AI development has morphed into a geopolitical race with China and the United States in a dead heat to be the victor. While the 20th century may be widely thought of as the American century, the 21st will be defined by the leader in this pivotal technology. At least, that’s the impression given by most commentators and echoed in Kai-Fu Lee’s recent book, AI Superpowers: China, Silicon Valley, and the New World Order.
Lee’s book fits into the common Cold War narrative, framing the development of this technology as a new space race, even going so far as calling AlphaGo’s victory over Ke Jie in 2017 China’s “Sputnik moment.” His chapters discuss the advantage of the US vs. China and vice-versa, with prognostications about who is ahead today and who will be in the coming years. For example, as far as business use of AI technology is concerned, Lee states (italics are mine):
Today, the United States enjoys a commanding lead (90-10) in this wave, but I believe in five years, China will close that gap somewhat (70-30), and the Chinese government has a better shot at putting the power of business AI to good use.
This strikes me as an incredible claim, and belies many of the problems I have with the book.
No doubt Dr. Lee provides great insight into China’s pioneering use of the technology and the cultural and political differences between the two countries. He himself is Chinese, but completed his PhD at Carnegie Mellon University, worked for Google in the US and China, and has been involved as a VC in many companies on both sides of the Pacific. His experience and expertise are valued and well received. As long as he stays within this realm, I think he provides excellent commentary. The difficulty is when he moves into political philosophy and economics, this is where these prognostications begin to flounder.
A Planned R&D Strategy
The biggest oversight Lee makes throughout the book is his belief in the efficacy of a centrally planned and funded R&D program. Time and time again, he refers to the amount of money Beijing is putting into AI research, its support of AI start-ups and so forth. The amounts are indeed staggering with billions of dollars being directed toward AI initiatives around the country with the goal of having a multi-trillions dollar AI industry by 2030. But it’s not so much the quantity of funds that is being invested as much as the quality.
Centrally planned and directed R&D work stumbles because of incentive, pricing, and knowledge problems inherent to the planners themselves. Hayek famously explicated this issue in his discussion of the knowledge problem which relies on the epistemological limitations of central planners because they are making decisions for the masses and lack the detailed, distributed knowledge of those they are making decisions on behalf of. This insight has been further developed and bolstered by critiques from public choice theory who focus on the role of special interests and incentives by the politicians and bureaucrats making funding decisions. Although we’re still in the early days of the AI revolution, we can already see these dynamics playing out.
AI R&D for the CCP, not the People
A quick glance at the largest and most well-funded AI companies in China shows that the lion’s share of the AI startup funding is going toward government focused security firms aimed at AI technologies such as facial recognition. According to data from CBInsights, 89% of the funding going to the top Chinese AI startups is devoted to companies that are pioneering facial recognition and security technologies for the CCP. This is compared to 1.4% for the top AI startups in the rest of the world. It seems clear that the Chinese funding decisions are decidedly tilted in their favor, whereas there is no market for these services elsewhere.
Another problem with the increase in AI research is the large spike in poor quality scientific publications in the field that have been emanating from Chinese universities. A 2018 study by the publisher Elsevier found that Chinese universities had published more papers than US universities from 1998–2017 in the field of AI, but when weighting by impact by accounting for citations and self-citations, the quality of Chinese papers plummeted. Additionally, Chinese research in general has suffered through a long-lasting spate of retractions of general scientific research due to fraud or faulty research. But this is what we ought to expect when non-experts (i.e., bureaucrats and politicians) direct funding in specialized areas; they artificially increase demand for researchers and incentivize more people to enter the field and produce when they otherwise would not or should not. In other words, a classic example of malinvestment.
Finally, we have seen the rise of fake AI companies, those who claim to leverage this new technology in order to attract additional investment and business. A high profile case of this occurred with iFlytek last year, a large Chinese startup that ostensibly develops AI for voice recognition, when they were accused of using humans to complete their translations rather than AI systems. This isn’t to say that other markets haven’t been affected by the rise of “fake AI.” it’s almost inevitable to see some fraud when a new technology reaches such a zenith of hype, but that this problem becomes exacerbated when those making investment and funding decisions have no skin in the game for the faulty decisions they make.
The Luddites: This Time Is Different!
Lee devotes another chapter to the issue of jobs and displacement of workers, citing (and siding with) many of the high-profile studies that have come out over the past five or so years warning for eventual automation of a large percentage of jobs. Lee does summarize the counter position well: every technological revolution in history has led to temporary displacement, but long-term increase in the standard of living for all and new and better jobs have always arisen, but fails to make a compelling case for why “this time is different.”
He relies on Brynjolfsson and McAffee’s book, The Second Machine Age, to argue that the increase in productivity over the past 30 years has led to the “great decoupling” because wages have stagnated and even declined for the poorest people in America leading to increased inequality. The reason given for this is what they refer to as the skill bias — whereby this information technology (and by extension, AI) makes high-skill labor more productive, but does little to nothing for low-skilled labor.
This argument has numerous problems that are readily apparent on just a moment of reflection. First, Lee recites the oft-debunked decoupling argument which is based on faulty statistics and happens to select a starting point at a high in the late 80s. Moreover, this number relies on ignoring benefits that are not counted by wages (which accounts for anywhere from 20-31% of compensation) and deflates using the CPI rather than the more relevant PCE which overstates inflation thus depressing growth. Moreover, it seems absurd to believe that the average worker in 2019 enjoys the same material standard of living as the average worker in 1989; everything from consumer electronics, appliances, cars, and services have greatly increased in quality and availability.
The second problem is with respect to Lee’s skill bias argument. This seems like it may have been a plausible argument in the early 2000s when the internet was still in its infancy, but as adoption grew, more tools and services became available thoroughly distributing the potential to take advantage of the technology to people around the world without the need for specialized skills. Even if this were the case then and is no longer, it still misunderstands why wages rise in the first place. Finally, Lee’s argument puts the blame for income inequality at the feet of technology when central bank policy and Cantillion effects are a far more efficacious explanation for the recent economic malaise of the West.
AI Development from the East
Overall, Lee’s book is an uneven read. The first half of the book is likely why most people (myself included) purchased it — to get a respected insider’s perspective of Chinese technological development. The second half, goes into Lee’s personal life, from his unbalanced life to his struggle with cancer and lessons looking back. This becomes his motif till the end as he focuses on human connections and the need to address potential job losses with retraining toward more communal and interpersonal work.
If you haven’t read much on China’s AI push or beyond China’s big-three (Baidu, Alibaba, and Tencent), then you’ll get a lot out of this book and what the tech landscape looks like. It truly is remarkable to step back and consider the story that Lee weaves and that the world has witnessed over the past few decades: China rising from an agricultural based economic backwater, to a knock-off manufacturer, to a competent manufacturer with rapid growth, to a technological power pushing the forefront of AI development. However, too much of the book is spent chasing economic phantoms and fails to deliver the deep insight one hopes for from someone as well regarded in the industry as Kai-Fu Lee.