Oliver Sacks on the nineteenth century’s love of facts

Oliver Sacks’ posthumously published essay collection The River of Consciousness is a surprise and a delight. While it has some pieces in his familiar style of reflections on neurological casework, the highlights are the truly wonderful essays on the history of science. Who knew that Darwin discovered the pollination of flowers by insects? Or that Freud did foundational research on the structure and role of nerve cells?

Informing these essays is Sacks’ deep affection for and engagement with the work of nineteenth-century scientists, particularly Darwin and Freud, but also many more obscure toilers. At one point, when investigating some of the peculiar visual hallucinations experienced by his migraine patients, he can find no help in twentieth-century psychiatric literature, so he looks further back:

When I searched the current literature, I could find no mention of these [phenomena]. Puzzled, I decided to look at nineteenth-century accounts, which tend to be much fuller, much more vivid, much richer in description, than modern ones.

Sacks found that those nineteenth-century writers, while often lacking a theoretical framework to interpret their observations, were meticulous recorders of what they observed. Twentieth-century psychiatry had a more developed theoretical system, but had little time for phenomena that did not easily fit into that system, and so ignored them. There is perhaps a parallel for this in anthropology, where the extremely detailed accounts of early fieldworkers can still be usefully mined for insights for decades afterward–something it is difficult to imagine happening with many contemporary works with a much more elaborate theoretical apparatus. A mindset that places value on facts is itself something of value.

In the nineteenth century, an era of naturalistic description and phenomenological passion for detail, a concrete habit of mind seemed highly appropriate, and an abstract or ratiocinating one was suspect—an attitude beautifully brought out by William James in his famous essay on Louis Agassiz, the eminent biologist and natural historian: “The only man he really loved and had use for was the man who could bring him facts.”

The nineteenth-century genius for, or mania for, the collection and description of facts is definitely one of the most distinctive traits of the epoch. Jürgen Osterhammel’s The Transformation of the World: A Global History of the Nineteenth Century, one of my favorite history books, describes this very well, though with more attention to the social than the natural sciences:

The novelty in nineteenth-century Europe was that, over and above a normative political and social theory, branches of knowledge arose with the aim of describing the contemporary world and grasping the patterns and regularities beneath the surface of phenomena. …

“Factual investigation”—which Joseph A. Schumpeter contrasted to “theory” in his great history of economic thought—acquired new scope and significance in the nineteenth century, when Europeans produced incomparably more self-observational and self-descriptive material than they had in previous centuries.

For the most important analysts of political and social reality—one thinks of Thomas Robert Malthus, Georg Wilhelm Friedrich Hegel, Alexis de Tocqueville, John Stuart Mill, Karl Marx, Alfred Marshall, and the chief figures in the German “Historical School” of economics, including the early Max Weber—factual investigation was closely bound up with the theoretical quest for connections and correlations.

Indeed, Schumpeter’s History of Economic Analysis has a lot of praise for “factual investigation,” and he particularly liked works that defied the stereotype of economics being excessively theoretical:

Of particular interest to us is the type of analysis that combines presentation and explanation of facts in such a way that the two cease to be distinct tasks and mutually condition one another at every step: the type of analysis that arrives at its results by means of discussing individual situations. … It is hardly possible to overlook the factual complement in the Wealth of Nations—though some critics seem to have accomplished even this feat.

A lot of those massive, fact-filled nineteenth-century tomes are certainly impossible to read today, but the greats of the era were able to integrate voluminous facts with theorizing and strong arguments. Such a style of analysis was precisely what Sacks enjoyed about Darwin’s later botanical works. These are little known compared to the Origin of Species, and yet Darwin spent decades of his life on them.

Darwin spoke of the Origin as “one long argument.” His botanical books, by contrast, were more personal and lyrical, less systematic in form, and they secured their effects by demonstration, not argument. …

Botany was not a mere avocation or hobby for Darwin, as it was for so many in the Victorian age; the study of plants was always infused for him with theoretical purpose, and the theoretical purpose had to do with evolution and natural selection. It was, as his son Francis wrote, “as though he were charged with theorising power ready to flow into any channel on the slightest disturbance, so that no fact, however small, could avoid releasing a stream of theory.”

Sachs accepts that the twentieth-century transformation of botany and zoology from sciences founded in descriptive natural history to more theoretical enterprises led to tremendous progress, but “it was clear that something was being lost, too.” With economics also having taken an empirical turn over the past couple of decades, perhaps there will be a swing back to appreciating some of those nineteenth-century virtues.

415uvbcj6ul._sx322_bo1204203200_

How much did outside advice really change things in Russia and China?

