Arthur Turrell on Economic Data, Modeling, and the Future of Nuclear Energy

Arthur Turrell is the deputy director at the data science campus for the UK Office of National Statistics (ONS). Arthur is also a former researcher at the Bank of England and a nuclear fusion scientist. He joins Macro Musings to talk about his work at the Bank of England, the future of economic data, and his new book on nuclear fusion titled, *The Star Builders: Nuclear Fusion and the Race to Power the Planet*.

Read the full episode transcript:

Note: While transcripts are lightly edited, they are not rigorously proofed for accuracy. If you notice an error, please reach out to [email protected]

David Beckworth: Arthur, welcome to the show.

Arthur Turrell: It's fantastic to be here, David. And I'm really honored that you've asked me.

Beckworth: Well, it's great to have you. And I was thinking back to when we first met, we met at the American Economic Association meetings, the last one they had in person down in San Diego. And you were there with your group, the Bank of England. I went to one of your parties, I believe was a Saturday night. I met you, I met a few other staffers there. So you came out of your way to greet me. So it was nice to start chatting. And then I learned about your fascinating research. You've done research on real-time economic data, and you've also done this research, which I learned after the fact that you were actually a PhD in physics on nuclear fusion.

Beckworth: So this is fascinating stuff. And this is Macro Musings, but a big part of macro is long run economic growth, and that's closely tied to how we have energy. How do we fuel this growth? And you cover this in your book, how are we going to sustain humanity's progress moving forward if we're running out of the traditional fuel sources on this planet? And tell you what, after reading your book about nuclear fusion, I'm very hopeful.  Not to get ahead of ourselves too far here, but it's a source of energy that really we can never effectively run out of. Is that correct? I mean, it's just so much if we get to that point.

Turrell: I think we probably worry about things like the sun running out of fuel if we really truly exhausted all of the fusion fuel in the solar system or continents moving around and things. So it's true that fusion fuel looks effectively limitless right now to humanity, if we can get those advanced forms of it working.

Beckworth: Yes. So fascinating stuff. And we'll come to your book in a bit. But I want to talk to you first about your work at the Bank of England. And before we do that, one other thing, I want to mention your wife, Alice Turrell as well. She's a listener of the show, is that correct?

Turrell: That's right. And in fact that's one of the reasons why when I saw you at San Diego, I felt compelled to go over. I felt that I would never be forgiven if I hadn't gone to speak to you given my wife was such a fan of the show. And actually I took a photo of us both-

Beckworth: That's right.

Turrell: ... and sent it to her. And she was quite annoyed at me for having this nice trip, hearing about all of this great macro at San Diego. But I think that will soften the blow a little bit.

Beckworth: Well, I'm glad I could help in the marriage relationship there, Arthur, back at home. But yeah, so she listens and a number of other staffers they come up to me and listen as well. So to all those at the Bank of England who are listening today, greetings to you. And to all our listeners around the world, it's great to have this audience that can join us week to week. And again, Alice, thank you for listening and getting your husband on this show. Okay. So Arthur, let's talk about your work. And in order to talk about your work, we need to first talk about your amazing career. Because you went from being a PhD studying nuclear fusion to being a researcher at the Bank of England. So walk us through that. How did you make that transition and why did you make that transition?

From Nuclear Fusion to Macroeconomics

Turrell: It's a difficult story to tell in some ways, because it's completely unplanned, and I've just been incredibly fortunate to have the range of experiences that I've had. It's been a surprise to me and it's been a surprise, I think, to everyone else as well that I've ended up where I have. But the context here is that I was working on nuclear fusion at Imperial College. I did my PhD there. I stayed on, did a post-doc for awhile and I did a fellowship for awhile.

Turrell: But at some point during that time, there was some things which were pushing me a little bit away from physics. And there's some things that were starting to attract me into economics. And in terms of the push factors, the UK physics landscape is very, very competitive. And for family reasons actually, because my grandmother was very ill at the time. And I think I was her only blood relative who was around still. I didn't want to move to US, which is a well-trodden path for people working on fusion. There are some great US labs, some great US universities.

Turrell: I had to plan maybe to go to Stanford and work on plasma physics there, plasma physics being a topic very much related to nuclear fusion. But with my grandmother ill and the state of play in UK permanent jobs in physics, so I started casting around. I could see colleagues who had been childhood math geniuses who are writing the codes for the UK Atomic Weapons Establishment, and even they couldn't get permanent jobs in physics.

Beckworth: Wow!

Turrell: So I thought, well, I'm just not smart enough or motivated enough to stay in this game. But at the same time, I was starting to get really interested in economics, because some extraordinary things were going on. I started my PhD in 2009. So when I was finishing it and doing my postdoc and my very short fellowship, the aftermath of the financial crisis was still playing out. And to me, it was extraordinary what was going on, that the whole world could be brought to a halt by the bets that were being made on financial markets, or at least that's how it seemed to me from the outside.

Turrell: And I think we could all agree that we've gotten ourselves into a bad situation. And somehow economics was part of the solution to that puzzle, both of how we'd got there and how we might prevent something similar from ever happening again. And also, it seemed to me that scientists had been banging on about climate change for decades. The world knew what it needed to do. We all agreed that it's a bad thing more or less, certainly in the UK. I know in the US there was a bit more contention about that.

Turrell: But most of the world agreed that the right thing to do was to tackle climate change. And yet somehow we couldn't collectively take action to do it. And again, it seems to me like human behavior or the behavior of institutions and organizations, incentives, those seem to be key to unlocking how we made progress on a global scale. So for all of those reasons and more, I was starting to get interested in economics. And then I found out that the Bank of England, they had decided to start hiring PhDs in subjects outside of economics for the first time to do research.

