The War Economy: Is America falling behind China in science?
Not yet. But we should make sure we act now to stay well ahead.
As you might expect from the picture at the top of this post, I’m a little ambivalent when it comes to breathless reports that America is falling behind its rivals technologically. That picture is from the movie Dr. Strangelove, which (among other things) lampooned America’s Cold War obsession with the “missile gap”.
On one hand, in a geopolitical contest such as the one we now find ourselves in with Russia and China, we need to prioritize which battles to concentrate resources on — a point made very convincingly by Hal Brands and Michael Beckley’s new book Danger Zone. But on the other hand, our obsession with the “missile gap” gave us the space race and the moon landings and all the technological spinoffs from those, plus a boost to our semiconductor industry. And having a sufficient number of missiles to roast the USSR did come in handy for Mutual Assured Destruction, the strategy that ended up successfully preventing nuclear war.
Therefore I think it makes sense to worry about technological gaps, but at the same time we need to pay attention to details and think critically when doing so. So it was with a cautious, skeptical eye that I read this thread about China zooming ahead in science:
The thread is eye-opening. I don’t think it’s time to panic yet — especially not from just looking at aggregate statistics. But a deeper look suggests that there are key areas where China’s rapid progress could threaten the kind of military-technological supremacy that America held in the previous century.
How formidable is Chinese science really?
One especially eye-catching graph in the thread above is this one from The Economist back in 2018:
That’s in math and computer science only. But Chinese universities have also been rocketing up the rankings in the Nature Index, with 8 out of the top 20 spots in 2021. Overall, China has almost pulled even with the U.S. in this measure of research output. Other measures say China is already ahead.
Now, these research output numbers are a more reliable guide to China’s scientific prowess than research input numbers such as total spending or number of researchers. We all know China spends a lot; these indices show that it now publishes a lot as well, including lots of highly cited papers and lots of papers in top journals.
The data sources that depend on citation counts, however, should be taken with a grain of salt. It’s well known that groups of researchers can create “citation rings”, in which they all cite each other’s papers. This happens everywhere of course, and sometimes it’s hard to draw the line between intentional citation inflation and simple clubby human networking. But evidence suggests that Chinese scientists engage in more of this than scientists in the U.S. Elisabeth Bik writes:
A recent study (paywalled) showed that the quality of research from China – as measured by the number of citations a paper gets – exceeded that of the US since 2019…
A 2015 article found that ‘highly cited Chinese papers are more likely than are similar highly cited U.S. papers to be cited by works from China and from the same institution or author; that is, substantial citation differences between U.S. and Chinese highly cited papers exist at all three levels of internal citation‘. As the authors explain, ‘The norms of interpersonal relationships (guanxi) in China may lead Chinese scholars to cite the work of their colleagues in the same institute, who they meet frequently, or leading scholars in their own country, who have an influence in proposal review and external evaluation for promotion.’
And a 2015 paper by Tang et al. found:
This research explores another possibility—whether there is a “clubbing” effect in China's surge in research citations, in which a higher rate of internal citing takes place among influential Chinese researchers…Both descriptive and statistical tests suggest that highly cited Chinese [nanotechnology] papers are more likely than similar U.S. papers to receive internal and localized citations.
But the Nature Index isn’t based on citation counts — it’s based simply on the number of publications in a set of 82 top journals, few or none of which appear to be journals that are themselves run by Chinese organizations.
[D]espite the rapid growth in articles authored by Chinese scholars in the Science Citation Index (SCI) over the decade to 2017, the average number of citations for each article was only 9.4. This is lower than the global average of 11.8, putting China in 15th place by this measure. The SCI tracks articles in high-impact journals…
According to the closely watched Chinese ranking website Netbig, even China’s leading laboratories or centres of excellence in such flagship fields as materials research, metals research and chemistry, including those affiliated to the prestigious Chinese Academy of Sciences (CAS), are not ranked among the world’s top 10…
My interviews with 19 young researchers and scientists in China over the past two years confirm the pressure they felt to publish articles in SCI journals as quickly as possible…These evaluation systems have also led to the proliferation of research malpractice, including plagiarism, nepotism, misrepresentation and falsification of records, bribery, conspiracy and collusion.
