Survey of the world's top AI talents: the most elite prefer employment in the United States, and China exports the most
Source: pixabay
Written by丨Zhang Tianqi
Those who win talents win the world.
In the field of artificial intelligence that affects humans, talent, especially top AI humans, is one of the keys to winning this battle.
Recently, the Paulson Institute's MacroPolo think tank released a survey called the Global AI Talent Tracker, which maps the flow of top AI talent and answers where top AI talent comes from and where it's going.
The McCropolo think tank leveraged author data from the Conference on Neural Information Processing Systems (NeurIPS). NeurIPS is recognized as one of the top conferences in the field of AI. The research published in NeurIPS focuses specifically on theoretical advances in neural networks and deep learning, two subfields that have driven many of the latest advances in artificial intelligence.
At the 2022 NeurIPS conference, 2,671 papers were accepted, with an acceptance rate of 25.6%. The acceptance rate was even lower in 2019 at 21.6%. The McCropolo think tank regards the authors of these included papers as the top talents in the field of AI research, accounting for roughly the top 20% of overall AI research talent.
At the NeurIPS conference, the opportunity to get an oral presentation is even more valuable. Only 190 papers received this honor in 2022, and the acceptance rate was only 1.8%. A total of 955 authors of these papers have been identified by the McCropolo think tank as the most elite (about 2%) members of AI talent.
Due to the large number of papers accepted at the conference, the McCropolo think tank randomly selected 186 papers for analysis, and the total number of authors of these papers was 867. Based on the authors' academic experience and work institutions, the McCropolo think tank summarized the characteristics and trends of top AI talent flow [1].
In 2020, the McCropolo think tank released data from a "Global AI Talent Tracker", which is an update of the database. Comparing the data of the two times, we can see the change in the trend of talent flow.
The most elite AI talent is preferred to work in the United States, followed by China
According to the data, China and the United States are the main sources and target places of top AI talent, with 70% of top AI talent working in institutions in China or the United States, and 65% of top AI talent coming from China and the United States. While other countries have their own characteristics, China and the United States still firmly dominate AI talent.
In the competition between China and the United States, the United States has a clear lead in top AI talent. The United States is home to 60% of the world's top AI research institutions, and is the preferred employment destination for the world's most elite (top 2%) AI talents, and 57% of the most elite AI talents prefer to work in the United States. China is second only to the United States and is the biggest competitor of the United States, but only 12% of the most elite AI talent prefers to work in China, and the gap is very stark.
The main countries where the most elite AI researchers work
This advantage of the United States is largely due to the strength of its leading universities, especially graduate education, and Silicon Valley's strong talent magnet. World-class universities such as Stanford, MIT, and the University of California, Berkeley, as well as tech giants such as Google and OpenAI, and an active startup ecosystem, put the United States in a strong position to attract and develop top AI talent.
In fact, the U.S. is quite reliant on STEM fields and also makes good use of foreign talent, which is based on the attractiveness of the U.S. graduate education system for foreign students. Of the 33,759 STEM PhDs graduating in the U.S. in 2019, 39% were foreign-born. The U.S. graduates almost as many foreign STEM PhDs each year as Germany does.
China is the world's largest exporter of top AI talent
The advantages of graduate education in the United States are fully reflected in the distribution of talent in the field of AI. China is the world's largest exporter of top AI talent, and the United States is the biggest competitor in terms of the quantity and quality of talent. The top (top 20%) AI talent with undergraduate education in China accounts for 47% of the world's talent, while only 18% of those with undergraduate education in the U.S.
However, at the graduate level, a large number of Chinese AI talents have flowed to the United States, and nearly 40% of Chinese AI talents choose to study in the United States, reversing the ratio of AI talents in China and the United States. After earning a Ph.D. in the U.S., 77% of non-U.S. students chose to stay and work in the U.S.
Among the top AI research institutions in the United States, there are even more AI talents from China than from the United States. Three-quarters of the top AI talent in these institutions comes from the United States and China, with the United States and China accounting for 37% and 38% respectively, with a slightly higher proportion of Chinese talent. On the other hand, the source of China's top AI talents is much simpler, with 75% of the talents absorbed receiving undergraduate and graduate education in China, not only are there few foreign talents, but even only about 10% of those who have studied in the United States at the graduate level [2].
From left to right, the movement of top AI talent from study to where they worked, from left to right, is where they studied at undergraduate, where they went to graduate school, and where they worked after graduation
Chinese AI talent is active in the United States. For example, about 20% of the core contributors to the GPT-4 team are from China. In terms of professional fields, the highest proportion of Chinese talents on the list of GPT-4 core contributors is in the field of computer vision, with 30% of the members coming from China. In GPT4's full list of contributors, 32 are from China, and the Chinese AI talent counted here does not include foreign-born Chinese scientists.
