31 A Dark Horse Comes Out of Nowhere
In 2017 the Chinese government officially announced its goal to achieve primacy as a center of AI innovation by 2030. That same year, a program called AlphaGo enjoyed a decisive victory over the best human Go player living, a nineteen-year-old Chinese man named Ke Jie. To Chinese people in Beijing’s equivalent of Silicon Valley, this match was both an inspiration and a challenge, what Kai-Fu Lee, a leading AI researcher and venture capitalist, calls “China’s Sputnik Moment.”
In the winter of 2018, a number of Western news sources reported on this Chinese plan. For example, on February 8, 2018, the journal Science, in an article ominously titled “China’s massive investment in artificial intelligence has an insidious downside,” announced to English language readers the new Chinese initiative in detail. China was investing heavily in all aspects of information technology, from quantum computing to chip design. “AI stands on top of all these things,” Raj Reddy was quoted as saying. He’s right.
The article went on to praise the Chen brothers, whose AI chip, backed by the Institute of Computing Technology of the Chinese Academy of Sciences, has been the foundation of a new company, called Cambricon, already worth $1 billion dollars by 2017. China’s State Council forecast that by 2030 the country’s AI industry could be worth $150 billion. (We’d later learn that another special chip designed in China had allowed the Chinese to hack into major Western organizations.)
“China’s advantages in AI go beyond government commitment,” the Science article went on. Because of China’s sheer size, vibrant online commerce and social networks, and scant privacy protections, the country is awash in data, the lifeblood of deep learning systems. Chen Yunji, the co-designer with his brother of a chip that can equal the performance of the 16,000 microprocessors Google Brain once needed to learn to identify a cat, also said that because AI is a young field, China benefits; AI’s relative newness has encouraged “a burgeoning academic effort that has put China within striking distance of the United States” (Larson, 2018).
Chen’s claim to a “burgeoning academic effort” in China is somewhat undermined by the reality that AI companies seek talent, offering salaries that no university can match. But the United States faces the same challenge, and some universities have met it by releasing American academics on semester- or year-long furloughs to make money and then welcoming them back to teach and do basic research. The Trump administration’s anti-immigration policies have worsened the U.S. situation, sending promising students to Canada and Europe instead of the United States, which had traditionally welcomed and trained them (and often retained them).
China uses its AI prowess in ways that make Westerners deeply uncomfortable, although Western governments also employ such practices: surveillance (especially fine-grained facial recognition), censorship, battlefield decision-making, and autonomous weapons. This is part of the insidious downside the Science article refers to. But at least in the West, civil liberties organizations, such as the American Civil Liberties Union, protest by bringing suits against such U. S. government deployments of AI. Myriad organizations continuously examine AI applications and their ethics; and Europeans have gone even further to control social media with, for example, rules about the individual’s right to be forgotten, or to have personal data deleted from public view. In Silicon Valley itself, valued employees are organizing to ask questions about the ethics of where their work might lead. Even a Chinese Internet conference in Wuzhen in 2018 featured sessions that grappled with unintended consequences of AI deployment: counterterrorism, data breaches, surveillance, issues around private enterprise (especially WeChat), and complicity with government intrusion (Zhong, 2018).
Kai-Fu Lee is an outstanding AI expert who emigrated from Taiwan to the United States without a word of English when he was eleven, graduated with honors from Columbia University, and received his PhD at Carnegie Mellon under Raj Reddy. His doctoral dissertation was called Sphinx, the first large-vocabulary, speaker-independent continuous voice recognition program. It was so extraordinary that he had to publish his source code to convince others that the program was what it claimed to be. Lee has had years of experience in American firms such as Apple, Microsoft, and Google, and is now a venture capitalist based in Beijing, with investments in both China and the United States. His lifelong experience in both the East and West makes him an especially perceptive observer. 
In his 2018 book, AI Superpowers: China, Silicon Valley, and the New World Order, Lee describes the contrast between the victory of IBM’s Deep Blue over the world’s chess champion, Garry Kasparov, with the victory of a Google subsidiary, DeepMind’s AlphaGo over the human Go champion, Ke Jie. Chess had been a brute force win, with specialized hardware and software applicable only to chess. The win was interesting, Lee says, but made little difference to the real world.
