It’s easy to get excited about new breakthroughs, but the real power lies in diffusing technological advances throughout the entire economy. Rob and Jackie sat down with Jeffrey Ding, Assistant Professor of Political Science at George Washington University, to discuss how technological revolutions influence competition and the implications for the United States and China.

Mentioned

• Jeffrey Ding. Technology and the Rise of Great Powers: How Diffusion Shapes Economic Competition, (Princeton University Press, 2024).
• Robert D. Atkinson. The Past and Future of America’s Economy: Long Waves of Innovation that Power Cycles of Growth, (Edward Elgar Publishing, 2005).
• Victor Appleton. Tom Swift Sr. Series, (Stratemeyer Syndicate, published between 1910 and 1941).

Auto-Transcript

Rob Atkinson: Welcome to Innovation Files. I'm Rob Atkinson, founder and president of the Information Technology and Innovation Foundation.

Jackie Whisman: And I'm Jackie Whisman. I head development at ITIF, which I'm proud to say is the world's top ranked think tank for science and technology policy.

Rob Atkinson: This podcast is about the kinds of issues we cover at ITIF from the broad economics of innovation to specific policy and regulatory questions about new technologies. So if you're into this stuff, please be sure to subscribe and rate us because it really does help us spread the word.

Jackie Whisman: Today we're talking to Jeffrey Ding, an assistant professor of political science at George Washington University. He researches great power competition and cooperation in emerging technologies, the political economy of innovation and China's scientific and technological capabilities. His book, Technology and the Rise of Great Powers was released this year by Princeton University Press, and that's what he's here to talk about today. Welcome.

Jeffrey Ding: Thanks, Jackie. Great to be here with you guys.

Jackie Whisman: I guess the title gives it away and it's a very ITIF- friendly title. So maybe we'll start and you can let us know the key message of the book and some of the most important things you found while you were writing it.

Jeffrey Ding: Yeah, the book asks a fundamental question, which is, how do technological revolutions affect the rise and fall of great powers? And the stakes of this question are obvious to us today as we speculate about whether China or the U.S. will take leadership in new emerging technologies like artificial intelligence. So to answer this question, I went back to three past industrial revolutions and examined how the technological advances in those revolutions shaped which country was able to gain economic leadership. And the country that's able to sustain economic growth at higher levels than its rivals for long periods of time eventually translates that into military and geopolitical influence.

So that's the mechanism of the rise and fall of great powers. So the message of this book is the key technologies in the past were general purpose technologies like electricity, like the steam engine, like broad processes of mechanization that only made an impact after multiple decades. They spread really slowly and the country that was able to take advantage of these general purpose technologies to adopt them across their entire economy, that was the great power that was able to sustain its rise. So that is the main message of technology and the rise of great powers.

Rob Atkinson: It's interesting because this is exactly what I've been focused on almost my entire career. My first book back in the '90s was a book author Edward Elgar Press called The Past and Future of America's Economy: Long Waves of Innovation that Power Cycles of Growth. And it was really looking at sort Schumpeterian long-wave theory with basically cyclical rises of GPTs. As you know, most for our listeners, that theory suggests that innovation isn't just this linear process, it's more of a wave process. You get these big clumps of what are called GPTs, they then diffuse through the economy. Carlotta Perez at Sussex has done a lot of the good work on that in terms of diffusion. But anyway, I just wanted to say really this is such an important issue and it's one, frankly, that I don't think most economists really think about very much. They're much more in this conventional model of, as Robert Solow once said technology innovation is manna from heaven. It just happens. So anyway, really like the whole focus, and I think it's super important for people to understand this and read your book.

Jeffrey Ding: Yeah. Thanks, Rob. I think a lot of it is building on some of the work done by long-wave scholars I think the key insight from those works is that technology is not continuous. Technological change is not continuous over time and space. It sometimes has the tendency to cluster at certain moments of time where you have these spikes in innovations, and I think a lot of people think we're in that moment today with all the advances coming out in AI. I think the one part where I'm trying to build on and push the long-wave people a little bit is their focus is mostly on the innovation phase.

