The latest 2019 Global AI Talent Flow Report shows that approximately 44% of AI talents in the world receive PhDs in the US and PhDs in China. Talents with degrees account for less than 11%, and cultivating more AI doctors does not necessarily benefit the country.
There is a lot of evidence that top AI talent is in short supply. However, how scarce such talents are, or where they are concentrated around the world, is barely known.
Recently, the CEO of Element AI of Canada released the latest 2019 Global AI Talent Flow Report, which will do the number and distribution of AI talents. Summary, it can be said that it is the most comprehensive report at present, mainly collecting three data sources.
Thesis published by 21 major academic conferences in the AI field,Such as AAAI, CVPR, etc., and analyzed the author's summary.
A targeted analysis of LinkedIn search results showing how many people claim to have a Ph.D. and what AI skills are required around the world.
Refer to external reports and other supporting sources to help understand the background and better understand the rapidly changing talent pool in the global artificial intelligence arena.
Reporting results show that in 2018, the number of people who published papers at one or more top conferences in the field of machine learning reached 22,400 People, a 36% increase from 2015, and a 19% increase last year alone. A supplementary survey of personal data on LinkedIn shows that a total of 36,524 people are eligible for AI experts, which is a 66% increase over the 2018 report.
The number of authors who have obtained doctoral degrees in China is close to 11%, ranking second in the world;
More than 11% of AI talents work in China, ranking second among the countries with the most employment;
The top AI researchers have 225 China ranks second in the world in terms of quantity.
The report also presents some valuable ideas and phenomena. For example, cultivating so many AI doctors will not necessarily benefit the country. The total proportion of women in the AI top conference is only 18%, and the gender imbalance is serious.
The following is a detailed report.
Proliferation: The paper rose 30% in three years.The meeting has thousands of authors
The main source of data for this AI talent report is the academic conference in the field of machine learning. The report covers 21 meetings and surveys the past. One year the author of a paper published at major international academic conferences in the field.
The 21 meetings are:
Calculation Linguistics Association North America Annual Meeting ( NAACL)
Artificial Intelligence Promotion Association Conference (AAAI)
Computational Linguistics Association Conference ( ACL)
Computer Vision and Pattern Recognition Conference (CVPR)
Natural Language Processing Experience Method Meeting (EMNLP)
Learning Theory Seminar (COLT)
Neural Information Processing System Conference (NeurIPS)
Artificial Intelligence Uncertainty Conference (UAI)
Genetic and Evolutionary Computing Conference (GECCO)
International Conference on Acoustics, Speech and Signal Processing (ICASSP)
International Conference on Artificial Intelligence and Statistics (AISTATS)
International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS)
International Conference on Computer Vision (ICCV)
International Conference on Intelligent Robotics and Systems (IROS)
International Machine Learning Conference (ICML)
International Conference on Medical Image Computation and Computer-Aided Interventions (MICCAI)
International Conference on Robotics and Automation (ICRA)
International Joint Expert Meeting on Artificial Intelligence (IJCAI)
Robot: Science and Systems (RSS)
Computer Vision Application Winter Conference (WACV)
Results show that one or more of the above are in 2018 The authors of the papers published at the top meeting were 22,400, which means that there are more than one thousand authors per meeting.
The country of AI researchers Top 5 (Number of authors who published papers at the 21 summits surveyed):
United States: 15747 people
United Kingdom: 1475 people
Germany: 935 people
Canada: 815 people
As a comparison, the report also counts the same in 2015, 2016 and 2017 The authors of the papers published at 21 conferences showed a clear growth trend: the number of authors increased by 36% compared to 2015, an increase of 19% compared to 2017. Research is also increasing: the total number of papers published at these 21 conferences has increased by 25% compared to 2015 and 16% over the previous year.
At the same time, the number of peer-reviewed papers has also grown at the same time: 25% more than in 2015 and 16% more than last year.
However, among the researchers who published papers at these conferences, the proportion of women was very low, accounting for only 18%.