For the game of Go, brute force was useless. Go is so complex that to play it, let alone win, seems to require unique human intuition, human art. Thus when the program AlphaGo (later called AlphaZero) decisively won three out of four games against a nineteen-year-old human champion, Ke Jie, it had a real-world effect.
The machine’s victory seized Chinese souls. Go was their own great game, one of the “four arts” all ancient Chinese scholars were expected to master, and the victory began what Lee calls an “AI frenzy” in China. To be sure, this was a moment of AI implementation rather than discovery, the frenzied apps built on earlier fundamental research done in the West. China’s venture capitalists also responded to their government’s challenge, and in 2017 were responsible for 48 percent of global AI venture capital funding, for the first time surpassing the United States.
It doesn’t matter that the great AI research breakthroughs came from North America, the United States, and Canada, Lee says. Those fundamental breakthroughs (deep learning, for example) have offered China the chance to dream up apps that build on this research in unexpected ways. Yes, the Chinese can be accused of being copycats (even copykittens, as Lee jokes) but they know how to read their market, they’re supple, they’re local, their data is massive (“China is to data as Saudi Arabia has been to oil”) and they make the Silicon Valley work ethic look slothful. Consider that in 2013, China had only two of the world’s largest publicly traded tech companies, whereas the United States had nine. But by 2018, five years later, China had nine of the top twenty, and the United States eleven. Twenty years ago, China had none (Friedman, 2018).
Lee’s tales of the Chinese entrepreneur wars make a gripping read—he compares them to the bloody gladiatorial combats in the coliseum, battles to the death, win—or die. Meanwhile, the Chinese government, seeing how vital AI is to the future Chinese economy, is “putting fingers on the scale”—that is, offering subsidies to venture capitalists and other promising enablers of AI apps.
Thus, without legacy systems (such as credit cards) to impede it, the Chinese Internet adapted quickly to mobile phones. People who couldn’t afford desktops or laptops could easily acquire a cheap mobile phone, their introduction to life online.
The Chinese app WeChat, however, wanted to move beyond online and reach into people’s offline lives. WeChat has thus become a super-app, a “remote control for life,” that dominates not just online apps, but allows users “to pay at restaurants, hail taxis, unlock shared bikes, manage investments, book doctors’ appointments, and have those doctors’ prescriptions delivered to your door.” Aside from contributing yet more data that AI algorithms can work on, it has blurred the distinction in China between online and real life (Lee, 2018).
Again, the Chinese government plays a big role. In its 2017 declaration aiming to achieve AI primacy by 2030, the central Chinese government laid out major economic goals for AI. (The Obama government, Lee points out, had issued a similar policy paper a few months earlier, but had the misfortune to release it the same week that presidential candidate Donald Trump’s Access Hollywood tape came to public notice. Later, President Trump would propose to cut funds for AI research, but the Pentagon wasn’t having any of that, and in fact Trump would eventually agree that AI research must be supported in the United States.
A more worrisome Chinese penetration is investment—the Chinese government is heavily invested in Silicon Valley ($35 billion over the last decade) buying startups to own their novel ideas. In 2018 the U.S. Congress passed legislation that expands government oversight on any foreign investments in “emerging technologies,” and the power to block deals if they’re considered unfavorable to domestic security (Canon, 2018).
The central Chinese government set major national AI goals, but implementation details were left to provincial and municipal levels of government. Thus the original Avenue of the Entrepreneurs, near both Peking and Tsinghua Universities, became a model for cities all over China, helped along by grants and subsidies from the central government, followed by private capital, itself encouraged by government policy. Lee notes that in 2009, when he founded his venture capital firm, Sinovation, manufacturing and real estate dominated Chinese investing. By 2014, venture capital in AI quadrupled to $12 million, and then doubled again the following year (Lee, 2018).
China has also responded by beginning to construct an entire new city, Xiong’an, sixty miles south of Beijing, a “showcase city for technological progress and environmental sustainability,” expected to reach a population of 2.5 million. It’s built specifically to accommodate autonomous vehicles and environmental protection, with AI embedded in every nook and cranny. At a presentation at New York City’s Asia Society in October 2018, Lee said that, after his book had gone to press, the city of Suzhou announced plans to rebuild a section of the ancient city, a two-level grid of streets where autonomous vehicles will be confined to the lower level, and human-driven vehicles, including bicycles, and pedestrians, will be on the top level. Lee admits that other technologically themed cities in China haven’t always succeeded, but many have, so how Xiong’an will fare is an open question.
