute nickel for copper and the economy reaches an equilibrium. If your time is competing for television and movies, once you run into diminishing returns in movies — you've seen all the good ones that are on — you tend to say, okay, I will balance that out with whatever, reading books or television or something. So providing we run into diminishing returns everywhere, the economy is assured of reaching equilibrium.
I had asked myself a question as a graduate student, what if there weren't diminishing returns? What if there were increasing returns? If something gets better the more you make it, what would happen? I thought of an example, pretty trivial. I had been to Hawaii as a graduate student. Milton Freedman was fashionable at the time, and I thought to myself, "Suppose cars were offloaded for the first time from a ship in 1925, or some time like that, in Kauai, the most remote island (I was in Kauai in the '70's when they installed their first traffic light). I imagined if there were a few dirt roads and Kauai got cars for the first time, and let's say the steering wheel in this thought experiment was in the middle of the car and there were no bias to the right or left. The car had just arrived and Milton Freedman was running the island, so everybody was free to choose which side of the island they would drive on. It was a perfect Libertarian island with no laws whatever. Cars all arrive on this magical island, you are free to decide which side of the road to drive on. I thought, what would happen if I came back six months later?
I remember talking to Stuart Kauffman about this and he and I agreed that what you would observe would be an awful lot of wreckage at the side of the road, but that cars would line up on one side or the other by a convention that eventually would emerge. Once more people started to drive, say, on the left, you would simply be a fool if you didn't do the same, because most of the cars you would meet if you drove on the right would be going the other way and you wouldn't survive that long. A system like that could tip one way or the other.
The examples I referred to were the QWERTY keyboard design, and other structures in the economy, that locked in more or less by chance. Economists were aware that increasing returns or positive feedback — positive feedback in this case being the more cars are on one side of the road, the more you are inclined to do the same. So it's an autocatalytic system. The more one candidate in a primary gets ahead, the more advantages they have, and so on. Systems like that tended to lock in to the dominance of one player or one side of the road or one typewriter keyboard, and everything else got pushed aside. The lock-in happened probably by chance, by small random events getting magnified by these positive feedbacks.
At the time, economics was aware of the problem. The great economist Alfred Marshall in 1890 actually had written about it in a footnote, but he had asked: what if there were N firms all with diminishing costs? They're all in the same industry. Aneroid barometers fascinated him — the more barometers a company got out, the cheaper it was for them to produce the next barometer. He posited that in a case of increasing returns or diminishing costs, after some time the market would have become dominated by one firm, which in his words would be whichever firm first got off to a good start. Economics couldn't say anything beyond that.
What I contributed was to show that, actually, we can say something beyond that. I realized that you could treat such dynamics as probabilistic, or technically as non-linear stochastic processes and look at how these by chance ended up with one outcome, or by chance with another. I started to work with Russian probability theorists. I didn't know enough sophisticated probability theory in the mid-'80's to solve these questions. But one colleague of mine from the famous Kolmogorov school in the then-Soviet Union knew about this sort of probabilistic mathematics. I had to get up to speed myself. It took me a year or two. We cracked these problems. The first paper I published was in a Russian journal in 1983 on increasing returns. All I could do was read my own name. I couldn't read anything else.
I had been told that all this stuff on increasing returns was “theoretical.” Then I was talking about it in Santa Fe to some students and was walking to give my lecture and I had a complete epiphany. I thought, this isn't esoteric stuff, angels on pins. This applies to all of hi-tech. I began to realize that all of hi-tech operated according to increasing returns, which meant actually that the more a firm like Microsoft got ahead of the market, the more its brand would be out there, the more money it would have to parlay into the next thing. The more Google gets prominent as a search engine and the more people get used to using Google, the more omnipresent Google becomes. Other search engines get pushed aside, like Ask.com or Alta Vista. But you can't predict in advance which one it might be.