18% of all authors of the 21 meetings surveyed. The proportion of female authors in Spain, Taiwan, and Singapore is the highest, but in absolute terms, female authors in the United States account for almost half.
The AI Index 2018 Report, released recently by Stanford University, shows the same situation. Women also have a low percentage of undergraduate artificial intelligence and machine learning courses:
74% of Stanford’s AI introductory courses in 2017 are male;
In the course offered by the University of California at Berkeley, the male ratio is 73%.
The percentage of women taking the introductory course in machine learning is lower, with 76% and 79% of men at Stanford and the University of California at Berkeley, respectively. The same report also found that in the United States, the majority of applicants for AI work were men, accounting for 71%.
In absolute terms, the United States is the country with the largest number of female authors, followed by China, the United Kingdom, Germany, Canada, France, Australia, India, Italy, and Singapore.
Nearly half of the authors of the paper received a doctorate in the United States.Only 11% in China
The data analysis of the authors of the conference papers can also make some observations on where the authors receive education.
First, the number of authors who published papers was the highest in the United States: more than 44% of authors received doctoral degrees in the United States.
Second, the number of authors who have obtained doctoral degrees in China is close to 11%, followed by the United Kingdom (6%), Germany (5%) and Canada, France and Japan (4%).
Employment data has a similar geographic distribution.
The survey shows that US employers continue to attract researchers to work, 46% of whom work for US employers; more than 11% work in China and are employed The second highest number of countries, followed by the United Kingdom (7%).Canada, Germany and Japan each account for 4%.
The place where AI experts work. The number of authors in the five countries of the United States, China, the United Kingdom, Germany and Canada accounted for 72%.
In general, the 18 largest countries accounted for 94% of the total number of authors. The top five countries - the United States, China, the United Kingdom, Germany and Canada - accounted for 72% of the total number of authors.
In addition, the conference authors work in the academic world (77%) and 23% work in industry.
Cultivating so many AI doctors does not necessarily benefit the country
The data shows that about 27% of researchers with doctoral degrees, their employers The country in which the country is located and the country in which the degree is obtained are different countries. In countries with more than 150 people in the report, this percentage has risen to 32%. Why?
First The data shows that some countries have a particularly strong appeal to researchers in the field of deep learning. US companies are most likely to attract overseas PhDs to work for themselves, China ranks second, China attracts The absolute number of researchers is about one-fourth that of the United States.
The data shows that the AI talent inflow ratio in the following ten countries or regions is higher than The outflow ratios are: Taiwan, Sweden, South Korea, Spain, the United States, Switzerland, China, Japan, the United Kingdom, and Australia.
Switzerland, Sweden and the UK are ranked in the top three in terms of the proportion of AI talents receiving academic training in foreign countries, with a ratio of 50%, 49% and 44% respectively.
AI talent inflow rate and The two indicators of outflow rate can reflect to a certain extent the ability of a country to attract foreign AI talent and retain local AI talent.
as shown above,The X-axis represents the talent inflow, the Y-axis represents the talent outflow, and the numerical value represents the standard deviation from the mean. In this picture, the countries or regions in the world are divided into four categories:
"AI talent attracting countries": in the upper right corner of the figure are Australia, Spain, Sweden and China. In Taiwan, the region indicates a net inflow of AI talent in these countries, indicating that these countries or regions have an advantage in attracting foreign talent and retaining AI talent in their own country (or region).
"AI talent producing country": The main countries at the bottom left are France and Israel. The AI talent outflow ratio in these two countries is greater than the inflow ratio and higher than the domestic AI. The proportion of the talent pool. However, from the position in the picture, the AI talent outflows in these two countries are only slightly lower than the inflows, and the net outflow is very small.
"AI talent anchoring country": The AI talent outflow rate and inflow rate in the United States are very low, basically not a big AI talent pool in the country. Influence, in absolute quantity,The United States is still the largest settlement of AI talents in the world, and the AI talent pool in the United States has remained generally stable. Countries with the same characteristics include China, Germany, Japan, India, South Korea and Italy.