The Chinese government’s system of encouraging investments is intricate, but largely successful. Inefficient, scoff American investors. But effective, say the facts. “When the long-term upside is so monumental, overpaying in the short-term can be the right thing to do,” Lee writes. “The Chinese government wanted to engineer a fundamental shift in the Chinese economy, from manufacturing-led growth to innovation-led growth, and it wanted to do that in a hurry.”
Lee’s sharp-witted descriptions of the social changes AI has brought about in China are compelling. Since I’d lived through decades of the West’s general sniffiness about computing in general and its resistance to AI in particular, I was astounded by the reactions of ordinary Chinese to this new science and its technology.
Almost overnight, great skepticism turned to avid fanaticism. Lee describes how difficult in the beginning it had been to recruit good minds for the startups funded by his venture capital firm, Sinovation Ventures, because a general view had long prevailed: one’s children and one’s spouse should aim for a lifetime job, an iron rice bowl, with the government. But once the government blessed AI startups, Lee found people knocking down Sinovation’s door—literally, in one case—for the chance to work with him. “…Scrappy high school dropouts, brilliant graduates of top universities, former Facebook engineers, and more than a few people in questionable mental states.”
The O2O—Online to Offline—Revolution in China was underway. It would turn online actions into offline services. E-commerce would make real-world services as convenient as things that arrived in boxes: hot food, a haircut, and a ride (the latter modeled on Uber, but done better for the Chinese, which drove Uber out of China and has become its rival in other countries). The list of services is awe-inspiring, and after the initial boom and bust (for the gladiatorial combat took place in O2O too), the urban service sector in China has been reshaped. WeChat, the super-app, offered a one-stop place to activate these services, in contrast to the constellation of apps that prevails in the United States.
Similarly, online services in China bundle related services, an approach Lee calls “going heavy.” For example, whereas the U.S. model for apps “goes light,” offering a single service such as handling restaurant orders (but leaving deliveries to the individual restaurants), the Chinese equivalent of Yelp not only rates restaurants but handles orders, delivers them, and is buying up gas stations and mo-ped repair shops. The Chinese equivalent of Airbnb lists homes, but also manages rental properties and handles the work of cleaning, stocking, and installing smart locks on each property. The long-run advantages of “going heavy” are in the data this yields about users’ consumption patterns and personal habits. Mobile payments, of negligible cost to merchant and customer, turn a data edge into a commanding lead. Data is the fuel of machine learning, the present boisterous star of AI: the more data, the more usefully the algorithms can work.
Advantage China? Maybe. Lee readily concedes that another breakthrough in AI, on the scale of deep learning, will change the game all over again, and it’s likely that such a breakthrough will come from the freewheeling West, rather than the implementing East. But such breakthroughs usually occur only every few decades. (After deep learning was invented, nearly three decades passed before sufficient computing power arrived to make it useful.) Meanwhile, the myriad implementations based on past breakthroughs are led by China, improving on those apps by dogged trial and error, and informed by the vast data offered by the Chinese population’s behavior. Not to mention old-fashioned spyware in a new-fangled form.
Although controversy surrounds the event, in October 2018, it was disclosed that a spy chip, barely the size of a grain of rice, somehow inserted by operatives in the People’s Liberation Army, had been discovered on motherboards manufactured in China. The spy chip had evaded detection for some years and affected nearly thirty U.S. companies, including Apple, Amazon, a major bank, and undisclosed government agencies. Probably China’s goal was access to high-value corporate and government secrets. For the record, Apple and Amazon called the story “erroneous,” but experts guessed that the details passed the sniff test. The China-U.S. confrontation is much more than a friendly competition between commercial rivals, as Lee presents it (Robertson & Riley, 2018).
Yet Western observers have already begun to wonder aloud if the Chinese aren’t on to something. Could the orthodox free-market ideology that has prevailed—indeed, achieved near cult status in the United States for decades—have its drawbacks? “I applaud the Chinese Government for supporting science and technology,” Yasheng Huang, a professor of international management at MIT’s Sloan School of Management says. “The U.S. should be doing that too” (Elstrom, Gao, & Pi, 2018). David Hoffman, the director of Intel’s AI policy, talks about the development of an AI ecosystem. “One approach to that is, the market is just going to take care of that and develop that over time. Most other countries are saying, well, even if that is the case, we want to invest and to provide direction” (Jamrisko & Torres, 2018).
