Brian Arthur
Santa Fe Institute

I wasn't sure what I was going to be talking about and then I saw a statistic yesterday in the Wall Street Journal that labor productivity had risen in the third quarter of this year by 8.1%. I don't expect you to get quite as excited as I got over that number-it's a number that would excite economists. But this is something of a mystery. Since 1995 productivity has been steadily increasing at quite a rapid rate in the US. We see figures like 5% or 3% per annum. These are large figures historically and there's been a lot of argument as to what's going on. We tend to think of numbers like that as rates, meaning the same thing is happening but more of it. But actually productivity statistics are different. They're telling us that something inside is changing. Something's being done differently and going through a transformation. That's what I want to talk about today: what the transformation really is.

I have a daydream that I go through every once in a while. I imagine myself being alive in the year 2100. If I were an economic historian looking back over 100 years time and it was pretty clear what happened-the real story of the economy-what story would I be telling students at that time? I believe in story and what I want to talk about is the story of what I think is actually happening now. For an economist, that's much more challenging. It's easier to look back 150 years and talk about what was happening then: railroads were emerging and so on. The prime candidate for the story of the economy at the moment is information technology. It's a bit too early for biotech and nanotech. They're coming along but not really directing what goes on in the economy.

If you look at information technology as just so many machines and teraflops and servers, you can argue that we have as much of that as we can handle. Businesses have absorbed huge amounts of this sort of thing. We have the Internet and all the fiber optics in the ground that we could want. Then you might conclude that all of this had become commoditized and it wasn't very interesting.

I find it interesting whenever I get lost over what's going on in the economy, to turn to history. Not to actual numbers, but whether I can fit what's going on into some sort of historical pattern. The answer is "yes." Think of it this way. The information revolution, or the digital revolution is merely the latest of five major technological revolutions we've had so far. The first one, the industrial revolution, happened in England. The dates are a bit fuzzy, but it was roughly from 1780 to about 1830. Think of that as the revolution that brought big textile machinery and mills and the factory system (largely water-powered). The next revolution, again pioneered in Britain, was the railway revolution. This was steam powered and it went on from 1830 to about 1880. The next revolution shifted to Germany and had to do with electric motors and heavy industry like steel. That's the era of the Brooklyn Bridge. The next revolution, manufacturing, is pioneered in the US from 1913 to the 1970's. This includes mass production, automobiles, big oil. Then there is this nascent revolution-the digital or IT revolution-and that started in the US in the 1960's. The ARPANET was 1969. The first Intel chip was 1971. That revolution is still going on.

When you look at these revolutions you see a pattern. At the start there's a huge amount of turbulence. Engineers are vying to get these new devices. Tiny companies are stealing ideas and jockeying for position. Devices that look very promising get suddenly sidelined as new things come along. The whole technology starts to fill out in the first ten to fifteen years. Then it becomes a matter of public attention. Newspapers discover all of this. When I looked at the railway revolution I went back to newspapers in 1836 and they were talking about a new economy and how different things would be. The next pattern that follows is frenzy of investment that leads to huge speculative frenzy followed by a crash. So each revolution is characterized by turbulence, media attention, huge amounts of liquidity coming into that technology, all kinds of schemes cooked up, things go wrong, and then it all crashes. The crash in the railway revolution happened in 1847 and what amused me was to go back and look at the Economist magazine of the day. There was a very familiar pattern in 1848 and 1849 where executives were hauled in front of boards of inquiry and stocks were worth 15% of their peak value three years earlier. There was a complete hiatus in investment. Nobody wanted to hear about railroads.