Recent polemics against neoliberalism have revived an old debate over the role of the economic advice given to developing countries by the World Bank and IMF. A crude but nonetheless influential interpretation of the relevant economic history holds that Russia’s failed “shock therapy” privatization of SOEs in the 1990s was the result of uncritical acceptance of free-market dogma pushed by the international financial institutions, while China’s successful “gradualist” approach to SOE reform was the result of wise officials ignoring those same institutions and carefully designing policy according to local conditions.

public-wealth

Chart from Novokmet et al. “From Communism to Capitalism: Private Versus Public Property and Inequality in China and Russia”

This interpretation may accord too much importance to the advice given by the international financial institutions, and too little to the domestic politics of the countries actually making the decisions.

John Nellis, a participant in the World Bank’s first mission to the USSR in 1990, has published an account of that period based on his notes taken at the time. It makes for fascinating reading. It’s particularly interesting that the famous Soviet State Planning Committee, or Gosplan, seemed committed early on to a “gradualist” approach to reforming state ownership:

Even here, in the principal basilica of socialist planning, no one questioned that the old system had failed and that a transition to the market, or something approximating a market, was urgently required. But those we met in Gosplan, and many of those we met in other Soviet ministries and central units, thought that the transition would be a gradual, lengthy affair, and that the outcome would be some sort of mixed approach. In this evolutionary process they thought (or hoped) that Gosplan would retain authority to forecast, analyze, assist, guide and even lead reform. …

As for the future of the real sector, the officials’ evolutionary vision was that the massive, multi-divisional state enterprise/ministerial complexes would be broken down into “correctly sized” units. These would then go through a process of “corporatization” and would become joint stock companies, with all shares initially held by the state. These would then undertake a process of finding private partners, Soviet or foreign, who would bring in capital, technology, management expertise, and access to markets. Some percentage of shares would have to be turned over to these partners, but it would at first be a minority share, particularly for foreigners. These processes had just begun to start and, in their view, years would pass before substantial results were seen. Central organs such as Gosplan would guide and assist this evolution. Majority private ownership was a long-term prospect.

This aspiration is not so different from the course that was actually followed by China. (Nellis also notes that a 1988 Soviet law had allowed for the creation of cooperatives, which, much like China’s township and village enterprises in the 1980s, often functioned as de-facto private companies.) The joint World Bank-IMF report that was published after the mission acknowledged that large-scale privatization was effectively impossible, and focused more on how to manage state enterprises effectively.

All this suggests that China’s “gradualist” approach to overhauling state ownership was less a strategy adapted to uniquely Chinese conditions, but more the approach most likely to be favored by a socialist government that wanted to maintain political continuity and control over the reform process. Yet by 1992, the Soviet Union had been dissolved, and the Russian government launched a program of rapid mass privatization using vouchers–a much more radical approach than anything that had been considered in 1990. Nellis asks the obvious question:

The overwhelming majority of persons we spoke to in 1990 were gradualists. They wanted to effect as painlessly and politically acceptable as possible a transition to the market. …

Why did the 1990 joint IFI mission not get a glimpse of the coming emphasis on mass privatization? How did it — we — miss the fact that the government of the Russian Federation would opt for audacity?

The answer, clearly, is the radical change in domestic politics after the vote to dissolve the Soviet Union in 1991. In particular, the failed coup against Gorbachev, which was led by representatives of the same conservative interest groups that had tried to stymie economic reforms. After the failed coup, the reform and privatization of state enterprises was no longer a technocratic matter of economic management, but an urgent political task to dismantle the strength of the interest groups that had led the coup. The new Russian government was driven by an “overriding fear that the communists might try again to regain power,” Nellis writes. And the reshuffle of domestic politics had elevated to decision-making positions people who were not that important in 1990, and had not previously had well-formed views. 

A more recent, if less detailed, summary of the World Bank’s involvement in Chinese SOE reform by Zhang Chunlin serves as something of a companion piece to Nellis. Zhang is currently the lead private sector development specialist at the Bank, and previously worked on Chinese SOE reform both at the Chinese government and the World Bank. He writes that

The Bank’s work in the 1980s focused on the reform of the traditional SOE model itself while maintaining state ownership. Recognizing the need for state direct control over some “important enterprises” such as public utilities, the [1985] report argued that once a suitable economic environment is created through price reform and competition, pursuit of profit should lead most state enterprises in economically appropriate direction. The fundamental problem remains of the proper relationship between the state and the enterprise.

The central theme of the World Bank’s recommendations for China was not the necessity of privatization, but of corporatization: giving state-owned enterprises the legal form of modern corporations. That promised to improve management and decision-making within SOEs. But it also posed the problem of how the state was to exercise its ownership rights to control these firms. Much of the Bank’s work since the 1990s has focused on finding the right institutional structures for effective state ownership, and it has advocated for reducing state ownership in many sectors.