Turrell: Now, there were PhDs from other disciplines, who've gone in through various channels, but this was an explicit mandate to hire people who had different skills. And in large part driven by Andy Haldane, who has just left being on the Monetary Policy Committee at the Bank of England. And I interviewed, and luckily for me, they thought I had some things to bring from the sciences to contribute to economics. And hopefully, I've helped a little bit with that. But it wasn't just physics. It was neuroscience, it was linguistics, it was mathematics, it was computer science.

Turrell: And one of the criticisms of macroeconomics in the run-up to the financial crisis and actually for a few years afterwards, is that it had been quite insular in the ideas, particularly in macroeconomics, and hadn't really tapped up those subjects enough or not enough recently. Obviously there's a bit of a link with engineering and signal processing. So I think the idea was that we could bring something new, and some diversity of thought, and some different ideas into macroeconomics. And hopefully, we've done a little bit of that.

Beckworth: Yeah. Fascinating story. And your research at the Bank of England says that you did make a nice contribution while you were there. And I want to look at some of that briefly before we get to your book on fusion. One of the things that you covered and wrote about was real time economic data. And let me tell you a little story where I started to appreciate how important this could be. So I had a former student and we happened to get in touch. And he was telling me, he worked for a hedge fund.

Beckworth: This is many years ago. He worked for a hedge fund and he and his colleagues had access to millions and millions of real time credit card and debit card data. And they would take this data and they would use it to inform their investments, they would buy certain stocks and different asset classes based on the day to day trends or changes they saw. I was blown over when I first saw this, and then I started reading around. There's more stories about people using real-time satellite images to engage economic activity.

Beckworth: And then it just became more and more abundant in different areas, Google searches, you've written about job postings and stuff. So I guess my question, let's start off with this before we get into your research. Arthur, do you think this is the future of economic policymaking, real-time data, or will there be a place for standard economic metrics like GDP, which comes out quarterly? That's very backward looking. I mean, where are we headed in terms of economic data?

The Future of Economic Data

Turrell: Well, those are some big questions. Maybe you can come back to me on that, but let me start with why we might be interested in this stuff. I think that's pretty important here. Where are we getting value from? So I think that real-time economic data give us two things, potentially. One is timeliness and the other is high frequency. And as Charlie Bean, he used to be a member of the monetary policy committee, said in his independent review of UK statistics, which I think came out in 2016. The longer a decision maker has to wait for the statistics, the less useful they're likely to be.

Turrell: And I think that's true. And it was certainly true a lot in the bank. In the early days of coronavirus, we were looking at job vacancy data, restaurant bookings, flight numbers, mobility data week by week. Because those quarterly data would have been far too slow. Can you imagine if we'd waited for the fully revised quarterly data to come out before taking decisions. It was already clear that people's behavior was changing because of the risk of this virus, and the bank took a special policy meeting.

The longer a decision maker has to wait for the statistics, the less useful they're likely to be, and I think that's true.

Turrell: So I think in practice, this stuff is useful. Sorry, there is another advantage as well, which is actually granularity. Because if we think about things like survey data, it's just not feasible to scale it up to the size of the whole country. But in some cases, and you mentioned job vacancies, something I've worked on a lot. If we take the data from online job portals, it's actually possible to basically get the universe of job adverts in a country, which is an extraordinary difference. We can drill down right into the details there of what that demand looks like.

Turrell: So I think that these have a lot of value to add. Your question is, is this the future? I think it's part of the future, but I don't think it's the only part of the future. And I very much see these real time indicators as complements rather than substitutes for the traditional economic indicators that we might get. And they've got faults and they've got flaws and they've got pros and cons relative to the other stuff. So I think Chris Giles in the FT said that they're like fast food, initially satisfying, but not that great for you.

Turrell: But I like to think of them more as a healthy snack between meals. So you've got this big meal, GDP. But maybe in between that, you want to find out a little bit about what's going on. You want to get a bit of a signal, but you've got to be careful about what you're mentioning. It can't all be sweets. You've got to take in the good stuff. So I think we've got to be really clear about the floors that they might have. They can be a bit untested. So like the Google mobility data that came out over coronavirus when it was first released. We really didn't know much about how it related to economic activity, because we didn't have a big background to say, this is how it correlates with GDP. And maybe coronavirus would have changed that anyway. So we've got to be a bit careful. It can be a bit untested. We've got to be honest that these are only approximate indicators of the underlying phenomenon or series we're interested in.

Turrell: So they might not correlate perfectly with official statistics. They may not have full coverage. There may be self-selection, they're likely to be biased. And if I simplify that a great deal, there's a bit of a tradeoff between what you'd really like to measure and what the faster indicators can tell you. But as you say, faster indicators also pass the market test. We know hedge funds have been using them for from years. And I was actually thinking about this the other day, and the earliest evidence I was able to find with a quick search was that firms were doing text analysis of filings back in 2003.

Turrell: And it's become so ingrained now. There's a great working paper by Sean Cao at Georgia State University, which shows that firms are actually changing the words in their filing to try and trick… influence is a better word, what the automated text analysis of their filings takes away from them. So yeah, this is not a new game. And in a way, I feel we have a duty to explore every possible source of information on the economy. I had lots of stories about people doing this. Morgan Stanley using air pollutions to work out industrial production in China. And I even heard a story that merchants in ancient Babylon measured the depth and flow of the Euphrates in order get ahead with the future supply of commodities. So this is not a new idea.