Rob Atkinson: Right.

Jeffrey Ding: This idea that the new breakthroughs, the big pioneering advances, they all cluster in one country. That country has these heroic inventors, and that's where the power resides. Whereas for me, it's not necessarily about that initial window where one country monopolizes all innovations in the leading sector because for a general purpose technology like AI, which is its own entire research field, no one country is going to control all innovations in AI and your guys' think tanks work has found that, that there's a lot of different countries competing in that innovation phase. It's more about this decades-long competition of who can take those frontier advances from their top firms, their top universities and spread them across their entire economy so that the small medium-sized businesses in Iowa where I'm from are able to adopt that GPT, that general purpose technology and improve productivity growth across the entire economy.

Jackie Whisman: Building on this, you talk about the conventional or dominant theory of technological advantage for a country, and you call it leading sector theory. What is this?

Jeffrey Ding: Yeah, I think it goes back to that earlier thread we were talking about where leading sector theorists, they believe three main things about how technological revolutions affect the rise and fall of great powers. First, they think that new technologies make the most impact really early on in their cycle. The country that's able to seize that brief window of time where they can monopolize innovation before it diffuses to all these other countries, that's the country that's going to win out. So oftentimes, that vision of technological competition is very much based on exports, which country was able to produce the most of this new good, whether it's automobiles, steel output, cotton textiles in the First Industrial Revolution. So the timeframe is that the leading sector makes an impact very early on.

The second main precept is for leading sectors it's all about that innovation phase, which country is able to dominate and be the first to bring new to the world innovations out in this new sector. Then the third thing is leading sector theory sees technological changes concentrated in a few specific industries. My argument is that it goes almost the complete opposite way in all three of those respects for general purpose technologies. So on the timeframe front, we have a lot of literature from economic history and economists that GPTs take about four decades before they actually make a substantial impact on productivity; from the date we got our first electric dynamo to when electricity actually ushered in a productivity growth wave in the U.S. economy.

On that phase of technological change front, which is that second dimension, for GPTs, it's much more about the diffusion phase. It's about embedding and spreading these technologies throughout the entire economic system. And then thirdly, for GPTs, the breadth of growth is much more dispersed. The breadth of technological change is much more dispersed. It's not just one or two sectors. It's not just about the AI sector, for example. It's about are all these other industries also adapting and coming up with complementary innovations to make the most of AI? Because AI's impact doesn't come from just one industry alone. It has to come through this broad base of technological change.

Rob Atkinson: It's interesting you say that because this is, again, music to our ears. We've been focused on that for so long, and obviously we've been focused on both with leading sector development and GPT diffusion. I was just out at Berkeley at a conference that the Babbage Center put on industrial policy, and one of the roundtable participants was the CEO of Alphabet or Google spinoff called Mineral, which is basically looking at applying these technologies to agriculture just like a Kleiner-backed firm in California my son used to work for called the Farmer's Business Network. And it's striking how little focus there is on this. In policy circles, you almost hear almost nothing about AI diffusion. It's all about how do we get better LLMs? To me, diffusion is really key, and to me, one of the key points you made was that there's going to be complementary innovations in diffusion in each sector. There are different institutional factors that are supported or detract from it, including regulations, and we just don't seem to really think about that. Any thoughts on why just because it's not as cool or...

Jeffrey Ding: I think you all would know the political dynamics much better than me with how the domestic political economy affects science and technology policy, but I have a few ideas. One is just we gravitate more towards these heroic inventor stories of technology competition. When people think about technology competition today, they think about can China come up with their own version of open AI, and can they compete with open AI on this leading edge innovation. If you really distill down President Biden's technology policy today when it comes to competing with China, it's about preventing technology leakage so that China can't catch up on the innovation front. I think we have developed good planks with the CHIPS and Science Act about diffusion and building up a STEM workforce, but they're not really the priority. I think we've authorized a lot of money towards STEM workforce, but the Department of Commerce is much more focused on making sure we have the next round of export controls ready to prevent someone else from catching up on leading edge innovation rather than building up the broad diffused STEM workforce to diffuse AI innovations at scale.