But the radical downsizing and privatization of SOEs that started around 1995 and continued through about 2002 was a domestic decision driven by the dire financial situation at many firms. A World Bank report in 1997 did call for state ownership to “completely withdraw from inherently competitively structured industries where small and medium sized firms predominate,” but it noted that this recommendation “would formalize a process that is already underway.” (And, of course, China did not actually follow this recommendation.)

Zhang also notes that in later years the World Bank contributed to the debate over the creation and structuring of an agency to represent the government’s interests in SOEs, the body now known as Sasac. It’s less clear if this is a contribution the Bank should be proud of: Sasac is widely regarded as a conservative interest group that has worked to strengthen the position of large SOEs, rather than to further their effective reform. But Zhang mainly wants to emphasize the “productive partnership” that the Bank has had with China. “In retrospective, a clear reason why the Bank managed to stay relevant has been its willingness to adapt to China’s own reform strategy,” he writes.

Yet that is perhaps not so different from how the World Bank worked with Russia in the 1990s: it was willing to adapt to both the gradualist preferences of the Soviet leadership in 1990, and the radical program of the new Russian government in 1992. In the case of both Russia and China, the World Bank seems to have mainly tried to help their governments find the best way to implement decisions that had already been reached by domestic political leaders. It’s not clear that the advice of international financial institutions really played a decisive role in making those decisions.

 

 

Does the state still love state-owned enterprises even after they’re not state-owned anymore?

That is the question raised in an interesting new NBER paper on the aftereffects of China’s privatization of SOEs in the late 1990s: “Can a Tiger Change Its Stripes? Reform of Chinese State-Owned Enterprises in the Penumbra of the State,” by Ann Harrison and four coauthors.

They report that in 2013, 45% of industrial SOEs received government subsidies, but only 15% of private industrial firms received them. Privatized SOEs, however, are different: 25-35% receive subsidies, much more than firms that were always private. Similar results show up in access to finance: privatized SOEs pay higher interest rates than current SOEs firms but lower interest rates than firms that had always been private.

nber25475-fig3

Their hypothesis is that “allocation of government support in the form of subsidies, tax breaks or low interest loans favors both SOEs and former SOEs.” They conclude that while SOEs do change their behavior after being privatized, the government does not change its behavior toward those companies: “former SOEs retain ready access to large loans, concessionary interest rates, and outright subsidies.”

But why should this be the case? The paper is curiously silent on this.

The whole point of the SOE privatization over 1998-2003 (after that, privatization basically stopped) was so that the government would not have to keep giving free money to money-losing SOEs. Privatizing the SOEs, and then continuing to give them lots of free money, rather seems like the worst of both worlds. So it seems rather unlikely that this pattern is the result of a deliberate, centralized policy choice.

My hypothesis would be: it’s about people. Let me propose that the managers of SOEs understand the government pretty well and are pretty good at extracting benefits from the system. Since, after all, that is a good part of their job.

Furthermore, the managers of privatized SOEs are often going to be the same people who managed the firm when it was state-owned. One survey of SOE privatizations found that sales to insiders (aka management buyouts) were by far the most common type of deal, accounting for 47% of their sample.

Therefore, privatized SOEs are often going to be run by people who have established relationships with government officials and managers of state banks. It stands to reason that they are going to be better at working those relationships than the executive of a run-of-the-mill private company. And that companies that have been able to obtain subsidies in the past would know best how to continue obtaining those subsidies in the future.

On the financial front, this behavior by state banks is not irrational. If a state bank has an established relationship with an SOE, that relationship doesn’t just evaporate once the SOE is privatized. The bank will still have a lot of knowledge about its customer, and have a history that makes it more comfortable continuing to lend on favorable terms.

So the persistence of favorable treatments for SOEs after privatization is probably mostly about the persistence of relationships–and firms’ understandable unwillingness to relinquish established sources of commercial advantage.

A preview of Nick Lardy’s new book *The State Strikes Back*

A new Nick Lardy book comes along regularly every few years, and each one is an event for the China-watching community. Anyone who cares about the Chinese economy will find The State Strikes Back: The End of Economic Reform in China? interesting and provocative. This is a preview, not a review, since the book is not officially out until next week and so my Kindle pre-order hasn’t downloaded yet. But I saw his book talk in Seattle last night, where he gave a characteristically clear and concise summary of the argument (he also has an op-ed in the FT.)

The new book has to be understood in the context of Lardy’s previous book from 2014, Markets Over Mao: The Rise of Private Business in China. In that book he argued that it was the rise of increasingly efficient and productive private-sector companies that has driven China’s economic growth over the last four decades, not state-owned enterprises, government planning and industrial policy. In contrast to the view in some quarters that China remains a fundamentally state-controlled economy, he laid out all the ways in which markets have been liberalized and competition increased since 1978.