What happens is that there is a crash. People assume that that particular technology is all over. The newspapers lose interest. The technology loses its glamour. People don't want to be told about it or invest in it. But what I find interesting is that the technology goes on in the decades that follow, each time into a massive build-out. Railroads in Britain railway track increased by a factor of 1,000% after the crash. The US had a similar crash in 1859 after which railways increased by a factor of 800%. Moreover, these technologies after a crash and during that huge build-out bring prosperity. I began to wonder why. There is a beautiful book written on this by Carlota Perez called Technological Revolutions and Financial Capital. It's an interesting book about this sort of thing. I began to wonder what happens in that build-out period. It's more than railroads building more track or carrying more passengers. What happens is a very deep transformation of the economy. If you look at the US case, by 1860 railroads are just getting started. There are 30,000 miles of track. Eventually there will be close to 300,000 miles of track. But look at what railroads did from 1860 to 1900. They connected the eastern part of the US economy with the western part. They made a much larger economy with attendant economies of scale. They made inputs cheaper so that if you were shipping iron ore or leather, those costs fell by an order of magnitude. Coal, for example didn't have to be dug up locally and used locally. It could be shipped several hundred miles. They brought new industries into being including the slaughterhouses of Chicago. Herds could be run by railroad to Chicago, the meat could be canned and then shipped to the East Coast. Entire new industries came into being.

I realized that if you had run statistics on the economy at the time, you would have seen major productivity increases. Maybe economists would have scratched their heads and said that the railroad thing was over hyped. The US economy on the international scene in 1860 before the Civil War was a backwater economy by world standards. By 1900, 40 years later, due to these transformations it was the largest in the world. Then the US goes on to build upon that in the mass manufacturing revolution that succeeded it. So this led me to wonder whether the same could be said of information technology. I agree with everything that Jeff Bezos said this morning and Bill Gurley as well. But I have a third take on this. Suppose you said that information technology is a revolution like the ones that preceded it. We had our period of turbulence with Microsoft and Intel coming out of nothing. We had primitive technologies and PC's and a lot of stuff in the 70's and 80's elbowing for room. Then we had a huge amount of media attention in the 1990's. Then we had an investment frenzy and the dotcom thing and the crash. Then we had all these people hauled in front of government committees and so on. So history repeats. But if that's the case, are we going to get a massive build-out in the years 2000-2030? And what will it be like? What are the transformations that information technology is bringing at the moment?

I want to point to three transformations. By the way, sometimes I use PowerPoint slides and I prefer not to, but what I'm saying is written up in an article that will appear in about a week's time in Fortune magazine. I think they titled it, "Why Tech is Still the Future." When I talk to CIO's I ask them what they're really doing all the time. Something very standard comes out. Every company everywhere is doing some version of connecting things to things and having them converse. So devices are connecting to devices in wired-up network ways and wirelessly. Machines are connecting to machines. Business processes are connecting to business processes and to machines and to devices and to human beings. Networks are being connected to networks. There's a huge standardization that has to be gone through to facilitate this. And then something will happen. The first deep transformation I would characterize like this: business is slowly but steadily emerging as a collection of automated conversations among business processes. Think of it this way. About twenty-five years ago, navigation aboard an aircraft was a matter of the navigator with calculators and radio fixes. He passed the coordinates to the pilot and the co-pilot. Now he's gone. In his place is a conversation between on-board navigational systems, a GPS system, stationary orbit satellites, ground stations, other satellites, and a fly-by-wire automatic electronic/hydraulic system that automatically steers the airplane. What happens here is not something about having a computer on board that is handling the navigation. Instead, the functionality of navigation itself has become a set of conversations.

I was in Italy a few years ago, and the espresso manufacturer, Cimbali, told me that they were testing a new system that should be in place now. They found that after 1,000 cups of espresso their machines would go out of sync. To keep getting a good cup of espresso there are relationships between some of the parts that must be absolutely correct. They were sending out repairmen each time and this was expensive. So instead they put sensors in their espresso machines that monitored the different parts, and every so often the sensors would contact headquarters in Milan to receive new settings by satellite and Internet. The machine would then self-adjust. What's happening is almost invisible. If we could see all these conversations we'd see the air full of them. Everything is becoming connected to everything. Occasionally there is human intervention and supervision but a great deal of the economy is becoming an automated conversation among machinery, processes and devices.