Now, if Huang is right about this — and keep in mind that the article is four years old — it’s an indictment of the scientific journal system that so many mediocre papers get published in top journals. In fact, the replication crisis and high-profile instances of fraud in the West suggest that the publication system is broken in important and worrying ways. But if the problem is indeed worse in China, it means that China’s apparent rise to scientific supremacy is perhaps not quite as impressive as it appears.
There’s also the question of what, exactly, Chinese researchers are discovering. It’s hard for even boosters of Chinese science to name big breakthroughs that have come out of the country in recent years, and researchers working in China have only won one science Nobel prize ever (Tu Youyou, who discovered a malaria drug and won the prize in 2015).
So perhaps some portion of the flood of Chinese research publications is simply professors cutting corners and/or pushing mediocre results through a semi-broken Western journal system in order to survive in China’s incredibly harsh publish-or-perish academia.
But I wouldn’t bet the farm on this conclusion. First of all, China made some moves to rein in its perverse incentives for scientists in 2020, banning cash incentives for publication and reducing the importance of publication quantity. Yet despite these moves, Chinese universities continued to charge up the Nature Index rankings in 2021.
There’s a heck of a lot more going on here than gamed statistics.
Specific areas of concern
My brief attempt to identify recent Chinese breakthroughs did yield a few concrete examples of impressive and groundbreaking achievements.
The first is in quantum computing. A few months ago, a multi-university team of Chinese researchers announced a result that blew away anything the U.S. has achieved:
In 2019, Google reported that its 53-qubit Sycamore processor had completed in 3.3 minutes a task that would have taken a traditional supercomputer at least 2.5 days. Last October, China’s 66-qubit Zuchongzhi 2 quantum processor reportedly completed the same task 1 million times faster.
Here’s the paper, in case you want to check it out. In fact it’s not China’s only notable breakthrough in quantum computing; around the same time, another team discovered a method to significantly improve the number of photons a quantum computer can detect.
Quantum computing is a big deal, not just because it could eventually replace normal computing and create an entirely new digital revolution, but also because of the potential military applications — especially the possibility that existing forms of encryption could be rendered useless.
In a similar vein, Chinese researchers are also now leading the world in long-distance quantum communications. They can now communicate with satellites at a distance of 1000 kilometers using quantum entanglement — a completely secure form of communications that previously was only possible over short distances. And another team increased the distance over which single photons can be transmitted through fiber-optic cable by 2 orders of magnitude. Obviously these results also have military implications, since they could create communications networks that are much more secure than current ones.
China has also seemingly managed to make big breakthroughs in the area of quantum sensing. Long-range quantum magnetometers, or SQUIDS, could be used to detect U.S. submarines — including our nuclear subs, the backbone of our second-strike capability — at great distances. (Update: A good explainer on quantum sensing here from Jyotirmai Singh).
Together, China’s breakthroughs in quantum computing, communications, and sensing add up to something very significant. The U.S., which invented quantum computing and did much of the fundamental research in quantum mechanics itself, now finds itself lagging in some of the technology’s practical applications.
A second area of concern is machine learning, or “AI” as it’s often called. There are tons of articles out there about how China is taking the lead in this field. But when we look at who’s actually publishing the papers, we see that America still has a commanding lead:
And few of the truly stunning breakthroughs in AI — for example, Google’s AlphaFold system to predict protein folding — have come out of China.
But that’s not necessarily reason for complacency on this front. First of all, many of China’s AI advances are in the military field — autonomous drones, advanced sensors, and so on — and these would tend not to get published. Also, it’s worth noting that although the U.S. publishes a lot more AI papers than China, a plurality of the top researchers in the field actually come from China:
Many of those researchers might eventually return to China as Cold War 2 gets more and more serious. And they may take with them the AI insights they gleaned while working at Google or Stanford or MIT.