A lot of the key changes in the relationship between the state and private sector happened in the 1980s and 1990s, and are well explained in that book. But Lardy also engaged with the argument that, as he put it, “state-owned firms returned to prominence of the decade of leadership of President Hu Jintao and Premier Wen Jiabao (2003-12)”.

While acknowledging that the Hu-Wen government wanted to make state enterprises and industrial planning play a a bigger role in the economy, he argued that the data showed they had not succeeded. In fact, the private sector’s share of economic aggregates had continued to increase, not because of continued privatization but because private firms are more efficient and grow faster than SOEs. This process was aided by a substantial increase in private firms’ access to bank credit.

The main point of Lardy’s new book, based on his slides and talk, is that the positive trends he had emphasized in his last book are now going in reverse. The data now show that private firms’ access to bank credit has sharply declined, and that their share of various economic aggregates is also falling. He puts particular emphasis on the drop in lending to private firms:

soe-private-loans

(Note: Lardy has a chart like this in his slides, but this is my chart not his. It is based on the same underlying data but my estimates come out slightly different.)

The big decline in bank lending to the private sector (the absolute volume of new loans to private companies shrank, not just the share) had major consequences. It forced private firms to rely even more on shadow finance. But in 2016 the government also decided (correctly) that the rapid expansion of shadow finance posed systemic risks. The tightening of regulation led to an outright decline of shadow financing in 2018, putting many private firms into dire financial straits. The financial pressure on private firms has allowed their state competitors to expand at their expense: SOEs in industry are growing faster than their private competitors. Lardy said this is the first time this has happened since 1978.

iva-soe-private

(Again, this is a re-creation of one of Lardy’s charts using public data.)

Lardy thinks all this is bad for China. He is right! He also puts most of the blame on the policies of Xi Jinping–tolerating SOE inefficiency, encouraging the creation of larger SOEs, tightening Party control over private firms–since these trends in the data did not show up until a few years into his administration.

Essentially, both of Lardy’s recent books are about the use of economic data to support a narrative about the direction of reform in China. In Markets Over Mao, he argued that the data did not support a narrative of the resurgence of the state sector, and in fact supported a narrative of the rise of the private sector to new heights. I think it is fair to say that a number of people felt that Lardy in that book was too forceful in downplaying trends that were in fact important, but perhaps were difficult to tease out in the aggregate economic data. Now, Lardy is arguing that the data support a narrative that the state is resurgent and the private sector is losing out. Since this is a recent reversal of a positive long-term trend, he thinks that if China changes course it could significantly boost its economic growth rate, by as much as 2 percentage points.

My own view is more that economic policy under Xi Jinping represents an intensification of trends that were already playing out under Hu Jintao. I think this is pretty clear if you pay attention to China’s official rhetoric and try to understand the underlying political economy. Since I think the problems go back further than 2015, I am less optimistic than Lardy about China’s longer-term growth prospects (thanks to Greg Ip of the WSJ for including a summary of my views in his latest piece).

I also think that it is tricky to tell a clear story about the rise or fall of the state sector using the official economic data–having spent a lot of time and effort trying to do that myself. As someone who very much appreciates Lardy’s careful work with Chinese data, let me offer a couple of caveats to the charts above, in the spirit of seeking truth from facts.

First, on the bank lending data. Lardy is right to highlight the sharp downturn in lending to private companies in 2015-16. But it is not clear to me that this is a result of government policy to favor SOEs. Recall that there was a pretty serious economic downturn in 2014-15. It would make sense for banks to respond to that by trying to reduce the risk in their loan books, and one obvious way to do that would be to curtail lending to smaller and riskier companies, i.e. private ones. (The fact that SOEs are seen as less risky than private companies is a structural problem, but it’s nonetheless true that banks are correct to make this judgment given the realities of China’s political economy.) In other words, the change may have been more cyclical than structural.

There is some preliminary evidence that supports this interpretation. The data that Lardy and I use to calculate lending to state and private firms is released with a long lag, and recent figures aren’t out yet. But banking officials disclosed last year that lending to private firms totaled 30.4 trillion renminbi as of September 2018. This is equivalent to 38% of outstanding corporate loans–which is roughly the same level as in 2013, and a big increase from the 32% in 2016 (again, this is the share of outstanding loans; the chart above is the share of new loans made each year). This suggests that new loans to private firms rebounded in 2017-18 (probably more in 2017) as the economy recovered.