"AI Talent Platform Country": Finally, there is also a category of countries where AI talent outflows and inflows are rising, and these countries are attracting more and more Overseas AI talents, as well as the outward flow of their national doctoral students are also above average. Given the state and trends of these ecosystems, countries in this category include Canada, the Netherlands, Singapore, Switzerland, and the United Kingdom.
Top research is basically monopolized by big countries: the top three in the US, China and Britain
This year's survey found that the total number of authors at top international academic conferences increased by 19% compared to last year. To assess the current impact of these authors in the field, the report analyzes the citations of their published papers in 2017 and 2018:
The research of 18% of the top academic conference authors has had a significant impact on the field, and their knowledge is deep enough to continue to make a substantive impact on the research area. Sexual contributions, these experts may also be a source of potential application talents dedicated to applying theory to the team.
Data shows that these top researchers are more concentrated in some countries. The top five countries are the United States (1095), followed by China (255), the United Kingdom (140), Australia (80) and Canada (45).
The next concern is that the most influential AI researchers in a given country account for the total number of researchers in the country, which may be somewhat Reflects the country's success in cultivating top AI talent.
In this indicator, Australia is among the best, among the country’s AI talents,18% have published high-impact results. This was followed by the United States, the United Kingdom and China (13%), Switzerland (11%), Singapore (9%), Sweden and Spain (8% each), Israel, Canada and Italy (7% each).
In all countries, the most influential research is more likely to come from academia than industry. China is the country with the highest proportion of high-impact research from academia (90%), followed by Italy (86%), the United States (84%), Germany (83%) and Taiwan (81%).
France is the country with the highest proportion of high-impact research from industry (30%), followed by India and Israel (29%), Spain (28 %) and the United Kingdom (27%).
Social Network LinkedIn data shows that a total of 36,524 people have positioned themselves as AI experts in their profile, compared to 22,064 last year.It increased by 66% year-on-year.
Globally, one-third of AI talents use computer science as a relevant subject for academic training
The data shows that these self-proclaimed AI experts have been trained in various subject knowledge. 28% of people have classified computer science as their relevant subject area. In some countries, this proportion is particularly high, including France (47%) and China (44%).
Similarly, in other countries, AI talents have a higher proportion of other disciplines, such as physics: overall, 9% of AI Talent indicates that he has received physics training, but in Germany,This ratio is as high as 28%. For mathematics and statistics, the overall ratio is 18%, but the ratio is 27% in Israel and the United States, and 35% higher in Russia.
However, the limitations of LinkedIn data are obvious.
All personal information and related materials about AI talents on LinkedIn are filled out by the respondents themselves, and LinkedIn is widely used in countries around the world. different. For example, in the United States, there are currently about 144 million people who have established personal information on LinkedIn, accounting for more than 44% of the total US population. In Canada, this proportion is also as high as 38%. In contrast, in Russia and China, Linkedin is not popular, with a population coverage of only 3-5%.
The flow of AI talents: highly international, talent development should be globalized
The AI exchange between China and the United States is particularly active. The balance of talent inflows and outflows between the two sides is generally balanced: of the 22,400 researchers, about 500 experts have obtained Ph.D. degrees in China, and then work for employers in the United States. More than 500 people have obtained Ph.D. degrees in the United States and then work for Chinese employers. There is a similar phenomenon between the United States and the United Kingdom.
At the same time, American universities receive A large number of graduate students from abroad. For example, in 2015, international students obtained about one-third of graduate degrees in science and engineering in the United States, and 76% of graduates said they wish to stay in the country. In some universities, foreign graduate students The ratio is significantly higher.
Overall, in the past year, both the number of authors published in the AI field, the number of high-impact papers, and The number of AI professionals on Linkedin has increased significantly.
The number of women engaged in AI research is still insufficient, but some countries have taken a step closer to the goal of gender equality.
From the geographical distribution of AI talents Look, the United States is at the forefront of the absolute number of almost every indicator.
However, today's global AI field is highly international, each country or The regional AI ecosystem has its own unique advantages and strategic position. The country that focuses on promoting AI professional knowledge construction needs to invest in the global AI talent training required for the development of AI in the future.