When I hear Silicon Valley libertarians bang on about how they want the government out of their businesses and their lives, I wonder at their ignorance of their own history. Without long, steady investment in both the Internet (which began as a military communications system) and in AI, there would be no Silicon Valley. The Defense Department was investing with discrimination in these technologies for much longer than any private investor would have tolerated with so little to show. The humans who made these investments on behalf of the Defense Department—on behalf of the American public—were visionaries and empiricists, not slaves of ideologies.
American reactions to China’s ultra-fast rise in AI competence were predictable. It would be us vs. them, a clash of the AI Superpowers. Because some of this rivalry was arising just as the Trump administration was blaming China for everything including Original Sin, the rhetoric got heated. “Who will set the key rules of the global order in the 21st century?” Thomas J. Friedman asked, America, “the world’s long-dominant economic and military superpower,” or China, “its rising rival?” (2018).
“Nations are seeking to harness AI advances for surveillance and censorship, and for military purposes,” wrote Christina Larson in Science (2018). Larson quoted several who fear this Chinese government investment in AI is less about delivering hot meals and haircuts and more about staying in power and stifling dissent. In The San Francisco Chronicle, war college instructor and retired U.S. Marine Thomas C. Linn (2018) writes:
China is using artificial intelligence to build an Orwellian state. Smart cities track peoples’ movements. China, netted with millions of cameras and facial and vehicle recognition systems, can rapidly identify individuals. Police wear facial recognition glasses that do the same. Biometric data provide even better identification. And people get social credit scores, which determine eligibility for loans, travel and more. This artificial-intelligence-enabled system enables political repression and strengthens autocratic rule.
All true. All distressing.
“China is reversing the commonly held vision of technology as a great democratizer, bringing more people freedom and connecting them to the world. In China, it has brought control” (Mozur, 2018). An experimental program in China even tracked the facial characteristics of tenth-grade students in a Hangzhou high school to detect their moods. “Educators in China have been sharply critical of the Hangzhou school, not only for invading students’ privacy—neither they nor their parents were asked for consent—but for charging ahead with an unproven system that purports to improve student performance” (Lee, 2018). That program was suspended, at least temporarily.
These are signs of a very different worldview from the Western ideal. “Today, few would confidently declare that the Chinese Communist party is on the wrong side of history,” says Yuval Noah Harari, the Israeli historian and public intellectual, in his 21 Lessons for the 21st Century (2018). Yet within 48 hours of the evening I heard Kai-Fu Lee paint a relatively benign picture of friendly if fierce commercial competition between the innovating West and the implementing East, factions in the Chinese government—“unharmonious voices”—were reported to be condemning the private enterprise that has brought China to such economic prominence, and supporting instead state-owned and -controlled enterprises, a return to old Marxist times now that prosperity has been achieved (Yuan, 2018). Those with long memories recall the Cultural Revolution and shiver.
A November 2018 report from Freedom House, the democratic watchdog organization, says the Chinese are exporting digital authoritarianism. Thirty-eight countries have installed large-scale telecommunications equipment from Chinese companies, allowing those countries to track citizens’ everyday movements the way China tracks and controls its citizens, and furthermore allowing the Chinese to spy on the countries that have installed these systems. China even sponsors training for its international governmental customers in methods to control dissent and manipulate online opinions. The report cites the example of Uigars in western China, tracked by the Chinese government and sent to “re-education camps” (Abramowitz & Chertoff, 2018). But the Chinese and their government customers also must deal from time to time with the Dionysian, despite Appollonian rigidity. Joe and I were eye-witnesses and in personal danger during the decidedly Dionysian 1989 massacre in Tiananmen Square.
As China continues to develop systems that enable its authoritarianism, we could have three major internets, an outcome lamented in a lead editorial in The New York Times (Editorial Board, 2018). Eric Schmidt, Google’s former chief executive, predicts that the global Internet will split into two within the next decade, a Web led by the United States and another by China, that one with fewer freedoms and greater censorship. Tim Berners-Lee, the inventor of the World Wide Web, thinks Europe should build its own Internet, protecting privacy and intellectual property in ways neither the United States nor China does.