Second, on the industrial data. The fact that industrial SOEs are increasing their value-added faster than private companies is certainly notable. But SOEs and private companies tend to operate in different industries, so it can be hard to tease out the difference between sector effects and ownership effects. Industrial SOEs are concentrated in upstream, commodity-producing sectors, while private firms are more in downstream manufacturing sectors. It seems quite likely to me that the big decline in SOE value-added in 2015-16, and its rebound in 2017-18, have the same source: swings in commodity prices that had big effects on their profitability (value-added is basically profits plus labor compensation). The chart below uses monthly rather than year-to-date data, and we can see that the growth in SOE value-added has recently fallen back below that of private firms as steel and oil prices have come down.

iva-ppi

Lardy is right that the fact that in these charts the red line (SOEs) is above the blue line (private firms) is significant and concerning. But if this year or next the blue line moves back above the red line, will that mean China’s private sector is out of the woods, and all is fine? I suspect not.

51elq7mgmxl

 

 

 

What will it take to lower China’s investment rate?

The below is from Paul Krugman’s latest on China, an argument also strongly endorsed by Brad Setser:

China really can’t keep investing 40-plus percent of GDP. It needs to shift over to higher consumption, which it could do by returning more profits from state-owned enterprises to the public, strengthening the social safety net, and so on. But it keeps not doing that.

Myself, I think it’s weird that people who want China to invest less tell it to do all these structural reforms rather than just, you know, invest less.

They miss the point that much of the low-return investment in China is done by the government and the state sector: it’s all those local infrastructure projects. That’s really where the buildup of excess investment is happening, not so much in the private business sector (which faces hard budget constraints and often tough access to credit). According to the World Bank’s latest China Economic Update, China’s public investment has averaged 16% of GDP since 1978, while OECD countries spend about 3.7% of GDP.

herd-wb-gfcf-chart

So if the government wanted to make a policy choice to invest less, it can just directly make the state sector invest less in those crappy low-return projects. It doesn’t have to overhaul social policy first.

The point of strengthening the social safety net, in this framework, is to reduce precautionary household savings. But high household savings don’t directly lead to excess investment. They do help keep the banking system liquid which enables a lot of borrowing by SOEs.

But trying to impose financial discipline on SOEs by improving the safety net and lowering household savings is pretty indirect. The central government could just require investment projects by SOEs and local governments to clear more hurdles.

Fundamentally, the reason that China invests a lot is that the government has made a decision to keep public-sector investment high in order to boost aggregate demand. If/when that changes, the investment rate will come down. And so will growth. Which is why China is not in a rush to make that call.

The hard choice that China has to make is not whether to undertake complex and difficult technical reforms to social policy. The hard choice is to decide when the efficiency losses from forced high investment start to outweigh the benefits of the boost to aggregate demand.

Some people are interpreting the government’s recent pledges to avoid “flood- like” (大水漫灌) stimulus as a sign that they have in fact reached this conclusion, and want to wean the economy off low-return infrastructure projects. Maybe a bit, at the margin. But the leadership is also going to a lot of trouble to create new funding mechanisms to ensure local infrastructure projects can continue, so it seems clear they don’t want this shift to happen right now.

Redirecting some fiscal resources from investment to consumption (i.e. more social programs) could certainly help soften the blow. But this is a compelling argument only to macro people; the Chinese interest groups that would lose out from less public investment are not going to feel compensated if more social benefits go to households.

Seizing the moment for artificial intelligence: my take on the US-China rivalry

Today I am republishing a piece originally written for Gavekal Dragonomics clients a few months ago, on the US-China rivalry in artificial intelligence. Usually these pieces stay behind the paywall, but with our partner company Evergreen Gavekal we’re making it available for a general audience.

I wrote this piece as a way of sorting out of my own thinking on how to place recent technological trends in the broader story of China’s economic development. I make no claim to be an expert on artificial intelligence: these are just the thoughts of a China watcher trying to absorb what the technologists are saying.

The full text follows below, or you can download the PDF:

 

Seizing The Moment For Artificial Intelligence

by Andrew Batson (originally published May 3, 2018)

In the escalating trade dispute between the US and China, technology has increasingly become the key issue, overwhelming more traditional economic topics like tariffs, deficits and currency valuations. Both countries see their economic future as depending on their success in high technology, and each is worried they will lose out to the other. One of the most intense areas of focus is artificial intelligence, where recent rapid breakthroughs have captivated investors and the media—and where China has emerged as the main US rival.

In this piece, I will try to provide some nontechnical answers to the questions of the moment: What is artificial intelligence anyway, and why is it a hot topic? Why does China seem to be doing so well in artificial intelligence? And how should we think about the rivalry between the US and China to develop this technology? In brief, I think China will do well in artificial intelligence, in part because the technology is now in a phase that plays to its strengths. But it does not make sense to think of the US and China being engaged in an artificial-intelligence “race” along the lines of the US-Soviet space race.