There are some other areas in which China is reputed to be ahead of the U.S., such as nanotechnology and perhaps in certain areas of biotech. I certainly don’t mean to claim that the examples above are an exhaustive list of research fields where China is already out in front.
But quantum technology and AI/ML are useful examples because I think they’re illustrative of the kind of research fields where China is surging to the fore. They’re not exactly what we’d call fundamental science — there aren’t new principles of the Universe being discovered here. Instead, they sit somewhere in the middle of the spectrum that runs between basic and applied science, featuring inventive and novel methods and big increases in the capabilities of emerging technologies. And they all seem to have key military applications.
In other words, this isn’t the kind of scientific leadership that wins Nobel prizes, but it might be the kind that wins wars.
How to win the science race
To win the science race with China, we need to think about what we’re actually racing towards. America is still probably a lot better than China at uncovering the fundamental secrets of the Universe. But the areas where Chinese research is catching up to or even overtaking U.S. research is in the area of invention, especially inventions with military applications.
In fact, the areas where China is gaining scientific supremacy are very much the kind of thing that DARPA handles in America. There has been no shortage of books, papers, and blog posts about the efficacy of the DARPA model, which famously gave us the internet and GPS. The basic model involves strong independent program managers who identify a target invention and pull top researchers from various universities in order to build it. (Looking at the collaboration between institutions in some of these Chinese scientific breakthroughs, I’m wondering if China is doing something similar.)
We should respond to the Chinese scientific challenge by making better use of DARPA. Funding for the agency has been basically flat in inflation-adjusted terms since the 1990s:
And as a share of defense research spending, DARPA spending has actually fallen. Given the agency’s storied and successful history and the particular nature of the Chinese scientific challenge, we should probably pump up DARPA’s funding substantially.
Another important arm of U.S. government research that we’ve been neglecting is our national labs. Arora et al. (2019) have a great graph showing that even as university and industry research spending has increased massively over the last half century, the amount of spending on research actually done by the federal government has risen only anemically:
This should be remedied by a big increase in spending on national labs (or Federally Funded Research and Development Centers, as they are officially known). These labs have produced a large number of important research breakthroughs in the past. And there are a number of reasons to do research this way, especially when the technologies being researched have important potential military applications.
For one thing, doing research through national labs gives the government more direct control over what’s being researched — instead of having to rely on what Google feels like it could commercialize, national labs can focus on things that might represent more basic, less monetizable breakthroughs, or on things that might be immediately useful for the military. Second, in the current Cold War 2 climate, vetting of researchers — especially Chinese nationals — has become more important, and national labs probably make this vetting easier than corporations or universities.
Spending more money on research is really the key here, because that’s what creates demand. Yes, supply of research inputs is important — we need lots of high-skilled immigrants, and we need to push more Americans to go into STEM. But unless there are jobs for those researchers, Americans will be discouraged from going into the field, and the presence of immigrants will seem like a cutthroat zero-sum competition. Spending more money ensures that anyone who has the talent can see their talent put to work.
So far, the current U.S. industrial policy revival has been a bit disappointing on this front. The CHIPS Act lobbed a bunch of money at the semiconductor industry, but this represented a vastly scaled-down bill that chopped out most of the best parts of an earlier version called the Endless Frontier Act. The Endless Frontier Act could have revitalized U.S. scientific research, far beyond the narrow confines of the chip industry, but our congressional leaders simply didn’t see the point.
Well, the point should be obvious now. When China has quantum computers a million times faster than Google’s, satellites that can talk securely with the Earth via quantum entanglement, quantum magnetometers that can (possibly) find our most secret submarines, and autonomous drone swarms that can fly through dense forests, Congress would have to be insane to worry about pinching a few pennies.
The U.S. hasn’t fallen behind in science yet. But in order to avoid falling behind, at least in many key areas, we need to take bold action now.