In AI Superpowers, Kai-Fu Lee is worried not about China versus the United States, but about global problems. He believes that the United States has a great lead in innovations and China in applications and that the two will complement each other in AI for some time to come. But information and communications technologies differ from former disruptive technologies. The steam engine and electricity led to the loss of skills. The tasks of highly skilled master weavers, for example, were decomposed, and machines operated by much less skilled workers took their place. This change was hard on the master weavers (and transformed the life of the son of one of them, Andrew Carnegie) but raised a whole population of the unskilled into gainful employment.
With information and communications technologies (ICT), however, the results are more ambiguous, Lee observes. Worker productivity has steadily increased over the last thirty years, but those gains have not translated into wage or employment gains. Instead, we see increasing economic stratification; in the United States, the economic gains of ICT go to the top one percent of the population. ICT is often, though not always, biased in favor of high-skilled workers. “By breaking down the barriers to disseminating information, ICT empowers the world’s top knowledge workers and undercuts the economic role of many in the middle” (Lee, 2018).
This presents not technological but staggering social and political problems. Kai-Fu Lee and many others believe the AI revolution will be on the scale of the Industrial Revolution, but probably larger. We know it will be faster. AI will invade and enhance both muscle power and cognitive power, outperforming humans at many such tasks. But it will not ease the lot of the unskilled. It will take over tasks that, using data, can be optimized and tasks that don’t require human interaction. New jobs will be created, but probably not enough to make up for all lost jobs. Displaced workers can theoretically retrain for new jobs in fields that are difficult to automate, but this is highly disruptive and time-consuming (and training is so far largely in fields that are poorly paid).
Lee (2018) goes on to say that algorithms that perform white-collar work can be improved and disseminated quickly and cheaply, unlike the improvements that took place during the first two Industrial Revolutions in the 17th and 19th centuries and that were only fitfully adopted. He also argues that the presence of venture capital (VC) has changed the chancy patchwork of capital (private wealth, patronage, bank loans) that the first two revolutions relied upon. Instead VC numbers tell another story: global venture funding invested $148 billion in 2017, and AI start-ups accounted for $15.2 billion, a 141 percent increase over 2016. VCs will continue to seek every profit they can out of every appealing idea that AI researchers propose. AI is the first disruptive technology where China, a fifth of the world’s population, equals the West, both in advancing and applying the technology. China’s participation will accelerate AI.
Although Lee’s book examines the effect of AI on jobs in persuasive detail, his biggest concern is the effects of the two AI superpowers, China and the United States, upon the rest of the world, driving an all but insuperable wedge, if AI is left unchecked, between the haves and the have-nots. AI is an inequality machine. Developing countries are losing the great, perhaps only advantage they’ve had: cheap labor. Put bluntly, China and the United States are going to divide up the world between them, even as the Pope once divided the world between Spain and Portugal, except this time it’s for real.
Kai-Fu Lee’s proposed solutions to AI’s expected social impact are shaped by his own brush with mortality and deserve a reading in their own right. He proposes a fundamental rewriting of the social contract that rewards socially productive activities the same way the industrial economy rewarded economically productive activities. His specifics offer one concrete answer to my own vague longings that the AI bounty be fairly shared. Surely others will be imagined; if we’re intelligent, carried out.
We face a new world, including a potential new conflict between two nation-state adversaries who wield power of colossal potency, a kind of power that has never before been seen or used on a global scale. This power could nullify past weapons of wars. The conflicts to come are economic and geopolitical, but also philosophical, and even spiritual.
- And disingenuous: when Kai-Fu Lee praises the face-recognition software the Chinese government employs, he tells Western audiences that if it were used at airports, terrorists could never get on planes. But Westerners are uncomfortably aware how authoritarian governments, East and West, can abuse such software. ↵
- I first visited Souzhou in 1981, where colorful boxes outside each pretty canal-side dwelling contained night soil to be collected as fertilizer for the surrounding countryside. Twenty years later, on another visit to a completely modernized Souzhou, I mentioned to my guide, a former mayor of the city and now a high-ranking national official, how much more picturesque I’d found the old city. He replied irritably: do you really think that was a superior way for ordinary people to live? No, I didn’t. He was right to be irritated. ↵