Machines don’t think, but can do useful stuff

A common-sense definition of artificial intelligence, what a layman might understand the term to mean on the basis of reading the news and watching television, is probably something along the lines of: machines that can think like human beings. Artificial intelligence in this sense does not exist, and according to most researchers in the field there is no prospect of it coming into existence any time soon. There are enormous philosophical and technical challenges in understanding how human minds work and replicating those processes in software, and most of these challenges have not been solved.

A more precise definition of artificial intelligence, closer to that used in the industry, would be: the development of computer systems that can perform tasks associated with human intelligence, such as understanding speech, playing games, or driving cars. In a way the term is misleading, because what is being mimicked is not human intelligence itself, but the practical results of intelligence being applied in specific contexts. Artificial intelligence in this narrow sense does very much exist, and its progress is now attracting feverish interest from business, venture capitalists, governments and the media.

The turning point seems to have come around 2014-15. Since then, software programs have been able to match or exceed human performance in tasks that previously could not be reliably performed by machines: recognizing faces, transcribing spoken words, playing complex games. One landmark that had particular resonance in China was the 2016 victory of Google’s AlphaGo software over a master South Korean player of the board game Go (known as weiqi in Chinese); AlphaGo subsequently also defeated China’s top-ranked player. While chess programs have been beating human masters for years, Go is much more complex; the number of potential board positions is traditionally estimated at 10172, more than the number of atoms in the universe.

How are these feats possible? Most of what is referred to as “artificial intelligence” in the media is a subset of the field known as machine learning, and in particular a subset of machine learning called deep learning. All software works by following clearly specified instructions, known as algorithms, on how to perform a specific task. In machine learning, the algorithms are not fixed in advance, but evolve over time by building data-driven models.

Usually the type of software used in machine learning is called a neural network, because its structure is loosely inspired by the connections between neurons in a human brain. The network takes an input signal and repeatedly processes it into something more useful; what makes the learning “deep” is that there are a large number of “layers” that process the signal. The use of these techniques means that an algorithm can improve its performance of a task by repeated exposure to data. They are particularly useful where writing algorithms the traditional way—by specifying all possible details and eventualities in advance—is cumbersome or impossible. The concept of machine learning dates to the 1960s, and much of the original work underlying today’s approach of deep learning dates to the 1980s and 1990s.

More power, more data

The rapid improvement in the results of specific machine-learning applications in recent years is thus not a result of fresh theoretical breakthroughs. Rather, it has happened because advances in computing power have allowed machine-learning algorithms to run much faster, and the increased availability of very large amounts of structured data have given them much more to work with. The lesson has been that lots of processing of lots of data is required for the algorithms to be effective in finding patterns. This in itself is a sign that machine learning is not very much like human learning: humans can learn quickly from small numbers of examples, by building internal mental models. Machine learning by contrast is a massive and repetitive number-crunching exercise of building up statistical regularities.

Researchers in the field sometimes describe machine-learning algorithms as being “narrow” and “brittle.” Narrow means that an algorithm trained to solve one problem in one dataset does not develop general competencies that allow it to solve another problem in another dataset; an algorithm has to be trained separately for each problem. The Go-playing algorithm is not also capable of analyzing MRI scans. Brittle means that the algorithm only knows its dataset, and can break down if confronted with real-world situations not well represented in the data it learned from. An often-used example is facial-recognition software that is trained on databases consisting largely of photos of white men, which then fails to accurately recognize faces of black women.

But while it is important to understand that machine learning is not a magic wand, it would also not do to underestimate its potential. Machine learning is essentially a way of building better algorithms. That means it could be applied to almost any process that already uses software—which, in today’s world, is quite a lot—as well as many new processes that could not be effectively automated before. The most obvious example is self-driving cars, which can already operate in restricted contexts and could be in general use within a decade. Machine learning is already being used to spot patterns that previously required trained human expertise, such as recognizing financial fraud or early-stage cancers. Because of this broad applicability, enthusiasts call machine learning a “general purpose technology” that, like electricity a century ago, can boost productivity across every part of the economy.

Throwing resources at the problem

The key point is that machine learning has now moved from a pure research phase into a practical development phase. According to Oren Etzioni of the Allen Institute for Artificial Intelligence, all of the major recent successes in machine learning have followed the same template: apply machine-learning algorithms to a large set of carefully categorized data to solve problems in which there is a clear definition of success and failure.

All parts of this procedure are quite resource-intensive. Huge amounts of computing power are required to run the algorithms. The algorithms need huge quantities of data to find patterns. That data must also be first carefully structured and labeled so that the algorithms can draw the right conclusions—for instance, labeling pictures of objects to train an image-recognition algorithm—a process that is extremely labor-intensive. The repeated training and refining of the algorithms also requires a lot of labor by highly skilled workers, whose numbers are necessarily limited. But the reason for the excitement over artificial intelligence is that there is a now a sense that the main remaining constraints on progress are these limitations of resources—and such limitations will be solved over time.

