The Life Scientific: Doyne Farmer

0:00This BBC podcast is supported by ads outside the UK.

0:30at Whole Foods Market. Apple Vacations, where your story starts. The July Savings Event from Apple Vacations is here. Book by July 23rd and save up to$400 off your flight and hotel package to Mexico, the Caribbean, Central America, Hawaii, or Europe. From all-inclusive escapes to bucket list adventures, Apple Vacations makes it easy to create memories with the people who matter most. Start today at applevacations.com or contact your trusted travel advisor.

1:03My guest today is something of a rebel. Back in the 70s, when he was 24 years old, he walked into a casino in Las Vegas, sat down at a roulette table and beat the house. To anyone watching the wheels spin and the ball clatter to its final resting place, Doyne Farmer's choice of number would have looked like a lucky guess. But knowing the physics of the game, and armed with the world's first wearable computer, which he'd designed, seemingly random wins were actually somewhat predictable. Randomness, says Dorn, is a concept that depends on the state of your understanding, something he's shown time and again over his career.

1:42Professor Doin Farmer is an American scientist and entrepreneur who pioneered many of the fields that define the scientific agenda of our time. Chaos theory, complex systems and wearable computing. Today he uses big data and ever more powerful computers to apply what's called complex systems science to the economy to make better predictions about our future. Much like roulette, the way the economy works can appear random, but with the right tools and understanding, they're anything but. Now director of the Complex Economics Programme at the Institute for New Economic Thinking at Oxford, Doin says there's a real need to use these powers of prediction to help resolve one of the most pressing questions of our time, how best to prevent climate change.

2:28Doin Farmer, welcome to The Life Scientific. Very happy to be here. We'll get to how you beat the roulette table a little later, Doin, but in some senses what you were doing, making the seemingly random somehow more predictable feels like the unifying thread throughout your career. Back in your roulette days, how accurate was your prediction? What edge did you have over the house? We had about a 20 % edge over the house, which meant that we could effectively predict six or eight numbers that the ball was very unlikely to land on. Oh, you say we, I mean, you're part of a big team. We had what we called the pie.

3:02Everybody got a portion of it based on how much they worked. had a real thing going there. Well, in total, there were 30 people who owned at least a small piece of the pie. What was it then that prompted you to stop? Well, every time we started to really push the stakes up, we felt a lot of casino heat. There were credible stories about people getting beaten up in the back room, and we decided our kneecaps were more important than making a lot of money. And we also were experiencing a lot of hardware problems. We were on the vanguard of computing at that time. This was 1978. And we built the first wearable computer.

3:39And we probably spent three hours fixing equipment for every hour we spent in the casino. And did you have any close shaves? There were several times where I scooted out the door as fast as I possibly could. Well, Dorn Farmer, let's wind back to your childhood. You were born in 1952 in Houston, Texas, but your family moved to Silver City, New Mexico, when you were five. What do you remember from that? Well, Silver City is a remarkable place. It's about 50 miles to the next town with nothing but rabbits and ranches in between. It was where Billy the Kid grew up. It was a tough place. I got beat up quite a few times, but it was overall, I think, a really good place to grow up.

4:19What about your family? Well, my father was a mining engineer, so he went off to the mine every day, five days a week. In your team, someone came into your life who had a big influence on you. That's right. I was a member of a Boy Scout troop, and a fellow showed up named Tom Ingersoll, and he was introduced to us as a physicist. I knew it had something to do with Albert Einstein. I'd seen a show on TV about relativity, and I thought, wow, I want to understand that. Well, of course, apart from being cool as a physicist, which I wholeheartedly agree with, what made him so inspirational? He was a remarkably bright person.

4:56I've met, I think, some of the most intelligent people in the world in my career, and he was right up there with the best. He was 27 years old. His dream at the time was to find his great uncle's gold mine and build a rocket so we could go to Mars. So need I say anything more as a 13-year-old, you can imagine how big my eyes got. Wow. And I gather at the age of 14, you actually moved in with Tom? I did. My family moved to Peru. There were some problems in my house and I asked my parents if I could just go live with Tom. So it was really a fantastically lucky thing. My friends would come over and hang out there.