And I think one of the central reasons why we don't really have a diffusion-oriented technology policy is there's not a concentrated faction that advocates for it. Innovation-oriented policy, there's always going to be one sector. The semiconductor industry, the AI industry, the big tech lobby, they might advocate for targeted sector- specific policy. Diffusion-oriented policy, it's much more broad based, but the benefits are much more diffuse by definition. So all the benefits a little bit goes to every single city around the U.S. or every single sector in the U.S. economy. So there's not a concentrated faction that's going to lobby for it if you see politics as this game of interest groups trying to lobby for different policies. So those are just some of the reasons, I think, we underemphasize diffusion policy to our detriment.

Rob Atkinson: We were quite involved and been supportive of the CHIPS and Science Act, including the earlier version, the Endless Frontier. One of the things in there originally was to significantly expand the NIST, National Institute of Standards and Technology MEP Program, Manufacturing Extension Partnership Program. And at the end of the day, it didn't happen. $49 billion for CHIPS fabs, which we helped support, love it. But again, just ignored diffusion. So anyway, I just find it strange.

Jeffrey Ding: The extension partnerships are some of the most vibrant and effective ways to enhance diffusion because you're getting the people working on the implementation side, the applied side, working with the people on the innovation side. And so that's how diffusion works is when you have people sharing ideas and collaborating and bringing different sectors of the economy together. So yeah, it's interesting to hear the history of that, and it's a shame it didn't make it to the final version.

Jackie Whisman: My ITIF colleagues wrote in a recent report on Chinese robotics that China installed more robots in 2022 than the rest of the world with the U.S. way, way behind. And this seems like an example of this lack of focus on diffusion.

Jeffrey Ding: Yeah, I think robots are potentially a good indicator to look at. China has certainly installed a lot of robots. I think one of the things that I mentioned in one of the last chapters of the book when I compared the U.S. and China on different diffusion indicators is I think sometimes we have to dig beyond the surface level on these robot indicators. China, when you look at installed robots per worker in the labor force, China's diffusion ranking on that metric is much lower than if you just look at total installed robot capacity. And then oftentimes the statistics we use for the labor force are a little bit skewed for China. And if you use the more accurate labor force estimates, I think, from the ILO, International Labor Organization, then China's diffusion rate on robots is even lower.

And that combined with a lot of qualitative analysis that sometimes the robots are installed but they're not utilized effectively, I think we want to be a little bit careful about on China's diffusion rate on robots. And that fits in with my overall point in the chapter where I compare U.S.-China competition, which is that the U.S. is still very much in the lead when it comes to capacity. I think we still have much stronger linkages between our universities and industry. If you look at the number of papers in AI that are authored by at least one researcher in industry and one researcher in academia, I think the U.S. rate is at least double that of China's. I did some digging of my own into universities that met a certain quality baseline for training just average AI engineers.

The U.S. had about, I think, around 150 of those based off of the CS rankings data set that I was using. China only had about 30. So when you get past the top Chinese universities that are competing on innovation and AI like Tsinghua, Peking University, when you get past some of that first tier, they don't have the breadth of education institutions that we have in the U.S. to train average AI engineers that can then diffuse innovations. I think robots are a good indicator to look at. We should still keep tracking, especially 'cause it's going to interact with AI in the future much more. But we want to be careful because when we're doing these diffusion capacity indicators, you want to control by population because it's about engineering density, it's about spreading innovations across the entire economy. You also want to look at other indicators including on talent and linkages between industry and academia.

Rob Atkinson: I don't disagree with what you said, although one of the things we did in that report based on another report is we created a model that estimated how many robots you should have per manufacturing worker based on your total compensation package. So if you're in Germany, you should have a lot of robots because each robot costs the same no matter where you are, but your labor savings is so much, whereas if you're in Philippines, like 30 years to pay off a robot. So using that cause China's wages are a quarter of ours kind of thing, they were actually quite high. So I agree with you, you got to look at all these different factors.

