Is China really that far ahead in AI? Survey says “No”

LinkedIn has joined the conversation on China’s AI industry—and it isn’t as bullish as other recent reports. Their survey of the global AI talent scene (in Chinese) reveals that in terms of personnel working in AI and the experience they have, China lags far further behind than many had imagined.

Recently, there has been a media maelstrom of China’s AI progress. Bloomberg had China’s Plan for World Domination in AI Isn’t So Crazy After All, the New York Times asked Is China Outsmarting America in A.I.?, Wired described How Baidu Will Win China’s AI Race—And, Maybe, The World’s, not to mention our own coverage including of the tireless Kaifu Lee.

While all compelling reading—and it is clear there are bright spots on the horizon for AI in China as local media has sought to point out—certain indicators suggest that prospects are bleaker than the hyperbole.

LinkedIn itself is in an unusual position in China and struggling with cultural and political issues. The core of its offering is contrary to how Chinese people conduct themselves and manage their contacts. To keep a foothold in China, the company decided to toe the government line and keep things compatible.

Yet this report was released in Chinese only (although reads as though translated from English) and was clearly aimed at the China market with all insights provided by Chinese sources. It appears to have paid off: it was picked up by Chinese media which highlighted the industry insiders who contributed.

According to LinkedIn’s survey and report, half of China’s biggest AI employers are in fact US firms and many Chinese talents are in academia rather than industry (for now). China has “50,000+” personnel in AI, of a global total of around 1.9 million. The US has over 850,000. India has the next largest talent pool with over 150,000 AI workers, the UK followed with 140,000+ then Canada with over 80,000. China may come in fourth, but it’s a joint fourth place alongside Australia and France. Italy, Germany and the Netherlands claim joint fifth with over 30,000 each.

The survey took a broad approach to AI, covering deep learning, voice recognition, autonomous driving, natural language processing, but a narrow approach to the roles it counts. Only the technical roles, such as engineers, are included, rather than other divisions within tech firms such as marketing.

43.9% of China’s AI talent come from the US, many of whom are returning Chinese or US citizens of Chinese descent (海外华人). The UK is the next largest source at 15.3% of arrivals. The survey does not indicate overall numbers of people arriving to work in AI in China. Making AI more of a national or patriotic concern, Baidu has recently announced a global AI training scholarship program (in Chinese), but only for people of Chinese origin (华人) around the world.

“Due to policies and other factors and the influence of the development of the domestic market, those Chinese talents who find themselves working in Silicon Valley feel ever stronger the high tech force coming from China and throw themselves one after another into the great tide returning to China,” says the LinkedIn report somewhat emotively.

Naturally, this is not just a quantities game. Yet in terms of personnel quality across measures such as experience, international exposure, and level of education, China has significant catching up to do, too. Chinese AI talents are less experienced and have less international exposure. The report states that 9% of people working in AI in China have experience from abroad, whereas in the US that figure is 11.1%.

China’s AI talent is younger than the US and global averages. 38.7% of those working in China’s AI sector have over ten years’ experience compared to 71.5% for the US and 65.4% globally. Scaling this up—71.5% of 850,000 vs 38.7% of 50,000—and a huge gulf in terms of man hours of experience opens up between China and the US.

LinkedIn also provides the educational level of AI workers in the US and China, which are more equal. However, a look into academia reveals a difference. All around the world students progress through degree programs with some staying on in academia. Given the opportunities that academia provides for basic research and to train others, the role played by academic institutions is crucial to the overall AI sector. In the US, up to 2016, 26.7% of the current AI workforce had previously worked in higher education or research centers; in China, just 10.7%. Those in academia are being lured into China’s ravenous AI industry, which could potentially have a damaging effect on the teaching and research of AI in the country.

This is where some of the local media took the story. The People’s Daily ran an article based on the study (though led with its own statistics from last year to generate a headline of “Chinas AI Talent Shortfall of 5 Million—Supply to Demand Barely 1:10” in Chinese). It gives examples of the money being thrown at AI talent in China. However, it cites people in the industry as saying that too many AI experts remain in academia in China, rather than moving into the industry side. The article suggests that returning Chinese will help China close the gap with the US and the nation’s big data supply will beget big numbers of engineers. 

Further details are brought to life by the report. For example, the geographical spread of AI talent across the US and China. Maps show how talent is concentrated in fewer parts of China. Beijing is often called the Silicon Valley of China, but that’s perhaps not going far enough. The San Francisco Bay area is host to 17.2% of US AI workers—still the greatest concentration in the country, while Beijing is home to 34.1% of China’s and Shanghai to 33.7%.

Lining up the top ten AI employers in terms of the number of staff in AI roles in the two countries reveals that while the US top ten are all homegrown companies, the China line up sees Microsoft in fourth place, IBM in sixth, Intel in seventh and AMD and HP in ninth and tenth.

The research also provided the proportions of Chinese people (华人, ethnic Chinese not necessarily from China) working in other parts of the world. 7.9% of AI talent in the US is of Chinese origin which works out at over 67,000—more than the 50,000-plus figure in China. They form 29.4% of Singapore’s AI workforce, 2.6% of the UK’s and 1% of Russia’s.

While China is making huge strides in the field, LinkedIn’s report provides a useful pause in the fervor. While the whole point of AI is that it relies on algorithms, human engineers are still the most vital component. No wonder Silicon Valley companies are resorting to hiring AI professors to then capture their students before anyone else does.