Beckworth: Interesting.

Turrell: But I think we can take it to new places into the next level. And maybe we'll come onto this, but I would make a bit of a distinction between some things like Google searches and some things like the job vacancy data for as well. So not all of this alternative data is created equal.

Beckworth: Wow, Arthur. So ancient Mesopotamia was using real time data as well, who knew? Fantastic. So let me ask this question. So I'm going to really think big and ambitious here. So you remember Mark Andreessen’s article, *Why Software Is Eating The World*?  So I'm imagining taking this real-time data and just take it to the next level, we have artificial intelligence, smart machines. And Jay Powell goes into his office, turns on his smart machine, and it's going through these billions and billions of data points looking for underlying relationships. So even if the relationships between certain measures change over time, the smart machine understands it all. Will we ever get to the point where we'll have something like that? Real time data, smart computing coming together, making the life of a policymaker just very easy.

Turrell: That's a very interesting idea. And at the bank, we used to joke about the Star Trek dashboard of real time indicators, where the maker could look out and see what was happening in real time. So you mentioned AI and things as well. I think that's a slightly separate thing. So I think even if we could get to that first level of having this amazing real-time data that could tell us about economic activity in a way that we really cared about and at a level of granularity that was helpful, that would be a huge leap forward for macro and for all kinds of decision-making.

I think even if we could get to that first level of having this amazing real-time data that could tell us about economic activity in a way that we really cared about and at a level of granularity that was helpful, that would be a huge leap forward for macro and for all kinds of decision-making.

Turrell: And it comes back to what Charlie Bean said, “The longer a decision maker has to wait, the less useful information is.” Now, you mentioned about AI as well. And that's the bit I'm less convinced by. But never say never, because the improvements in AI and machine learning have really been astounding. And I'm not going to be foolish enough to predict the future. But the reason why I'm perhaps a little bit less optimistic on that front is that machine learning is mostly adapted seeing the relationships that have already happened, so the relationships in the past. So that's a problem. And I think that there are ways that we can use machine learning better in some developments in the economy, so modeling, some prediction things where we understand the economic structure. But yeah, I think just guessing that real time data together would be a big leap forward.

Beckworth: Indeed. So you are probably aware of China talking about the introduction of its central bank digital currency. And there are reasons to be uncomfortable about it. I think we both [would] be concerned about the privacy issues. They're using it to spy on their people. But tying this into our conversation, the flip side of that is, they're going to have amazing real-time data. They're going to know spending day by day, hour by hour. If everyone starts using a central bank digital currency in China and the Chinese are going to monitor it like they say they are, I mean, talk about real-time feedback. I mean, this would be an amazing feed. Again, it comes at a big, big cost and one that I'm probably not willing to take. But any thoughts on that, any thoughts on where that could lead?

Turrell: So I think that some types of real time data there is a trade-off with privacy and with what I might broadly call liberal values. And especially those of us who deal with data, we have to be incredibly careful that we respect people's privacy when we work with sensitive data of any kind, financial transactions or anything. And as you know, financial transactions have been used to monitor what's going on in the economy. And I think as people working on that data, we should and do and invite scrutiny, actually, from ethical committees that are independent to make sure that we're using the data in the right way. And like you, I would feel nervous about a system that knew everything about me. And also there's no need to do that. As a macro economist, what you really care about is consumption of groups, not of individuals. That's not important, that's not macro-economically important. So there's no good argument for a level of granularity that compromises people's privacy.

Beckworth: Okay. Arthur, so you have been working with real-time data. So we've been talking about it, but you've been working with it. So share with us a few examples from your own research, how you've used this and how it's helped inform the Bank of England.

How Real-Time Data Has Helped Policymakers

Turrell: Yeah, sure. And it's not just the Bank of England. I've also been heavily involved since I moved to the ONS on real time indicators that we make available to the UK Treasury and to the Bank of England to use for economic analysis. So the work that I've done personally and most of this was at the bank, was looking at two types of alternative data, if you like. So one of them was actually newspaper text, which might sound like a strange place to go and get data. But actually, if you start with the proposition that if people care enough about a topic enough to write about it, and people care enough about a topic to pay to read about it, then there's some useful information in that.

Turrell: And the idea we had was just, look, if we take this kind of information stream of text and try and turn it into numbers somehow using, either very simple models, like ordinary least squares or the more fancy models from machine learning, can we better forecast what's going on with macroeconomic variables? And that's a really interesting question, and we didn't really know the answer. But not just us, many other papers now as well have found that there does seem to be some information in the news.

Turrell: And I'm not sure which way the causality is going, whether journalists write about the news and that influences people, or whether journalists saw, if you like, a real time indicator in themselves. They're picking up on the things that are being felt or experienced by people before those things happen. Either way, it seems like the news can give us an early read on what might happen, based on the evidence that we found and to what various other people have done. There's been some great work from Norges Bank looking at some of the questions. So that's one type of real-time indicator that I've had personal experience with.

Turrell: And I think there's a lot more one could do on that. We've only scratched the surface. And in terms of the other one that I've spent quite a bit of time working on, that would be data and job vacancies. So early on in my time at the bank, I went around and started asking people, looking for interesting things to work on. And I went around and said, "What are we missing? What don't we have good information on?" And people kept saying, "We don't know enough about labor demand." And so in UK we have a very good survey of job vacancies, national survey, and it has a sectorial breakdown. But it doesn't have an occupational breakdown. And of course, the job of a rocket scientist and a dentist are quite different. And you imagined that it might be difficult for one of those to become the other overnight.