Suddenly I realized that there was a dynamic with hi-tech or Silicon Valley, and with hi-tech all over the US, that markets within hi-tech tended to tilt or tip into the dominance of one player. Hi-tech has what came to be known as "network effects": the more people that use Google, the more likely they are to use Google. This has huge up-front effects. Take Microsoft. They used to give you Windows on one of these little disk things, but the first copy of Windows — NT or, Windows 2000, or whatever it would be these days — might cost Microsoft something like $2 billion. The next copy might cost them fractions of a cent. So their costs would very rapidly go down per unit the more they get out there. Hi-tech has this. But this is not true for dog food. The next unit of dog food costs per unit just about as much as the first one.
Ideas operate very much according to increasing returns. It's costly for someone to dream them up and almost free for anyone to distribute them. It turns out that information is virtually free and we are seeing, as the economy runs more and more on ideas rather than on bulk commodities such as processed corn, processed iron ore, steel, or cars, we have different rules. Increasing rather than diminishing returns. A different economics applies.
We are coming to a commerce of ideas, i.e., an interchange of ideas that are simply out there. People use them as a kind of coinage. We will exchange ideas. We can use other people's ideas to construct yet further ideas. There is a natural tendency for information to be free. I don't know if it wants to be anything, but certainly once it's out there it's fairly costless to swap around. There are some superficialities and we are going to learn that things are much more complicated than that. Information is free, but grokking information, being able to take it in and use it for yourself, is not free. It's not sufficient to tell somebody something. That's just an idea. Ideas have to come with backing and they have to come with understanding. It's a bit like if I showed you a Chinese painting. You, or somebody who is skilled, could reproduce it, but what they couldn't reproduce is the depth of understanding and the culture, the context, the history, everything that led up to it.
A lot of complexity had to do with how systems unfold and by chance could go one way or the other. I was asking the same questions in economics, given positive as well as negative feedbacks: How do certain patterns form in the economy? How could we end up driving on the right rather than on the left? How come we speak English? But more, how does Silicon Valley end up in a bunch of apricot orchards rather than somewhere over in the East Bay? How do we get the dominance of something like Microsoft? Those were questions of unfolding but they weren't trivial questions such as saying there is another stage and another stage. These were systems that according to chance could fall into certain patterns and then the next pattern could fall on top of that.
There is another element to hi-tech. Microsoft may have been the thing of the moment in the early '90's, locking in most of the tech market with huge increasing returns. But just when you think it is going to be forever, something else comes along, Google. What fascinated me was that the economy had kind of layer after layer after layer, like an archeological site, Troy, with 14 layers of cities. But you could watch them in your own lifetime, five or six of these layers forming on top of each other.
I became fascinated by another question. We take an awful lot of things in the world for granted. We take it for granted that as technologies progress, they become more complicated. Something told me in 1993 to read up about jet engines. I thought, I'm not interested in jet engines, but it was an instinct. Of course, within two days I began to realize that the original jet engine was quite simple and the ones we have now are tens of thousands of parts and really, really complicated. You could say, oh, that's just the way technology is, but I began to wonder why it was that way.
By the mid-'90's, a series of questions emerged that I really started to obsess about. What really is technology? I had no clue. Somehow I began to realize that the economy forms out of its technologies in some way. The difference between a hi-tech economy, which we have here, and say the one in the Trobriand Islands, is the difference between having sophisticated technology versus fishing and canoes. I wondered how technologies develop over time — like how the jet engine became more sophisticated and why that should be. It's like saying why is the sky blue? We just take it for granted. Then above all I began to wonder, was there a theory of evolution for technology?
I was very aware as a graduate student that the economy somehow arose from its technologies. That wasn't new. Karl Marx had written a lot about that and other really good economists in the mid-1800s. But we tend to see an economy as out there and fixed and it uses factories and factories use technologies and machines and things like that. I began to wonder, how does something we call the economy arise out of technologies? I began to realize that every technology I knew about, from the computer to jet engines, started off pretty simple and wound up incredibly complicated, orders of magnitude more complicated than what started off. Then there was a question that had been hanging in the air since the 1860's, posed by archaeologists and anthropologists and a guy called Augustus Pitt Rivers, and by Samuel Butler: does technology as a whole evolve?