According to its many boosters, China has all of the necessary resources to make progress in machine learning. It has large and well-funded technology companies, including publicly traded giants like Tencent, Alibaba and Baidu, but also private companies with multi-billion-dollar valuations like ByteDance, which runs a popular news app with personalized recommendations, and SenseTime, which specializes in image and facial recognition.

AI-chart1

China also has the world’s largest population of internet and mobile phone users, who are creating huge amounts of data on daily basis through their interactions with software. It has a huge population of relatively low-cost college graduates, for doing the more repetitive work of categorizing data. And it also has more top artificial-intelligence researchers than any country other than the US; indeed, many of the top Chinese in the field were educated in the US and have worked for US companies.

China also has a government that has decided that artificial intelligence is going to be the key technology of the future, and that will not accept being left behind. An ambitious national plan released in July 2017 calls for China to lead the world in artificial intelligence theory, technology and applications by 2030 (for detailed analysis of the plan, see these reports by the Paulson Institute and Oxford University’s Future of Humanity Institute). While it is difficult for government plans to create fundamental research breakthroughs on demand, such plans can be good at mobilizing lots of resources. So to the extent that advances in machine learning are now about mobilizing resources, it is reasonable to think China will indeed be able to make lots of progress.

AI-chart2

The prospect of China being something close to a peer of the US in a major new technology is a shocking development for many Americans. Everyone knows that China’s economy has grown rapidly and that it has accomplished a lot. But most of its past successes in technology involve deploying things developed elsewhere, such as mobile phones, wind turbines or high-speed trains, on a large scale. China has a per-capita GDP of roughly US$8-9,000 at market exchange rates, lower than Mexico or Turkey—and no one is talking about their dominance in the technologies of the future. It is tempting to try to resolve this paradox by focusing on China’s state support for artificial intelligence, implying that its advantages are unfair. The rivalry is also not a purely economic one, since there are military uses for machine learning.

The paradox is more apparent than real. China is such a huge, diverse and unequal country that averages are not a good guide to the location of the cutting edge. The reality, as anyone who has visited Beijing, Shanghai, or Shenzhen in recent years can attest, is that the income, skills and education levels of its best people can be comparable to those in the US. That elite of course is not representative of all the hundreds of millions of their compatriots, but neither is the Silicon Valley elite representative of middle-class Americans.

The fact that China now has the capability to contribute to cutting-edge research is also in large part a result of its integration with the US: it is the decades of sending top Chinese students to top US universities that have built up the necessary human capital. Rather than say there is a competition between the artificial intelligence sectors in the US and China, it might be more accurate to say that there is a single, global field of machine-learning research that has a significant presence in both North America (Canada also has some top people) and China.

There is no AI race

More fundamentally, it is wrong to think of China and the US as being in a “race” for supremacy in artificial intelligence. Evoking the “space race” with the Soviet Union in the 1960s is the wrong analogy. The space race was about achieving clear technical landmarks defined in advance: first satellite in orbit, first human in orbit, first human on the moon, etc. Today, it’s not clear what the technical landmarks for an artificial intelligence race might be. There is a vague goal of “general purpose artificial intelligence,” which means the kind of thinking, talking computers that are familiar from decades of portrayal in science fiction. But there is no race to make one, since no one knows how.

Rather, there are multiple related efforts going on to make progress on diverse sets of specific technical challenges and applications. If China is the first to achieve some technical breakthrough, that does not prevent the US from also doing so, nor does it guarantee that a Chinese company will control the market for applying that breakthrough. Recall that machine-learning applications can be “narrow” and “brittle”: software that is excellent at predicting, say, the video-watching habits of Chinese will not necessarily also dominate the American market. What we can say is that there are economies of scale and scope in machine-learning research: teams of experts who have successfully developed one machine-learning application themselves learn things that will make them better at developing other machine-learning applications (see this recent paper by Avi Goldfarb and Daniel Trefler for more).

Artificial intelligence is not a prize to be won, or even a single technology. Machine learning is a technique for solving problems and making software. At the moment, it is far from clear what the most commercially important use of machine learning will be. In a way, it is a solution in search of problems. China is making a big push in this area not because it knows what artificial intelligence will be able to do and wants to get there first, but because it does not know, and wants to make sure it does not lose out on the potential benefits. China’s development plan for artificial intelligence is mostly a laundry list of buzzwords and hoped-for technical breakthroughs.