5:34We all fixed our motorcycles over at his place. We built electronic equipment. So it became a kind of community centre for the smart kids in Silver City. It sounds like you had a great time. Your lifelong friend there, Norman Packard, who I'm sure we'll be mentioning later, was a big part of your subsequent career. He was also much inspired by Tom, wasn't he? Very much so. Norman was two years younger, and he was a very geeky kid, kind of tiny, big head. When he got introduced to our Boy Scout troop, he held up the crystal radio that he had built, and we passed it around and all looked at it and listened.

6:10Norman and I became good friends. We were both pretty serious science kids. And it certainly sounds like much of your education happened outside of the school. Well, I would say all my education. Everything I learned of substance, I learned from Tom or out of a book. And by the time you went to college, you were convinced that the only true path to understanding how the world ticks was through physics. Of course, I agree with you. In 1970, you got a place to study physics at Stanford. but you had a lot of big questions that you're looking for answers to, which by that point you felt physics couldn't answer.

6:44Yeah, that's right. I mean, I had good physics courses, but the things I really wanted to understand, I wasn't finding out about. When I was about 15, living with Tom, I read a book by Pierre Teilhard de Chardin, a very remarkable Jesuit priest who was also a physical anthropologist and a big thinker. He had this vision of the noosphere, That is technology, culture, biology, all evolving in tandem. And I was fascinated by that. Where are we going in the long run? And how does that interaction work? And in the 70s, certainly nobody was talking about that. Nevertheless, three years later, you moved on to the University of California, Santa Cruz, to do a PhD in physical cosmology.

7:30Why that particular choice? I wanted to understand fundamental origins of everything. And I found it fascinating thinking about quasar formation and galaxy formation. But I realized a lot of smart people had picked over these problems before. And, well, I got distracted. Well, this leads us back to that roulette table. How did the idea to beat the odds at roulette come about? Well, there was a summer where I was working as a construction inspector in Montana. and Norman was trying to make his living as a blackjack counter in Las Vegas. Norman didn't do so well on the blackjack. I, in contrast, started playing poker on the weekends and poker had just been legalized in Montana,$100 pot limit poker.

8:16And I realized I was making more money playing poker on the weekends than I was making working for the Forest Service. So at the end of the summer, Norman and I had a big debate because we both needed to make money outside of school. Norman was standing watching the roulette wheel and realized that roulette was just a physical system, so we should be able to beat it. But it's a physical system that is very, very complex. So how can you use physics to beat the roulette wheel? Yeah. To a casino, a roulette wheel is a random number generator. To a physicist, a roulette wheel is a rolling ball on a beveled circular track with a counter-rotating rotor in the middle.

8:55If you know the position and velocity of something and you know the force is acting on it, you should be able to predict its future motion. Now, the other key fact about roulette is there's typically about 15 seconds that elapse between when the croupier spins the ball and when the bets are closed. So we realized in those 15 seconds, if we could just measure the position and velocity, we should be able to predict where the ball was going to land. But you needed to build quite a bit of kit to help you calculate the ball's final landing position. We certainly did. We ended up building the first wearable computer, which was bigger than a package of cigarettes, under one armpit, 12 AA batteries under the other armpit, which is what it took to run the computer for an hour and a half.

9:42I had programmed the computer. That took me about a year and a half to write that program. Wow. It had to be written so succinctly that it would fit in just 3 ,000 bytes of memory. But it contained the forces, the timing, operated all the equipment. But this is the 1970s, we have to remember. Yeah, we started building it in 1977 with a very first commercially available microprocessor, the 6502, the same one that Steve Jobs and Steve Wozniak used to build the first Apple computer. We did this at the same time they were doing that. So you can see the challenge. They were first building a home computer, but they could build something that was the size of a small case.