Turrell: So labor is not this homogeneous blob. It's very heterogeneous. So we thought by going out to the online job portals where firms post their jobs and actually express their demand for labor, and taking that data and analyzing it, we might be able to get a better grip on things like labor market tightness across the country. And one of the reasons why I was so excited about this and why I think it's such a worthwhile exercise, is that these data are generated in the course of economic activity.

Labor is not this homogeneous blob. It's very heterogeneous. So we thought by going out to the online job portals where firms post their jobs and actually express their demand for labor, and taking that data and analyzing it, we might be able to get a better grip on things like labor market tightness across the country.

Turrell: When firms post these, it costs them some effort. It cost them some money, even if not explicitly. And so they have a bit of skin in the game, that they want to get this advert out. And I think I'm right in saying in the UK it's illegal to put out job adverts that you have no intention of hiring for. So, these are a genuine measure of demand, and that's why I prefer this kind of data perhaps to things like Google searches, which the marginal cost of an extra Google search is basically zero for a user. So you can see that these relationships might change a bit more at a time, whereas at the job vacancy data that is posted on portals is a really clean measure of labor demand.

Beckworth: Okay. Very fascinating. Now this is contributed to our understanding of macroeconomics, better data. But you've also researched in another area that has contributed to our understanding of macroeconomics, and that is modeling. So you alluded to this earlier, maybe some of our models weren't up to speed come into 2008. And I'm sure there's macroeconomists out there that’ll push back against that claim. But you've been working on heterogeneous agent models, and they are the rave right now. There's different forms of them. There's the HANK model, which I see a lot of, the heterogeneous agent new Keynesian model. But I believe you work more with the heterogeneous agent modeling that looks at them in as in a simulation setting. So maybe you could walk us through the differences and, and what are the insights you're able to glean from that exercise?

Heterogeneous Agent Modeling

Turrell: Yeah, sure. So I think you've got to bear in mind that I was coming to macroeconomics with a completely different perspective by design, as I've said. Thanks to the great initiative of people at the bank like Andy Haldane, bringing in people with different backgrounds. So when I joined the bank, I felt that macroeconomists were a bit skeptical of the idea that representative agent models might be missing something. And it's only down to the great work by Ben Moll and others in really pushing the idea that heterogeneity matters with models like HANK. They've shown that it leads to new channels or different channels dominating, and that it's important. But when I arrived, that view was a lot less popular. And frankly, when I started looking into the assumptions of these models, they seemed quite heroic.

Turrell: They were a little bit surprising for someone who'd come from science. And I think partly that was because they are. But partly it's because what I didn't appreciate initially, was that people weren't really trying to capture reality. They're much more interested in a formal way to get flight to their thoughts. And that's a really legitimate way to use modeling, but only if you want to create models that reflect your strong priors and beliefs about how the behaviors are going to come out. So you build some of those channels in from the start. And what isn't going to happen is that those models aren't going to surprise you with phenomena that you never expected to see, because the assumptions are built in from the start.

Turrell: So I think that modeling is really useful and that model exists in science. But I also think that models that can tell us about what we can't anticipate, what we haven't baked in from the start are really useful. So models that we can throw in a punch of interacting agents, and we just tell the agent how to behave. And then we sit back and say this is too complicated for any human to solve. It's too complicated to solve mathematically, let's just throw it in a computer and let the computer solve it. And nature made the world incredibly complex. No one is smart enough to think through how all of those eventualities will fit together. But that model can do that work for us. And that's the kind of model that I've spent more time thinking about while I've been in macroeconomics. So agent based models. And the difference is that we have individual agents, discrete units and we give them some rules and then we just let them go.

I think that modeling is really useful and that model exists in science. But I also think that models that can tell us about what we can't anticipate, what we haven't baked in from the start are really useful.

Turrell: And this is philosophy famously in science, the paper in nature by Phillip Anderson that says more is different. So having one agent on their own behaving looks very different than if we have a group of agents, even with the same rules all interacting. They’re different in that sense. And the other thing is that, once you do that, and you're letting the computer do a lot of the work instead of semi analytically solving these models, what you end up with is simulations that are as rich and as complex almost as an actual experiment or real-world data. And I think this is the thing that puts a lot of mainstream macroeconomists off agent-based models, because it's just very complicated. What's going on? And I like to think about it as an experiment in a computer. And just like in a real-world experiment, you have to then unpick what happened.

Turrell: And maybe you see some interesting phenomenon, but then you have to work backwards. And this is a very standard approach in the sciences. You think about voids in computer science, you think about models of the spread of infectious diseases and epidemiology, like the famous Imperial Model in the UK that Professor Neil Ferguson used to recommend lockdowns, or the models of the [inaudible] that I used in physics. Those are all those kinds of models where they're individual based models, agent-based models and you throw things in. So they're quite different from the HANK models, but they do recover some of the same benefits in the sense that we can have lots of models with a completely different distribution and just see how that plays out.

Beckworth: That's interesting, the way you compare that to a randomized controlled experiment in a natural science setting. In fact, this is like the holy grail from macro. We can't rerun Earth's history a thousand times with a different shock each time to see how things... There's human suffering and it's just not ethical. But what you're saying is, look, we do have the ability to do a randomized controlled experiment via computer simulation. But you alluded to this and it's been my impression as well, is that this hasn't been widely adopted or not everyone's on board. But what is your impression? Is it gaining more acceptability, more credence in the profession, or is it still an uphill battle?