Of the 32 Chinese AI talents who contributed to GPT-4, 11 completed their undergraduate studies in China and 21 completed their undergraduate studies in the United States. At the graduate level, nearly eighty percent of these talents studied in the United States and stayed in the United States. It can be said that receiving undergraduate education in China, studying and working for a PhD in the United States, and publishing top AI research are the portraits of a generation of Chinese AI talents.
The GPT-4 team is not unique in this regard. In the field of computer science specifically, Chinese graduate students account for 14% of the total number of computer science graduate students in the United States and more than a quarter of the total number of international students studying in the United States. Among them, 90% of Chinese graduate students choose to stay in the United States to develop their careers [3].
For the United States, the influx of Chinese AI talents has contributed to the rapid development and innovation of the United States in the field of AI. For China, these talents, with graduate education at top U.S. universities and work experience at the world's leading technology companies, are what the government and top tech companies want to strive for. Their return will be of great help to China's AI education and industry.
The gap between China and the United States is narrowing
In fact, China has made rapid progress in cultivating and attracting AI talent in recent years. According to the McCropolo think tank, the proportion of the world's top AI researchers trained by China rose to 47% in 2022, a significant increase from 29% in 2019. AI talent from China has also improved qualitatively, with 26% of the most elite (top 2%) AI talent now coming from China and 28% in the United States, which is very close. In the last survey, only 10% of these most elite AI talents came from China.
Country of origin of top AI researchers (top 20%, based on undergraduate degree)
In terms of job selection, in the last survey, only 11% of top AI talents (the top 20%) chose to work in China, while the United States hired nearly 60% of the world's top AI talents. According to the survey, 28% of top AI talent now choose to work in China and 42% in the United States. It can be seen that in recent years, the gap between China and the United States has narrowed significantly, and the choice of AI talents' work location has changed from the dominance of the United States to the situation of competition between China and the United States.
Top countries where top AI researchers (top 20%) work
This is largely due to the growing demand for talents brought about by the rapid development of China's AI industry, as well as the increasing attention of domestic universities and technology companies in the cultivation and introduction of AI talents.
For example, in terms of research institutions, Tsinghua University and Peking University were the only two Chinese research institutions in the last list of the top 25 AI research institutions listed by the McCropolo think tank (based on the affiliation of the authors). But in this year's statistics, Chinese research institutions occupy six seats, Shanghai Jiao Tong University, Zhejiang University, Chinese Academy of Sciences and Huawei also make it into the top 25, and Tsinghua University and Peking University are ranked third and sixth, respectively.
More AI talent is choosing to pursue graduate education in China rather than in the U.S., with a 2020 report by the McCroplo Intelligence Institute showing that about 56% of Chinese AI researchers who receive undergraduate education go to the U.S. for further study. In this year's report, the figure is only about 36 percent. The close connection between the country where the postgraduate education is received and the place where the future will work also means that more talent will work in China in the future.
This may also be related to the inconvenience of studying abroad during the pandemic, as well as the restrictions imposed by the United States on studying abroad in STEM for China. From 2019 to 2022, the total number of Chinese students studying computer science and mathematics in the United States fell by more than 10,000. Chen Xi, co-founder and CEO of AI robotics company Covariant, has said in an interview that further restrictions by the U.S. government could hurt the entire field, which will definitely affect the company's ability to recruit talent, and if it continues, his company may face a dilemma of not being able to recruit people [4].
But the problem cannot be blamed solely on Covid hindrances and U.S. restrictions on studying in STEM fields in China, as India has seen a similar trend of returning talent. While India remains a significant exporter of top AI talent, its ability to retain talent is growing. In 2019, almost all Indian AI researchers (according to undergraduate degree statistics) chose to develop overseas, but by 2022, one in five talents had stayed in India to work.
The trend in China and India seems to reflect the overall declining mobility of top AI researchers over the past few years. In 2022, only 42% of top AI talent was foreigners working in other countries, down 13 percentage points from 2019, meaning more talent chose to stay in their home countries. Although the United States still has a clear advantage in top AI talent, countries are also accumulating their own talent strength to impact the dominant position of the United States.
Bibliography:
[1] Methodology for Global AI Talent Tracker (2023) - MacroPolo. (2024, March 6). MacroPolo.
[2] The Global AI Talent Tracker 2.0 - MacroPolo. (2024, March 6). MacroPolo.
[3] Cortese, A. (2024, January 2). The Chinese Talent Behind Your Favorite Generative AI Product - MacroPolo. MacroPolo.
[4] MENG JIANGUO, CADE METZ (2020).U.S.AI.U.A.'Chinese TALENT.NEW YORK TIMES