Darwin had answered that question in biology. The many species, in what we now call the biosphere, are somehow all related by threads of common ancestry that go way back. That wasn't new to Darwin. His grandfather had said as much, and other people had said as much in the 1700's and early 1800's. What Darwin was able to supply was the mechanism by which that branching and then speciation had occurred. His book was about the origin of species. It wasn't called evolution.
If you take certain technologies — you can take jet engines or you can take refining certain types of metals; the steam engine historically; lasers; computers, or methods for doing things, like sorting algorithms, which are methods. If you take all of those technologies together, can there be a theory of evolution by which they are somehow related by threads of common ancestry to earlier technologies?
Economists, historians and technology thinkers for decades, at least since Darwin, have been trying to apply Darwin’s theory to technology. Say, okay, different designers have different ways to solve problems. That gives you variation. Then the better solutions are selected. So you have variation and selection in technology. That would give you a Theory of Evolution for technology. I'm condensing 150 years worth of thinking since Darwin's book.
But those theories weren't satisfactory. At least they didn't satisfy me. The jet engine didn't evolve out of variations of the air piston engine. In Darwin's scheme, if you get a new species of finch, it's related to some older species of finch but adapted gradually and changed the structure of its beak because of some circumstance. There is a gradual progression until a new species is formed in a slightly different niche and branches off from the old species. You couldn't say that the laser or the jet engine branched off from anything. They were completely new. So I began to realize that there wasn't a satisfactory theory of evolution for technology. In fact, there wasn't a satisfactory theory of anything I could think of in technology.
There is no theory of technology. People greeted the very idea that there could be a theory of technology either with, "I guess you could do that but why would you want to do that?" Or, "I don't know, do we need anything like that? Do we need a theory of technology?" I didn't start to do this academically. It was kind of an obsession. I began to think, yeah, there are some common principles that I could use to think about all of this. So I went underground. I went back to my lab notebooks a few weeks ago when I finished the project, because I was curious to see how long I had been working on this. When I looked at my lab books, I realized I had been working on this project for 13 years. But most of the work was certainly not writing — it was reading and thinking about technology and the history of how technologies came into being — very specific ones — maybe a dozen to twenty technologies, everything from computers to search algorithms. Certain technologies I learned in great detail, just as a biologist would have to study certain types of beetles or something to make sense of their arguments.
What happened was that I went underground in this whole project for about thirteen years. I once read that the mathematician Andrew Wiles, who solved Fermat's Theorem did exactly that. Wiles, I think out of instinct, as a very good Princeton mathematician, decided that he didn't want his embryonic thoughts to be hammered by criticism. He needed space and time to think out his ideas, that he needed to put things together where he wasn't being bothered and questioned all the time. He didn't want to give any competition any ideas that this was getting worked on, so there would be a frenzy of activity somewhere else. Instinctively for 10 or more years, as far as I know, he worked in the attic of his house in Princeton and worked with one or two colleagues who were in on this and decided to hatch the whole plot and bit by bit built up his understanding until he could construct a working version of his theorem. Then he would give it out to a wider set of people to probe and test and so on. I made no such resolution when I started to work on the project around '96. But it is essentially what happened. I told very few people. The word came back from the Santa Fe Institute — what has Arthur really been doing? He has some very interesting early work and some promising work, but we haven't heard anything from him in years.
At any rate, for better or worse, I became convinced that there are so many questions unanswered in technology. What really is technology? How does it work? How does it develop? Is there an overall theory of evolution? Technology was actually a goldmine.