There's a second transformation. Once these machines and devices are all connected and talking with one another there comes a natural opportunity and temptation to make things smarter. Once the system is connected, you can make it a lot more intelligent. So imagine, for example that you have all the traffic lights in Los Angeles and they all begin to talk to each other. Along every street in Los Angeles, sensors tell this system how much traffic is moving along that route. These traffic lights start to talk to each other and tell each other about the traffic on their streets. There is a strong opportunity at that stage to automate the system and make it smarter. It may be that it speeds the traffic on one street and slows it on another to avoid a traffic jam on one of the freeways. What's emerging is networks of sensors that are becoming intelligent.

This is happening in spades at the moment in supply chain systems. We have RFID (radio frequency identification) technology. These are tags that go on containers and trucks and even individual product packages and they talk to each other and the depots they're passing through. Once you get a system like that operating, you can get some sort of primitive intelligence emerging.

Another example comes from the military. It's either cute if you like this sort of thing, or maybe a little bit horrifying. The Army last July did this experiment. They had a large minefield, maybe a mile wide and they drove a tank down the middle of it with an enormous net like a huge snowplow to clear the field in front of it. The mines in this field are sitting on the ground and about the size of small melons. So far this is standard operating procedure. But these mines have sensors and they have RF ability to talk to each other. They have a primitive form of intelligence built into the system. They had also little miniature rocket devices. As the mines on each side of the cleared area began to communicate with each other, they discovered they weren't communicating with anything in the middle-there was a gap there instead. They were preprogrammed if they discovered a gap to use their little rocket motors to hop towards the gap. So as the tank clears the minefield, you look behind it and see all these little mines hopping to fill in the minefield behind the tank as it clears. There's a sort of Michael Crichton movie feel to this. It's just one step away from having the mines sense the tank as it approaches the minefield and hop into its path.

This is research that is going on at the Santa Fe Institute and at XEROX PARC. You can't organize these types of intelligent networks top down. You can't preprogram the whole Los Angeles traffic system with some consultants. It gets too brittle and too automatic and gets into trouble if it encounters circumstances that the programmers did not foresee. Instead, a very primitive intelligence must be programmed into the devices themselves like the little mines. All they can do is contact one another, detect gaps in the signal and then hop towards the gaps. This primitive intelligence builds into something much bigger.

If I put these first two transformations together-everything is getting connected and everything is getting smarter-what that tells me is that something in the economy is emerging. There's one definition of a cognitive system that I like. A system is cognitive (like Ecoli that can sense a food gradient and move up the gradient) meaning it has a primitive sort of intelligence if it can sense something and react appropriately. So when I say that everything is starting to connect and converse, what I believe is emerging is a rather primitive and powerful neural system for the economy. Humans will still be part of this. We're supervising and watching what happens, but this is a set of computations that are happening when things are sensed. When someone logs into Amazon.com, Amazon senses them. The servers know that someone is there and they react appropriately. I wouldn't say that it's primitive intelligence but what's emerging is a primitive neural system for the economy and this is entirely new-we have seen nothing like it.

There's another thing happening, and this is the third transformation. When you start to look at the deep revolutions we've seen so far in economic history, you see completely new industries. So I was wondering how many new industries have we seen with information technology besides computation itself. Surely, information technology is just adopted and then it is diffused. The business that adopted them does what it does before but does it better, faster, cheaper. Nothing really new is happening, you could argue. But this isn't true. When industries meet up with information technology, traditional or not, and don't adopt IT, they encounter it. This means they come up against information technology and in that encounter something new and different and surprising nearly always happens. What happens is that traditional industries take some of their activities and combine them with some of digitization's activities and create things that are entirely new.