Turrell: I can talk about why I think it's been an uphill battle, actually. And I think part of it is that when you're moving to an agent based model, it's a very different setup, and it's actually much harder to do rational expectations. And also I think the tribe of people who are working on agent based models tend to be contrarians anyway. So what tends to happen is they say, "We'd like to get close to reality. So not only are we going to throw out the representative agent with rational expectations, we're going to throw out the general equilibrium, we're going throw out the rational expectations part. We're going to throw it all out."

Turrell: And so, they are quite a long way away from the RANK DSGE model. And the RANK DSGE model, people are looking at the agent-based model and say, "That doesn't look like anything we recognize." In the agent-based model people are saying, "Wow! Rational expectations, that's mad." The evidence in the real world just doesn't support that as how people make decisions. I can't believe you'd include that. And I'm mis-characterizing the strength of position. But I think one of the problems has been that we don't have models that have plunked themselves right in the middle of that.

Turrell: But HANK is a wonderful, wonderful step towards that middle ground, because HANK shows with a continuum distribution how important heterogeneity is. And I think mainstream macro is really sold on that point about heterogeneity being important now. And agent-based models, you can imagine an agent-based model, that all it does is it has general equilibrium, it has rational expectations. And the only extra thing it does is go from a continuum representation of a distribution in productivity or assets or whatever you like, and just turns into a discreet one, whether individual agents, and they're still following these similar rules. And there's no reason why you can't do that.

Beckworth: All right. So Arthur, you've described to us agent-based modeling. And you have a paper where you plant that flag in the middle. Talk about that paper and what you learned from it.

Turrell: Yeah. So in that paper, we were saying, let's keep rational expectations, let's keep general equilibrium, but let's switch to using those atomistic individuals. And the problem is that it's actually quite hard to solve these models that progress in time with rational expectations. So what we decided to do was go to the, frankly, astounding developments in machine learning that have been made and use that to help agents solve for the rational expectations solution to these models. Thereby we're getting much closer to mainstream macro, but we're still able to introduce much more richness in the heterogeneity. And to give you a sense of how powerful this reinforcement learning is, have you ever played the game Capture the Flag?

Beckworth: Yes.

Turrell: Okay. So two teams, each trying to steal a flag from each other, bring it back to their own base without getting tagged. It's a complex multi-agent game. Well, DeepMind and Google have a paper where they showed that they can produce players, artificial intelligence players of that game that do just as well as humans, even when there's a mixture of humans and machine learning agents on the team. That's a complex environment that requires teamwork. So we decided, well, we'll have that. We'll pinch that technology. We'll let our agents in our model use that to solve the best decisions.

Turrell: And actually the impetus for this model, which we tried to do this in, was quite interesting. It was actually an email from Andy Haldane that I got about 6:00 in the morning, one Saturday, early on in the coronavirus crisis saying, how can we model the economic and health parts of what's going on together? How can we combine these things? And because agent-based models were already being used in epidemiology to do lots of the modeling, I thought, well, let's try and see if we can put both of these things into an agent-based model. And in that paper, we do indeed include some of the way that the infection spreads through individual agents, alongside the macroeconomic effects. So hopefully we've planted that flag, although it’s early days.

Beckworth: And the title of that paper is *Solving Heterogeneous General Equilibrium Economic Models with Deep Reinforcement Learning.* Is that correct?

Turrell: That's right. Yeah.

Beckworth: Okay. We'll provide a link to that as well on our show notes. So again, I encourage listeners to check this out and I will also provide a link on our show notes to Arthur's page titled *Coding for Economists*, and there you can find more resources regarding agent-based modeling as well as other applications, like scrap scraping data, textual analysis, and such. All right, let's move on to your book. That's the big reason we brought you on. You have a new book coming out, and it's a book on a topic that I was not familiar with, but I think it's very important. Again, the title of the book is *The Star Builders: Nuclear Fusion and the Race to Power the Planet.*

Beckworth: And it's got a great cover, at least the US version. I like the cover. It looks like it's the sun, but instead of the sun, you got a nuclear reaction symbol there going on in the middle. It's been an amazing read. And again, this is consequential to us. It's not just a discussion about the physics of fusion. This is a discussion about how do we sustain this planet over the long run. We've got climate change. And even in the absence of climate change, there's a limit to some of the existing fuel sources we have. And you highlight this in the book. So fossil fuel sources are running low. At some point they'll be depleted. So we need some alternative energy source that can sustain us over a long, long time.

Beckworth: So, Arthur, let's begin maybe by defining some terms and things... Honestly, I'll confess. I did not know the distinction between fission and fusion, and maybe many of our listeners don't either. But it really helped me understand the difference between a thermo nuclear hydrogen bomb and an atomic bomb, as well as... Of course, that's not what we're getting at in this book. But tell us what is the difference between fission and fusion, and why we are focused on the fusion side of it?

Fission vs Fusion: Energy Prospects for the Future

Turrell: Yeah. Great question. The simple answer is that fission is taking big unstable atoms and splitting them apart to release energy, and fusion is taking small atoms and smashing them together to make bigger atoms to release energy. And they're both nuclear reactions, and they both happen in nature somewhere, and they're both potentially sources of energy for humanity as well. So we can use fission for many, many years, and we haven't yet cracked fusion.