I began to realize that all technologies, all new species of technologies, such as the laser printer that evolved out of Xerox PARC, are composed of already existing elements. When Gary Starkweather started to work on the laser printer in the early 1970's, producing a working version in 1972-1973 down in Palo Alto, he was basically saying, we don't want line printers. They can't print images. They can't change font size. They are basically computer driven typewriters. After a lot of thinking about other ideas, say, writing on cathode ray screens and all, he had the idea that maybe he could get a computer to control a laser beam, very highly focusable, and write (or the word they would have used is paint) an image onto a Xerox drum. So the elements existed: a computer, computer controlled lasers, the elements to control the laser operation and Xerography. And we got the laser printer.
When I started to look at any new species of technology, be it a jet engine or laser printer or sorting algorithms, all of their components already existed (or could be constructed from things that existed). I began to realize that it was possible to put a Theory of Evolution together with combination at the heart of it. So what I'm saying is that technologies evolve from previous technologies by selecting building blocks and putting them together in new ways.
Some people had realized this, and some had written a bit about it, but most of the writing was maybe a paragraph here and maybe a few sentences there, by somebody really smart like Schumpeter 100 years ago, who hadn't gone beyond some preliminary ideas which are very widely quoted. Nobody had worked out this idea about combination. I came to it independently, but didn't do much about it in the '80s. Then I discovered a lot of kind of hand-wavy literature in which people talked about it.
I was faced with another question, and that is, if everything is a combination of something that existed before, we don't have sophisticated technologies like magnetic resonance imaging constructed out of flint or obsidian or whatever we had 10,000 years ago. It's not the original components of 10,000 years ago that are getting combined to give us something really sophisticated like magnetic resonance imaging. I began to realize there is another leg to it, and that is that every so often technologies are used to capture phenomena, in this particular instance nuclear magnetic resonance.
Every so often we use instruments and modalities, methods and technologies, to capture some phenomenon, say, that you can affect the nuclear spin of certain atoms and use that to make certain measurements. That becomes some sort of diagnostic technique. So the two legs of the Theory of Evolution that are in technology, are not at all Darwinian. They are quite different. They are that certain existing building blocks are combined and re-combined, and that every so often some of those technologies get used to capture novel, newly discovered phenomena, and encapsulate those and get further building blocks. Most new technologies that come into being are only useful for their own purpose and don't form other building blocks, but occasionally some do.
What really excites me about this is not so much technology. What really interests me is that astrophysics or cosmologists have a very similar idea, both of the formation of life on earth and of the formation of the universe. Now this is off my beat so maybe I'm a little hazy on this, but after the Big Bang, if you wait long enough — maybe 10 to the minus 27 seconds, eternities of very short magnitude — elementary particles, whatever, quarks, begin to combine in different ways to form elementary particles, which begin to combine to form the basic hadron building blocks, which further combine to give you atoms and the molecules, which over time lead to very rapidly expanding gases, which in turn form stellar systems.
All of these steps are formed by combinations of elements forming new building blocks that give you further combinations. The same is true of early life. I have been talking to people recently at the Santa Fe Institute and they are talking about various reactions in what could be called the evolution of metabolic pathways.
You get terribly simple metabolisms forming very early life, like four billion years ago. Then they form certain elements that in combination can give you further elements that are catalyzed by some building blocks that give you yet newer molecules. So the whole of what we call life, building up to RNA and then DNA, form out of structures that are combinations of simpler ones that give you combinations of yet further ones.
So in one phrase, to go back to technology, my argument up until my project, I think it's safe to say, was that if you ask people, "What is technology," as a whole they would have said it's a bunch of standalone methods or devices: the Solvay process, the computer, laser printers, and so on, that are sometimes interrelated and have some sort of ancestry.
What I'm saying is no, all of this forms a gigantic chemistry, that the simpler molecules that have formed in technology — the computer, the laser, Xerography — those molecules can be put together to form yet a new molecule, the laser printer, which can be put together to form something more complicated. Technology viewed as a whole is chemistry and its chemistry is still building.