But I do want to bring in this other point, though, which is, and I know you're not saying this, but I think it's important to bring about, is one of the reasons for focusing on leading sector capabilities as opposed to just diffusion is really around national defense and just capabilities. We could have the highest productivity in the world through AI, and our garbage collection would be super high and our insurance industry would have super high productivity, but if we ever fought a war, we would lose miserably 'cause we couldn't make the drones and things like that and the missiles. First of all, do you agree with that and secondly, how do you see the interaction between the leading sector theory and diffusion theory?

Jeffrey Ding: Yeah, I think you're right to point out the book is not saying we should forsake leading sector innovation. I think there are a lot of forces that will make sure that we're still going to be spending a lot of money on cutting-edge R&D, especially for military applications. The one thing I will say, though, is when we think about how AI will affect military competition, it might not be about who's going to bring forth that first initial leading capability. I did this other paper with Allan Defoe who's now leading DeepMind's AI governance research. We looked at the history of how electricity affected military competition. And back in the late 19th century, everyone was pontificating and speculating, and they thought that electricity would allow countries to create these electric rays of destruction and sweep entire battlefields. But really, the impact of electricity wasn't this super weapon.

It was diffused throughout all different sectors of the military used in coastal defenses, used in communications and transportation, back end logistics to make militaries more effective. So even if we think about economic productivity, I think there's an analog for militaries as well, and military power. Well, sure, we're drawn to those electric rays of destruction that can wipe out entire battlefields. But what's really going to contribute to military effectiveness, at least how AI, that the channel through which AI might contribute most to military effectiveness is through those diffused channels of is this random intelligence unit in the Army going to be able to adopt this new AI technique effectively? Not necessarily, who's going to build the AI super weapon? So I think a lot of the insights actually from the book about pathways of how technologies and institutions interact, a lot of those insights might also translate to military power, not just economic productivity.

Rob Atkinson: Yeah, I used to read Tom Swift books, which dates me, not when they first came out. Probably, Jackie, you don't even know who Tom Swift is or Jeffrey [inaudible 00:19:16]

Jackie Whisman: Is he a relative of Taylor?

Rob Atkinson: Probably Taylor's great-great-grandfather. But Tom Swift was this kid series came out in early 1900s, and it was Tom Swift and His Electric Runabout. Tom Swift and His Giant Magnet, and it was this Tom Swift and His Aerial Warship, so it was exactly that sort of thing. He had the giant magnet. So anyways, the whole thing, he just looked at all this technology, create all these cool things. So you talked about your nephew, have your nephew read Tom Swift to see what he thinks.

Jeffrey Ding: Yeah, I'll check it out.

Jackie Whisman: What are the implications of your theory or approach for both China and the U.S. going forward?

Rob Atkinson: Maybe start with China and then we'll move on to the U.S.

Jeffrey Ding: Yeah. I think the interesting thing about the implications for China is I think China has also overly optimized for this leading sector model. When it comes to R&D targets, for example, and ITIF has done a lot of great research on this, China has hit most of its R&D benchmarks in terms of the goals that it's laying out for how much it wants to invest in cutting edge R&D, how much it's putting into supporting these top frontier firms, how much it's trying to create these clusters of innovation, pump up its big startups, on all these fronts, China is very much succeeding when it comes to trying to follow the leading sector template of innovation policy. But where I think China has misjudged things, it has not been as effective when it comes to diffusion. Now, there are some things that China has done well.

Especially when you can implement diffusion from the top down, there are some things you can do really well, such as building infrastructure with the high-speed rail network. But for other things, especially general purpose technologies that benefit from these fast-acting market-based processes of diffusion where you can't just mandate we're going to require all companies to adopt AI, especially if it's going to go against making profits and it's not going to be the best technology, then it's harder for these top-down initiatives to encourage diffusion. I think one of the examples I draw in the book, and I cite research from a think tank affiliated with China State Council is that while China has hit its R&D benchmarks, it's invested much less in education.