Here are some examples. The movie business is a traditional business. The actual technology hasn't changed a lot in a good 80 years or so since talkies in 1929. The movie business is still a matter of capturing images and projecting them on a screen. Now you can take those images and with digital functionalities you can change it, warp it, morph it, color saturate/desaturate it, lay in animated layers with human beings, and get something entirely new. You have a sub-industry there of digital special effects. Banking went through this in the 1970's. Banking had certain activities that had to do with swaps, futures and options, and when you combined those with digitization you got completely new combinations and sub-businesses. Modern financial risk management was one of the outcomes. A great deal of the derivatives business would be impossible without information technology. The same is true in science. Genomics is not possible without information technology. It's the combination of components of molecular biology and certain capabilities that computers bring like gene sequencing algorithms. From there we get gene therapy and modern diagnostics using DNA arrays. Digitization is creating entirely new businesses and new industries.

Let me summarize these three transformations. Business is steadily becoming a conversation among devices and activities and machines and business processes via digitization. Second, once you have everything connected in these networks, the networks become very smart, changing how business is done. Finally, digitization is bringing in new industries. Nanotech is impossible without it and so is biotech along with more mundane things. There is one fly in the ointment in all of this, one downside. It's a pretty optimistic picture overall. The digital revolution is far from over. It's only in its early days. But there's a difficulty. The technology itself is going offshore. For the first time, we're finding very unlikely countries giving the US a run for its money in technology: India, China, Finland, and Ireland. Twenty years ago none of them were considered contenders in technology. Yet we're starting to get worried here that India or China will really take over. There will indeed be a squeeze from the low end. Ireland is the largest software exporter in the entire world. Bangalore has a great deal of software technology as well. China is coming online in the production of all kinds of devices and hardware. This may well squeeze companies like Intel. The US will be relying on staying on the edge in high tech. Can it? I think that it can. I went back into economic history. What you find is this. The countries that lead in advanced technology (England, then Germany, then the US) are the ones that lead in science. It's not a coincidence. If you want to put together something like the laser, you must understand quantum physics. There were two places that understood it better than any place else in the 1940's: Columbia University and Bell Labs. You really have to do science at a very advanced level. The US is far enough ahead in science to maintain its lead. It need not worry that China will take over in the next few decades because China doesn't have the advanced science. They have good science but the US has enough of the cutting edge to maintain its lead into nanotech, biotech and quantum electronics. But there should also be no cause or place for smugness or complacency. Countries like England were doing well in 1880 but they didn't support science and then Germany took over. The US will be able to maintain its lead, providing it maintains its lead in science.

Will Japan and Germany come back and take over? Probably not. What they did best was mass manufacturing on a large scale. They haven't broken fully into the digital revolution, nor are they fully at the edge of science the way they might have been otherwise. I don't think they'll come back. I find it significant that it's small countries like Finland that were largely agricultural that are taking over. It's because they didn't have manufacturing before, nor did they have all the institutions like large labor regulations that go with that kind of economy that could hamper them from getting into high tech. So they are small and nimble. India is in the same position. Yes, the US should watch and be extremely vigilant and there will be lots of pressure from the low end with high tech manufacture.

I was fascinated by what Bill Gurley said yesterday about IT mattering and being the main thing that's happening. I agree. I agree also with Jeff Bezos that it's day one for the Internet. I want to go further and say that if you look back in 100 years time, the story we'll be telling is not one of interest rates or bond rates. It will be instead of deep transformations in the first two decades of the 21st century when things started to talk intelligently to one another and a primitive neural system for the economy started up. Completely new industries took off that no one ever thought of before. And a lot of technology went abroad but at least for the first two or three decades the US managed to maintain its lead. It's a rather optimistic scenario but I think it's realistic. The first two revolutions in Britain that were copied elsewhere provided heavy machinery and energy for the economy. They provided a muscular system for the economy. This digital revolution isn't providing muscle or energy. It's providing the neural system for the economy. In the long run, it will turn out to be much deeper and more important than the industrial revolution.