Beckworth: So you provided a nice... If you can think of a two-by-two matrix, and you have controlled and uncontrolled... So let's talk through these two different types of nuclear reactions. You got fission, you got fusion. So fission, you're breaking the nuclear part. With fusion, you're building, making it bigger. In both cases, they release energy. So spell it out for people like me out there who didn't know any better. So with fission, you have controlled fission and you have uncontrolled fission, and I'm going to describe what I think I understand, but you can tell me if I'm wrong or right. So controlled fission is like a nuclear reactor, a power plant that lights our homes and lets the city run. Uncontrolled fission would be an atomic bomb. So is that the correct classification there?

Turrell: That's how I'd describe it. Yes. Of course, people design atomic bombs. They're meant to do that. But in some sense, the idea is just to make it go as fast and as big as possible. That's what the designers wanted to achieve. So in that sense, it's uncontrolled. Whereas, of course, when we're doing this stuff for energy, we want to keep it as controlled as possible. And actually the technology and the way that these things work is very, very different. And that's why I make such a distinction between them.

Beckworth: Okay. So again, you got controlled and uncontrolled. So with fission, what probably most people think about, controlled fission is like a nuclear reactor power plant. Uncontrolled is, you're letting it rip, the atomic bomb goes off. Now we take those same two labels and apply them to fusion. So let's start with the one that's easy. Uncontrolled fusion is the hydrogen bomb, correct?

Turrell: That's right.

Beckworth: And so the holy grail, the missing piece is the controlled fusion, this energy source that we can tap into, and it's effectively the energy we get from the sun, right? That's what you're trying to replicate, hence the title of the book.

Turrell: I was just going to say that, so in a hydrogen bomb, there are both fission and fusion reactions. And three of this quartet were demonstrated actually within a relatively short period in the '40s and '50s. So we've done all of that, and we've known how to do it for years and years. The bit that we don't know how to do is control the fusion. And that's a bit of a surprise when you look around the universe and you see every day when you go outside, daylight is provided by nuclear fusion reactions.

Turrell: And if you go outside at night, those thousands of pinpricks of light that you see in the sky and the many, many billions more than that are out there that you can't see, that’s all light from fusion as well. And fusion reactions happened at the start of the universe. And most of what we're made out of in terms of atoms comes from fusion reactions. So in many ways, it's the university's most ubiquitous energy source. And it's a bit of an embarrassment, frankly, for humanity that we haven't it working on Earth yet. Although, there's been some interesting progress to the end.

Most of what we're made out of in terms of atoms comes from fusion reactions. So in many ways, it's the university's most ubiquitous energy source. And it's a bit of an embarrassment, frankly, for humanity that we haven't it working on Earth yet. Although, there's been some interesting progress to the end.

Beckworth: Okay. So why is it that it's so easy for stars to generate this energy, but it's so hard for us to figure out? And again, to reiterate the point you just said, the other three categories controlled and uncontrolled fission, and then uncontrolled fusion, those three things we've done. We've mastered them back in the '50s. And here we are, 60, 70 years later and we're still trying to make it work. Now, your book paints a very optimistic case that it was just around the corner. But why has it been so hard compared to the sun that does such a great job with it?

Turrell: It's hard because it needs some of the most extreme conditions that humans have ever tried to create and control. To do it on earth requires temperatures over 100 million degrees Celsius. I think that's about 180 million degrees Fahrenheit, but don't quote me on that. Familiarity with…

Beckworth: Very, very hot.

Turrell: Very, very hot indeed. Hotter than the center of the sun, is the optimal temperatures to it on earth. So it requires extreme temperatures, it requires extreme pressures and densities like found in the sun. And the reason why the sun is so good at it is that the sun is just absolutely enormous. And so it can use gravity to create those very, very dense conditions. And the collapse of a star matter generates the heat right in the center, and it gets hot enough for fusion reactions to happen when stars are born. And then the fact that there's this huge, huge gravitational well and lots and lots of matter around traps the released fusion energy, which comes out in the form of fast particles of light inside the sun. Whereas on earth, if we think about the scale of a fusion reactor on earth, it might be, I don't know, six meters across. So when light is created from fusion reactions, that light escapes straight away. Some of the fast particles that come out of the fusion that we tried on earth can come out straight away.

Turrell: And we can't just have an earthbound fusion reactor just floating around in space, because that's not where we are. We have to put it in some container. But no container out of any material known to humanity could withstand that temperature without the energy of the plasma, this is the type of fuel that we put into a fusion reactor, just dissipating its energy and the reaction just stopping straight away. So fusion is really hard to do on earth. That said, some of the most ingenious people on the planet are working on ways to do this. And we've developed some amazing experiments and apparatus that have got us really pretty close to demonstrating in principle that we can generate energy from a plasma.

We've developed some amazing experiments and apparatus that have got us really pretty close to demonstrating in principle that we can generate energy from a plasma.

Beckworth: So you have a number of interesting people in places you go to in the book, and we'll mention some of them. But you said the big struggle is not to generate this form of energy. It's being generated, but to do it in a net energy gain perspective. In other words, you want to be able to pull out more energy than you're putting into create it. And so far, that big hurdle hasn't been met. So you call it break even, or self-perpetuating ignition. Get to the point where you put a little bit of energy in an out comes a lot more. And you argue we're close to that. So there's just a funny, humorous saying in this industry that has been working on this, that fusion is 30 years away and always will be. So 30 years from now, it will be another 30 years. But you argue that saying we can put that to the side now, because we are truly close. We're on the cusp of this net energy gain achievement.

Turrell: Well, it's a great joke. And it's one that always comes up in the context of fusion, so much so that The Economist editors have banned people from using it in any article about fusion. So it's a long refrain, and it's a genuine question. Why can I say taking so long, and are we actually close? But we've got to be careful about what we mean by fusion. So do we mean net energy gain, which is really the scientific principle, or do we mean commercialization and people actually being able to plug their laptops and TVs into the grid and get fusion energy.