The fact that machine learning is now in a resource-intensive phase does play to China’s strengths. There is an enormous amount of venture-capital money and government largesse flowing toward anything labeled “artificial intelligence” in China, and Chinese companies have had some notable successes in attracting high-profile figures in the field to join them. But fears that China will somehow monopolize the resources needed to make progress in machine learning are fanciful. After all, most of the key resources are human beings, who have minds of their own. And many of the key tools and concepts for creating machine-learning applications are in the public domain.

Will today’s advantages endure?

It is also not certain that the current resource-intensive phase of machine learning will last forever. As is usually the case when limited resources constrain development, people are trying to find ways to use fewer resources: in this case, refining machine learning so that it does not require so much human effort in categorizing data and fine-tuning algorithms. Some of the current buzzwords in the field are “unsupervised learning,” where the machine-learning algorithm is trained on raw data that is not classified or labeled, and “transfer learning,” where an algorithm that has already been trained on one dataset is repurposed onto another dataset, which requires much less data the second time around. Progress in these areas could lessen the advantages of China’s “big push” approach, though of course Chinese researchers would also benefit from them.

China’s current strength in machine learning is the result of a convergence between its own capabilities and the needs of the technology; since both are evolving, this convergence may not be a permanent one. But China’s government is correct to see the current moment as a great opportunity. China was already becoming one of the global clusters of machine-learning research even before the government decided to throw lots of subsidies at the technology. The self-reinforcing dynamics of clusters mean that today’s successes will make it easier for China to attract more machine-learning experts and companies in the future.

The biggest loser from this trend, however, is not the US, which already has well-established clusters of machine-learning research, but smaller nations who would also like to become home to such clusters. European countries, for instance, seem to be struggling to hold their own. The perception that there is a rivalry or race between the US and China ultimately derives from the fact that the two countries are rivals rather than friends. Artificial intelligence may indeed be the first example of a major cutting-edge technology whose development is led by geopolitical competitors—the US and China—rather than a group of friendly nations. The rising tensions between the US and China pose the question of whether a global artificial-intelligence field structured in this way is sustainable, or will be forced to split into national communities. The loss of those exchanges would slow progress in both countries.

Nick Lardy on the crowding-out of private investment

The Australian National University’s annual free China Update book is bigger and more interesting than usual this year, in honor of the 40th anniversary of the start of reform and opening up in 1978. It’s got contributions from lots of prominent China economists that I have only begun to work my way through.

Naturally, I immediately checked out the chapter on private sector development by Nick Lardy. It’s quite a useful update to his work on the economic weight and role of the private and state sectors, and includes a careful, data-driven assessment of the resurgence of state enterprises under Xi Jinping. Here is a section where he identifies the symptoms of the crowding-out of private investment by SOEs:

The most plausible explanation of the waning of private investment is crowding out—an explanation supported by several pieces of evidence.

First, the share of bank loans to nonfinancial corporations that went to private firms fell from 57 per cent in 2013 to only 19 per cent by 2015, while the share that went to SOEs almost doubled over the same period—from 35 per cent to 69 per cent.

Second, financing of private firms through microfinance companies stalled after 2015. Lending by these companies grew rapidly from 2008, when the People’s Bank of China and the China Securities Regulatory Commission first issued formal guidelines on microfinance companies. The volume of such lending levelled off at just less than RMB1 trillion in 2014, but has not grown since.

Third, between 2011 and 2015, SOEs’ profits rose by only RMB30 billion, or 1 percentage point, while the investment of these firms rose by almost RMB2 trillion, or more than 20 per cent. Much of the differential between the growth of investment and the growth of profits must have come from increased borrowing from banks.

Fourth, indirect evidence suggests that SOEs have borrowed increasing amounts of funds to cover their financial losses. In 2005, 50 per cent of all SOEs were lossmaking. By 2016, the share of lossmaking SOEs had declined slightly, to 45 per cent. Thus, roughly half of China’s SOEs for more than a decade have been unable to fully cover their cost of capital. Moreover, the magnitude of losses generated by lossmaking firms increased sevenfold, from RMB243 billion in 2005 to RMB1.95 trillion in 2016. As a share of GDP, these losses doubled, from 1.3 per cent in 2006 to 2.6 per cent in 2016.

The interesting question at the moment is how this crowding-out of the private sector evolves in response to the government’s campaign to rein in financial risks. Surprisingly, one of the big casualties has been investment by state entities: there’s been a sharp slowdown in infrastructure investment as the central government has tightened controls over local government fundraising. Mostly as a result, the non-state share of fixed-asset investment rose to 65% in the first half of 2018 from 63% in 2017.

But it would be unusual for tighter financial conditions to really benefit private-sector firms, which tend to be smaller and riskier borrowers than state firms. And there is indeed a great deal of official concern at the moment over small businesses losing access to credit.

(Also see this previous post for other references on the crowding-out of private-sector investment.)