10:22We had to build something that was tiny. Now, just to complete how the rest of my kit worked, we had switches in the toes of our shoes that we could operate with our toes, wires running up our trousers, and a little plate with buzzers on it under our belt, and a radio transmitter wrapped around our shoulders. So you had this wearable computer, and you knew the physics, but at the roulette table you needed a partner in crime. So one of you would calculate where the roulette ball was likely to land and the other would then place the bets. Take me through how that worked without anyone noticing what you were doing.

10:57One of us, almost always me, was the data taker, meaning the person who collected the information to make the prediction. The second person would place the bets. Right. Usually a woman because we thought women were less likely to be suspected of doing something like this. the data taker had to measure how fast the ball was going. To do that, I would tap the switch in the shoes whenever the ball made a circuit on the wheel. Then the computer under my armpit would make the calculation and predict where the ball was likely to come off the track. Then buzz against my stomach that would tell me what number the ball was likely to land on.

11:39And when I say the number, we divided the wheel into eight octants. So these are segments of a pie. So you wouldn't predict exactly what number, but a group of numbers within a particular segment. That's right. And it would transmit the signal to the other person. And so they would casually lay their bets down. Meanwhile, I would be sitting there looking like some spaced out geek, just betting on red and black at very low stakes. Well, you did make a bit of money, but you eventually stopped, as you've explained, computer hardware problems, but also that fear of getting your kneecaps broken. But your roulette experience did prompt you to switch your PhD topic to what's known as chaotic dynamics.

12:20Yeah, the biggest source of randomness in a roulette wheel is when the ball leaves the track, a small difference in where the ball goes can cause an enormous difference in where it goes subsequently. And so when I heard about chaotic dynamics, I thought, wow, that's exactly what we've been doing in roulette. But it plays out in many different systems. And it was a new idea at the time to try and understand how randomness came out of order. It was such a new idea, in fact, that there wasn't even anyone to supervise your PhD research. So how did you manage? Four of us banded together, Norman Packard, my partner in roulette, Rob Shaw and Jim Crutchfield, and we co-supervised each other.

13:03Well, your PhD thesis title was Order Within Chaos. And you were looking at the weather, which, of course, is a great example of chaotic behavior. We looked at many different systems, actually, not just the weather. And we found that, indeed, some of them were chaotic. Our main contribution was to be able to take data and analyze that data and draw pictures of the chaos. And you could identify what kind of chaos it was. The chaotic behavior of weather systems. Many listeners will probably have heard of the butterfly effect in weather. Ed Lorenz, who originally coined this phrase, said that a butterfly flapping its wings might influence whether a hurricane goes here or there.

13:43Ed Lorenz is regarded now as the father of chaos theory. You actually spent some time working with him, I gather. Yeah, I was lucky enough to be offered an internship because I was short on money. I camped out next to the National Center for Atmospheric Research where I was not supposed to camp. So I had to hide my tent every morning under a tree and set it up again in the evening. I began stopping by every day to ask Ed whether it was going to rain because if it wasn't going to rain, then I didn't have to set up my tent. I could just sleep outside. And Ed was a wonderful guy, would look really carefully at all the clouds and stroke his chin.

14:18he'd go, nope, not going to rain, or yep, it's going to rain, or can't tell. And he was never wrong. Well, you got your PhD in 1981, when chaos theory was just starting to find its way into popular culture. So I want to jump forward briefly here, because in 1993, Steven Spielberg's film Jurassic Park came out. Jeff Goldblum plays a mathematician specializing in chaos theory, and I gather he called you when he was preparing for the role. Yeah. So the phone rings in my house and I pick it up and this guy says, hi, this is Jeff Goldblum. Do you know who I am? And I went, yeah. And he said, well, I'm going to be in this movie about dinosaurs and stuff where I'm playing a chaos scientist.