Turrell: So the second part I think is further away, and there's lots of challenges to go there. The first part, net energy gain, people have made big progress on. So, net energy gain is 100% of the energy that we put in coming out. And it's like at the moment we've got a match and we're trying to light a fire. But at the moment, we're just getting a little bit of flame. We're not getting as much energy out. But net energy gain is where the fire starts and maybe you get as much flame as you did putting the match. And it goes beyond that as well.

Turrell: Well, actually in experiments today, we've got to 67% of net energy gain or net power gain, I should say, using magnets to control the plasma, and to about 3% using lasers to ignite and control the plasma. Doesn’t sound like they are that close to 100%, but this is physics, and it's plasma physics. So what we care about here are orders of magnitude. And I forget the number exactly, but I think it must be factors of 60 or so that experiments on laser fusion have improved just since 2011 to get to that 3%.

Turrell: The factors by which the experiments improve are factors of three or six or 10. It's not going from 3% to 5%, it's going from 3% to 30%. So that's why I'm optimistic, because we have seen progress. And also because money has been flooding into fusion as well. Governments have always spent some money on fusion and people have long argued that it's not really enough to see it happen. And like with any technology, it's responsive to the level of investment. So if you want to make this happen, then you have to give it the resources to make it happen. And for a long time you have just not had enough.

Governments have always spent some money on fusion and people have long argued that it's not really enough to see it happen. And like with any technology, it's responsive to the level of investment. So if you want to make this happen, then you have to give it the resources to make it happen. And for a long time you have just not had enough.

Turrell: But what's really interesting about the last few years and why people are so excited about fusion now, is the private sector has been getting involved. There are 20 or so fusion firms around the world. A lot of them are building experimental reactors. I think there are about a hundred of fusion experiments in construction or built around the world at the moment. People put an estimate of about $2 billion footed infusion over the past five or six years. So the fact that the private sector is starting to show interest, suggests some people think this could happen, certainly net energy gain.

Turrell: And a lot of those private firms are saying, "We're going to show net energy gain within a few years." Not within 30 years, definitely within 15. I also think that there's good scientific reasons to believe that net energy gain is attainable. There's a great theory paper that shows that if you can get the right combination of environmental variables in your plasma, then you can get to net energy gain. There's no physical reason why you can't get there. And obviously that helps people working on fusion stay optimistic over the years, but the exciting thing is we're really close to those conditions now.

Beckworth: Yeah. So, Arthur, in reading your book, I was struck by how much private industry has stepped up to the plate. And my impression was this could be the pivotal turning moment, the inflection point where we do see the achievement of that 100% threshold that's needed to get a net energy gain. And it reminded me of the space industry. Jeff Bezos and Elon Musk and all the advances we've made with rocket technology, but it took them getting involved. They had to step forward, they had skin in the game, they had commercial profit motives. I mean, I'm a big fan of capitalism for this very reason. We may need capitalism to get this on a working scale, working level. So do you think that is the key catalyst here that's going to push this over and make it successful, is getting private industry involved?

Turrell: I think that private industry can only help fusion. I'd be interested to see whether a private firm gets there before one of the national labs, because the national labs have been doing this for a long time. And at the moment, all of the machines that are close to net energy gain are government labs.

If you can get the right combination of environmental variables in your plasma, then you can get to net energy gain. There's no physical reason why you can't get there. And obviously that helps people working on fusion stay optimistic over the years, but the exciting thing is we're really close to those conditions now.

Beckworth: Okay.

Turrell: But that's not to say that a private firm couldn't catch up. But what I would say is that commercialization, rolling this out on a kind of global scale and providing energy down the line, that is something that is really hard to see happening without the private sector involvement. And there are ton of extra challenges to solve that. But it's amazing how once the proof of principle happens, things can just completely change. And I think we forget this when we think about these... What seemed like wild ideas or difficult to achieve feats.

Turrell: And I always think about the New York Times in 1903 reporting that man won't fly for a million years, to build a flying machine would require one to 10 million years of effort from mechanics and engineers. And I think that was nine days before the Wright Brothers first flight. And now we take it for granted. But it's an extraordinary thing, these huge hunks of metal flying through the sky. And there's countless other examples… maybe using MRNA to fight disease. Grant funders wouldn't believe this. It was a small number of people and a passion for science and the idea it could work pursuing it. And now look at where it got us. It's helping save the world, and that's not really hyperbole.

Beckworth: Okay. So let's move on from the science of it and the cool factor. And it really is a cool factor. This is going to be amazing when this does come online and we can tap into it. Let's move into the discussion of why it's really needed. So tell us about the alternative energy sources we have, why we really have to find an option and why fusion is the option.

The Argument for a Fusion Future

Turrell: Most of the energy that the world gets today is from fossil fuels. And that situation in the long run is just simply untenable. And there are very good reasons for that. The first is that fossil fuels create a load of air pollution, actually, and have lots of negative health consequences. And if you've been downwind from even a small generator, you'll know what I mean. So we need to get off them for that very good reason. We need to get off them for the reason that they're simply going to run out.

Turrell: I think the engineer and information theoretician David [inaudible] said something like, the world was lucky to start the industrial revolution with billions years worth of accumulated fossil fuels. And we've been burning through that. And if we just rely on fossil fuels, we're going to run out. But the biggest and most immediate threat is actually climate change. By pumping CO2 into the atmosphere, we are changing the planet and the consequences of that are pretty stark.