Q: Can you explain increasing returns and how does it differ from economies of scale?
A: I'll hark back to the work I did in the 1980's-very technical economic work. Basically, increasing returns is about things that have advantage gathering further advantage. I published an article in Scientific American in 1990 called Positive Feedbacks in the Economy. It was very widely read. A year-and-a-half later I found myself in a Senate office trying to explain increasing returns on a summer day when everybody was rather droopy. It was the Senate office of Senator Al Gore. It was later in the afternoon. I was going into technicalities. Gore says, "just give me an example I can understand." Al Gore is a very bright guy. I said, "think of it this way; Presidential primaries." Everybody sits up at once. You have eight candidates. If a candidate looks more likely to be president, they attract more money and then they attract more television time and they then look like an even better bet to be president. It's a feedback loop. This happens for all the other candidates. In Iowa you have these competing bandwagons and if one of them starts to roll a bit faster and gets more momentum, that momentum builds up and one candidate can completely lock in the candidacy for that party and shut the others down. We feel very strongly in California by the way that our primaries were so late in the game that they never counted because the winner was essentially locked in throughout the Eastern states. Microsoft MS-DOS vs. the Mac OS in 1983 is the same kind of story. If one gets a sufficiently good start, it gets better contracts and adherents and it locks in. People are writing software for that operating system that it gets enough momentum that by increasing returns it locks out something else. Increasing returns means more advantage accrues to early advantage. Many of the things that are locked out were superior to the ones that took over. Any computer person would have told you in the mid 80's that MS-DOS was a kludge. The Mac system was mean and clean but at its peak it got about 12% of the market. MS-DOS took over and it was parlayed into Windows and then into Windows NT and so on. Small historical effects can make large differences. If something is absolutely terrible it won't take over the market, but things that are not necessarily the best can take over. When I said this in the mid 1980's it was regarded as a shocker. I remember saying it at a meeting at Harvard and one of the professors got really upset and said, "under perfect capitalism we are assured of the best of all possible worlds." This was technically not true either. I gave the same talk in Moscow during the Soviet days and I said that by increasing returns something can get ahead and get further advantage and lock in and not be the best of all possible worlds. A guy who was a member of the Supreme Soviet got upset and said it was not possible: "under superior socialist planning we always choose correct solution." I figured with the capitalists on one side and the communists on the other I had to be correct.

Q: Can you give us some idea of what impact this high level of productivity will have on both inflation and employment?

A: That's a good question. One thing I want to mention, prompted by Joy Covey, is that it's worth mentioning who's going to benefit from these changes I'm talking about. It's not entirely clear that the main beneficiaries will be the tech companies themselves. There could be so much competition and so much commoditization that successful companies like Microsoft could have competition from similar companies and there's not that much profit to be made. The tech companies will do quite well. But the main beneficiaries will be two classes. One is the consumer. If it is true that everything is getting connected and if traffic does start to flow after all these eons in LA, then we will all benefit in very subtle ways including the quality of what comes to us. If the supply chain delivers in a smarter and automated way, there will be inventory costs saved and supply will be cheaper. Things will come to the consumer more cheaply. The main beneficiary will also be the companies that are in the lead of introducing these things, like Walmart and FedEx. A great deal of Walmart's success is not what they put on the shelf-anyone can do that-but a keen appreciation over the years of what digitization can bring. That story is long from over. Watch the standard economy using the new technology. In the railroad days you wanted to watch a coal company, not a railroad company-someone using the new economy to make more profit.

What about jobs and inflation? If technology allows us to do the same things more cheaply there won't be the push for inflation. Yes, I think inflation is a kind of natural thing that happens but if the same things can be done for half the price and then half that price in five years, there's deflation technically. Whole sectors of the economy will be doing this. If productivity continues to increase that will expand the economy with the same labor force and that should offset inflation. I wouldn't say we will see no inflation.

As for jobs, I think that's a real problem. If you look at today's New York Times there's been an increase and companies are starting to hire. But in the longer term, two things are going to happen. One is pretty obvious: as everything becomes more automated, people are going to be automated out of jobs to some degree. But whole new industries are starting up and they're still nascent. They will be able to employ more people. And the baby boom people are going to start to retire in five or ten years and that will lead to a lot of empty positions and that's good news for jobs as well.