15:02So I want to find out how they talk. So do you mind if I talk to you for a while? Wow. And so we chatted on the phone for half an hour or something. And I was quite thrilled when the movie came out to see that several of the things I had said actually appeared. What was it that came up in the movie then? Well, it seemed almost out of place. I remember one of the questions he asked me was, how many children do you have? I said, I have three children. And so that's part of what he says in the movie. Why he had to say that, I don't know. There's something about a glass of water? Yeah, that's right.

15:33He's trying to seduce Laura Dern, despite his wife and three children, And he starts talking about the way the water swirls as an example of chaos. So trying to bring her in with the beauty of nature, I suppose. A good chat up line. Back now to the early 80s again, fresh from your PhD, Dorian. You went to Los Alamos National Lab in New Mexico of Manhattan Project fame, of course. But you were at its then newly formed Center for Non-Linear Studies, where you developed an interest in what's now called complex systems science. What is a complex system? So a complex system has what we call emergent phenomena.

16:10The system as a whole can do things that are surprising from the point of view of just looking at the individual building blocks. So take ant colonies. Ants are very simple creatures, but ant colonies, on the other hand, can ranch aphids, fight wars, and take slaves. And this all emerges from the way they communicate with each other. If you just looked at an individual ant, you would never guess. Well, we're talking about ant colonies here, but you and your friend, Norman Packard, were developing computer models to simulate all sorts of complex systems. By making small changes to your input data, you could see how your model system would end up.

16:50So running these simulations many times and even observing some of this emergent behavior. What systems were you looking at in your computer models? Well, we were very interested in the immune system, how the cells in your immune system can signal to each other, and how the ability could emerge to recognize what's self and what is other, which is what the immune system has to do. We also looked at the origin of life and how could you have a precursor to life that would have spontaneously emerged to effectively set the stage for life to emerge. By the late 80s, you were becoming something of a world leader in the field.

17:27And in 1988, you founded the Complex Systems Group at Los Alamos. But around that time, you were also getting interested in how complex system science could be applied to economics. Why economics? I'd taken economics in college and I didn't really believe it, to be honest. And then in 1987, I was invited to a meeting where a very famous physics Nobel Prize winner, Phil Anderson, and a very famous economics Nobel Prize winner had basically each picked a team of about 10. And we spent two weeks having very intense discussions to see whether physics and things like chaos might have something to do with economics.

18:07In parallel, Norman and I had been developing methods for predicting things. Given some data, if the data was coming from chaos, could we actually use that to make better predictions? That might sound paradoxical, But actually, because there's order in chaos, if you identify that order, then you can actually use it to make better short term predictions. What was wrong with the standard economic models? Economists operated under the idea that people are rational, meaning they always take the best possible choice. So if you're a firm, you do everything right. If you're a household, you make the right decisions all the time.

18:43And we just felt that was very far fetched. So give me an example of when a standard economic prediction was wrong. Certainly one of the most dramatic was the financial crisis of 2008, where the whole system for lending collapsed globally. Why weren't economists able to predict that crash? Well, because in their models, they assume everything is optimal and everybody's rational, missing the way people really act. We're not perfect. We respond to things emotionally and we respond to things collectively. We heard. And that really wasn't in the theories at the time. So when there was a downturn in the housing market, everything went to hell.

19:25Well, in 1991, you set up Prediction Company with your old friend Norman Packard. What was your ambition? Well, Norman and I thought back to our roulette days. We wanted to find a casino where they didn't break your kneecaps and where you could bet as much as you want. So I bought a suit, first suit I'd ever owned, and spent a year traipsing around Wall Street, found a partner, and they basically gave us the money to bet in the stock market. So how did you get on? Well, it took us about five years to make a system that worked. But once we got it working, we did very well. So there you were applying the science of prediction to the stock market and winning.

20:04How did this go down with those working in finance? Word began to leak out that this kind of thing was working, and then lots of other people started doing it. Ten years afterwards, it became a lot tougher because there was a lot more competition. I guess like any complex system, to get a better idea of what will happen to the economy in the future, you need to understand the system from the inside. So we're talking about individuals, businesses, households, how they interact. And that'll often produce something quite different from what you might initially expect. This is what you were doing, right?