And actually, its education spending as a proportion of its economic development is much lower compared to comparable countries like Brazil, for instance. So when it comes to building a broad base of skills in AI or other emerging technologies, China has not done as much of that STEM workforce development that we talked about, at least compared to R&D. I think for me, my main takeaway when I was looking at different things that China was doing, and China has not really learned the lessons of my book, at least when it comes to the history of general purpose technologies and how they affect the rise and fall of economic leadership.

Rob Atkinson: It's interesting you say that because back until recently, there was fairly consistent visits of Chinese government delegations to ITIF, always wanting to know what's the secret sauce we would have and all that stuff. And I remember once I met with the Minister of Science and Technology for China when I was head of the U.S.-China Innovation Experts group with the White House, I met with him a number of times. But anyway, the point is I made this point every single meeting I was in.

And I said, "Look, unless you get broad based adoption of these technologies across almost every sector, you're not going to catch up to the U.S." and virtually every single meeting I was just ignored. Blank faces, blank stares, "What are you talking about? We want to do biotechnology," or whatever. So it's just crazy. We've got only a couple of minutes left. So Jeffrey, maybe the last point would be you're advising whoever is going to be president in January and who knows who that's going to be? What would you say to the president's economic policy team about diffusion and generally what would be your message to them or Congress?

Jeffrey Ding: Yeah, I think my main message would be the U.S. should rebalance its technology policy away from innovation centrism and more towards a diffusion-oriented approach. And so I think we have some of these building blocks already. We've talked about the CHIPS and Science Act that I think authorized a lot of money to be spent towards STEM workforce development. Now, the next step is actually implementing that, and I think that requires then allocating that money, dispersing that money and making sure it's spent. I think we've done a good job with the fab building. We're making sure that that money is going towards the fab building. We should try to do the same for the other important planks of that act. So some of it is just implementing the infrastructure we've already put into place and pushing for that to be done.

Then the other set of recommendations I would say would be to build up what I call these technology diffusion institutions. Those are things like the manufacturing extension partnerships, investing in those. Other countries have adopted things like what they call innovation voucher programs where they encourage small and medium-sized firms to bring in people from frontier firms to introduce new technologies. That also includes investing in a broad range of education institutions, not just focusing on the centers of excellence, but investing in community colleges, alternative ways to pick up data science skills like certification programs. So those would be a few of the recommendations I would put at the table. But the overall message is rebalancing our technology policy towards one that's centered on diffusion.

Rob Atkinson: Yeah, that's fantastic. Yeah, we've been long proponents of innovation vouchers, and we actually worked with, I think, two or three different states to get the state programs. We've been encouraging a federal program as well, and everything else you said absolutely been pushing for, so we'll keep our fingers crossed. I do think that there's more awareness that we got to do more of this. It's funny because when you look at history and then we do have to stop, one of the most successful systems that we've ever had was really agriculture 'cause we combined research with extension. So the extension and the ag researchers worked hand in glove, and it was-

Jeffrey Ding: Exactly.

Rob Atkinson: ... a core reason. Besides having a lot of fertile land, it was a core reason why the U.S. led, dominated the global agriculture industry, and there's no reason we couldn't do that with other agents. You look at Singapore for example, they have a retail diffusion program to help their retailers become best in class. We don't even think about that. So boy, Jeffrey, we could just keep going 'cause as you know, I'm super interested in this stuff. And I love your book, and I hope everybody buys a copy and reads it. So anyway, thank you for being here. It was really great.

Jeffrey Ding: Thanks for having me. Nice to meet you both.

Jackie Whisman: And that’s it for this week. If you liked it, please be sure to rate us and subscribe. Feel free to email show ideas or questions to [email protected]. You can find the show notes and sign up for our weekly email newsletter on our website itif.org. And follow us on Twitter, Facebook, and LinkedIn @ITIFdc.

Rob Atkinson: We have more episodes and great guests lined up, and we hope you will continue to tune in.

Jackie Whisman: Talk to you soon.