Turrell: The World Health Organization estimates that it is already contributing to the deaths of 150,000 people per year. So we need to do something about that, and we need to do something about it pretty quick. And because so much of our current energy as a planet comes from fossil fuels, the scale of the change we need is absolutely extraordinary. And we're going to need everything. We're going to need renewables. We're going to need nuclear fusion. And if we can get nuclear fusion working in time to contribute, then we’ll happily make use of that and plug that in, too.

Turrell: And I think it's a bit naive to think we won't need more energy in the future as the population grows. And as we exploit new technologies as well. And it's naive to think that we won't need to provide everyone on the planet with the similar kind of energy that people in Europe are lucky enough to enjoy per capita today. So for all of those reasons, we've got to scale up and we've got to do it quick. And mature technologies, immature technologies, technologies that might seem a little wild, but could pay off big time, are all worth exploring.

The scale of the change we need is absolutely extraordinary. And we're going to need everything. We're going to need renewables. We're going to need nuclear fusion. And if we can get nuclear fusion working in time to contribute, then we’ll happily make use of that and plug that in, too.

Beckworth: So tell us why fusion, though, is the best among these other alternatives to fossil fuel. Why not do geothermal, why not do more nuclear fission, the traditional nuclear reactor? Why is it that fusion, do you think, at least over the long run should be the number one source of energy?

Turrell: Well, I wouldn't say that. I think that’s actually the choice for how people want to get their energy. It's not for me to say what people would prioritize. But what I would say is fusion has some amazing pros as an energy source. So one of them is that it doesn't produce any greenhouse gases. So, yeah, that's an immediate solution to our problems in that respect. And if we compare it to something like nuclear fission, fission produces long-term radioactive waste.

Turrell: Now I think that long-term radioactive waste is much less of a problem than climate change is. So I'm someone who's relatively relaxed about that, because I think climate change is the bigger problem right now. But if we could have an energy source that could produce like fission does, on the same scale as fission, with the same very small physical footprint as fission that wouldn’t produce long-term radioactive waste, well, that would be quite amazing. That would be a very good thing to have.

Turrell: And fusion does produce radioactive waste, but it's very short term. And after about 100 years or fewer, the level of radiation is safe to handle. And it’s only the chamber that gets activated as well rather than being a waste product. So from that point of view, it's very promising. As I've mentioned, it has quite a small footprint. So renewables are fantastic in the sense that solar and wind are already two of the cheapest forms of power.

Turrell: Now, fantastic in the sense that they don't produce any greenhouse gases, at least not while they're operating. Obviously, the building and construction of all of these things produces some. So renewables are absolutely going to be a part of the solution, they're going to be a really big part of the solution. But one of the downsides of them is that they have very low energy density. So we have to build enormous plants for renewables if we want to get the same amount of energy as just one nuclear fission plant or hopefully nuclear fusion plant would do.

Turrell: People object to that for all kinds of reasons, because of displacement of other things you could use the land for, because of how it looks, and because people don't want to necessarily live next door to big industrial plants like that. And Princeton University has a really interesting report called *Net Zero America*, which saw renewable energy generation, even if it was only just a big part, not the whole of US energy generation for net zero in the future, swelling up a land equivalent to about eight different states. So when I talk about scale, I really do mean that. This is on a massive scale. And we can do that, it's just a very big thing to do. And it will change the face of the countryside. But absolutely we're going to have that, but we're going to have a load of other stuff as well, and fusion could be one of those alternatives.

Beckworth: Okay. So what is needed to get us there in the time we have left today? What should we be doing? You have a chapter at the end of your book that says, "Can We Afford Not to Do Fusion?" So what should be our next steps, Arthur, moving forward? What would you recommend policymakers do to make this come to fruition?

How to Make Fusion Energy a Reality

Turrell: So if I were given the keys to the kingdom, so to speak, on fusion policy across the whole world, I would certainly support pursuing a couple of different lines of inquiry. So I think that for a long time that the investment in using magnets to do your fusion research has relied on a certain shape, like a donut of reactor. But there's an exciting, different way of doing it, which uses something that looks more like a cored apple.

Turrell: And it basically brings the fusion fueling much closer to the magnetic fields. So I would definitely support those. And actually some of the private firms are working on those. The UK government has a plan to build one of those. So people are doing that, and I think that's definitely worth looking into. But I think for laser fusion, which is the other approach, the advances that they've been able to demonstrate with the laser at the National Ignition Facility, part of Lawrence Livermore National Laboratory in California, have been quite amazing.

Turrell: And based on what we know from secret underground experiments done in the 1980s, and from pen and paper physics, and from simulations, you don't have to go from a laser the size that they have already to one that's very much bigger to get to net energy gain. So for the cost of whatever it would be, a one-off investment to extend the size of the laser there, I would probably go for that as the cheapest way to get to net energy gain. Of course, net energy gain is just the first step. Once we show that we can get more energy out than in, everyone working on fusion has to then show that this can be a bonafide power source. But I think once people have seen the principle, just like the Wright Brothers, showing that first flight, everyone can see the route to making this happen in a way that's actually useful to people.

Beckworth: Yes. And our long run economic growth depends on it. So we need to take this seriously. Well, with that, our time is up. Our guest today has been Arthur Turrell. His book is *The Star Builders: Nuclear Fusion, and The Race to Power the Planet.* Arthur, thank you for coming on the show today.

Turrell: Thanks so much, David. It's been great to talk to you.

Photo by Clement Mahoudeau via Getty Images