China’s AI Talent Base Is Growing, and then Leaving

When China decides that it wants to establish leadership in a particular strategic area, its approach has tended to follow the mantra “the more, the better.” So, too, has the pursuit of strengthening artificial intelligence (AI) talent followed this principle.

AI talent is usually cultivated within universities and their extended research ecosystem. And in China, shifts in college majors are often perceived as an indicator of national priorities and resource allocation. So when Beijing deemed AI a “special discipline” as far back as 2012, dozens of universities rushed to set up their own AI specialization and degree programs, attracting thousands of students.

Credit where credit is due. China has been successful in producing AI talent, evidenced by the rapid growth of AI human capital over the last decade. But talent acquisition is only one part of the puzzle—equally important is retaining that talent so they contribute to China’s AI aspirations over the long term. On the retention front, however, China has not done nearly as well.

Figure 1. Exponential Growth of China’s AI Talent BaseSource: NeurIPS and author calculations.

China’s AI talent has grown about ten-fold over the last decade. In 2009, only about 100 Chinese AI scientists (defined as those who did their undergraduate studies in China) were accepted to the NeurIPS conference. A decade later, nearly 1,000 were accepted, making up about a quarter of the total amount of authors accepted to the conference.

The focus on the NeurIPS is because it is considered one of the most prestigious, so those accepted to the conference represent the pool of top global AI talent, defined as “upper tier.” In the past ten years, a total of 15,616 AI scientists were accepted to the conference, and nearly 2,800 of them (~18%) were Chinese.

The explosive growth in China’s AI talent base has been accommodated by a flurry of second-tier universities that have gotten into the AI game. Whereas AI talent used to be concentrated in a handful of elite Chinese universities, such as Tsinghua University and Peking University, it has now spread into institutions like the University of Science and Technology of China, Nanjing University, Beijing Institute of Technology, Wuhan University, and Dalian University of Technology, to name just a few.

Figure 2. Current Location of China-educated AI ScientistsSource: NeurIPS and author calculations.

When it comes to retention, of the 2,800 Chinese accepted to NeurIPS over the last decade, about three-quarters of them (>2,000) are currently working outside of China, according to my estimates (see Figure 2). And of the 2,000+ AI talent that have left China, about 1,700 (85%) came to the United States. Some currently work for US tech giants such as Google and IBM, while others hold professorships at top institutions such as UCLA and University of Illinois-Urbana Champaign.

So while Beijing has cultivated an army of top AI talent, well over half of that talent eventually ended up in America rather than getting hired by domestic companies and institutions. That’s because most of the government resources went into expanding the talent base rather than creating incentives and an environment in which they stay.

Beijing seems to have recognized its failure in retaining talent. The well-known New Generation Artificial Intelligence Development Plan, released in 2017, vowed to lure top-notch AI scientists in neural network, machine learning, self-driving cars, and intelligent robotics by opening up special channels and offering up competitive compensation packages.

Still, it’s not clear that Beijing will be able to reverse the Chinese AI brains from draining to its biggest competitor, the United States. In the meantime, it is up to the United States to decide whether it should keep that pool of talent.

Notes

This analysis is based on the NeurIPS conference authors database I created at MacroPolo. All estimates are based on a sample that follows hypergeometric distribution. For the total number of China-educated AI talent, the estimated value of 2,788 on average (+/-10%) is at a 95% confidence level. For the number of China-educated AI talent who are currently working outside of China, the estimated value of 2,054 on average (+/-19%) is at a 95% confidence level. For the number of China-educated AI talent currently working in the United States, the estimated value of 1,712 on average (+/-23%) is at a 95% confidence level.