20:35This is what's called complexity economics. At Prediction Company, we were actually just taking the data and building models like you do in AI or machine learning. We just knew if this happened, then the market was likely to go up. But that made me really ask why. So I started reading more economics papers, and it felt like they were really thinking about things the wrong way. In complexity economics, we instead assume that people or firms take in information, They make decisions, act on those decisions, which then changes the economy, which generates new information. And they make new decisions and then the process repeats itself.

21:16So there's no assumption that things are going to settle down. We actually use computers to simulate what the world is going to do. OK, Donny Farman, I want to jump forward to 2012. When you moved to the UK and Oxford University, it was there that you hired to run its complexity economics program. What was the attraction? It was a chance to really advance the state of the art and do things on a bigger scale. And I gather that when the pandemic hit in early 2020, you decided to predict the economic impact of COVID. Why? COVID didn't just make people sick, it affected the economy very strongly. So we thought this is our opportunity to build a model that will really help the world.

21:59And what sort of data did you need to collect to try and predict the different scenarios? We first of all had to predict how big the shock would be. My ex-graduate students gathered detailed data about things like how close together do people in different occupations work. And with that data, we could predict which occupations wouldn't be able to go to work and therefore which industries would be short of labor and therefore which industries wouldn't be able to produce their goods because of that. And by combining that with information about what businesses would people not go to because they were afraid, we were able to make a good prediction about what the shock would be and then how it would reverberate around the economy over the course of time.

22:43And of course, it should be said that it wasn't just the economy you were trying to protect. People's lives were at stake too. People's lives were at stake. And at that point, the government was trying to decide how to emerge from the lockdown that we'd gone in back in March. And we knew that they were going to announce something. And so we found a policy that we thought was the least bad. We specifically recommended that the government keep open the upstream industries like mining, forestry, manufacturing, but close the downstream customer-facing industries. And that kept most of the economy functioning, but not that many people more would be affected than if everybody stayed home.

23:25So we recommended that policy to the government, and that was the policy they took. Now, how big a role our model played, they didn't tell us. But we did give them that information about a week before they made the decision. How accurate would you say your prediction was in the end? Very accurate. We predicted in the second quarter of 2020 that there would be a 21.5 % hit to UK GDP. When the dust settled, it was 22.1%. Very impressive. Well, Doin, your current focus is on climate change. And last year, you co-founded a company called Macrocosm, where you use your complexity economics models to predict the economic impact of the transition to green energy.

24:08Why did you choose to focus here? Transitioning away from carbon emissions is the biggest problem facing humanity. So the techniques we're using are really well suited to do that. So how do you predict how much a particular energy source like solar or wind will cost in the future and become profitable? For that, we collected data on lots of technologies. So from this data, we built models that just made use of those patterns to make predictions. And we discovered that technology transitions are exponential. And those predictions turned out to be much better than the predictions that the more standard economic models were making.

24:48Quite a lot better. What then did your models predict? Well, in 2010, I predicted that by 2020, solar cells were going to be as cheap as coal-fired electricity. At the time, that was a crazy prediction. The economist in 2014 said that solar energy is the most expensive way to deal with climate change. But it turned out I was right and they were wrong. What do you then predict for the future? We predict in the next decade a rapid transition to solar and wind and batteries and storage. Within 20 years or 25 years, we will have displaced most of fossil fuels, not going to nuclear because nuclear is too expensive.

25:30And that's actually going to save us a lot of money. We calculate around$11 trillion in total globally. And does it matter that the politics of certain countries mean they're going to resist this sort of change? It does matter because it holds things back and we need to get there as fast as we can. But I think that resistance will ultimately be overcome simply because it saves us money. You've applied your research, Doin, in a range of fields over your career. You seem to be showing the experts in these fields that what they're doing is wrong. I don't want to say what they're doing is wrong. I just think we have another way to do it that may do better to do it right.