Shulou(Shulou.com)06/02 Report--

Reproduced from: qubits (ID:QbitAI)

This article is 4657 words, it is recommended to read for 12 minutes.

This article introduces the global AI strength list.

Here comes the annual AI research ranking!

This time, the ranking analyzes the two top AI meetings-NeurIPS and ICML.

The ranking studies 2200 accepted papers, lists a list of authors and their affiliated organizations, and calculates the publication index of each organization.

The US publication index is close to 7 times that of China, and China ranks second.

Tsinghua University and Peking University are among the top 20 universities in the world.

Tencent, Alibaba, Baidu and Huawei are among the top 20 global companies.

It is worth noting that Tencent ranks first in China, scoring 8.8 points, surpassing Alibaba, Baidu and Huawei.

So what is the method of ranking?

According to the author, the ranking method is inspired by the "Natural Index" (Nature Index).

In order to collect the contribution of a country, region or institution to an article without being double-counted, the Natural Index uses a fractional measurement (FC) method, which takes into account the contribution share of each author.

The total FC of a paper is 1, and if each author's contribution is the same, then the total value will be shared equally. For example, if a paper has 10 authors with the same contribution, the FC for each author is 0.1.

If the author belongs to more than one organization, the author's FC will be distributed equally among each organization.

The FC of an organization is the summation of all the authors FC who belong to the organization.

The only difference between the ranking of this study and the Nature Index is that it counts overseas laboratories in the country (not the country) where the headquarters is located.

For instance.

If a paper has five authors, three from MIT, one from Oxford and the other from Google, each author will receive a score of 0.2.

In terms of institutions, MIT will score 3 x 0.2, or 0.6, while Oxford and Google will each score 0.2.

By country, the United States will score 0.8 and Europe 0.2.

If an author belongs to two institutions, such as an author from Google who also belongs to Stanford, then Google and Stein get a score of 0.2 prime 2, or 0.1, respectively.

So why did you choose NeurIPS and ICML in the research rankings? The author explains:

They all have similar popularity, similar institutional participation and similar paper acceptance rates among top AI researchers (21.2% of papers accepted by NeurIPS and 22.6% of papers accepted by JICML).

▌ 2019 AI Research ranking ▌

The top 40 leading organizations in artificial intelligence research in 2019 (industry and academia)

Google (USA)-167.3 Stanford University (USA)-82.3MIT (USA)-69.8Carnegie Mellon University (USA)-67.7UC Berkeley (USA)-54.0Microsoft (USA)-51.9Oxford University (UK)-37.7Facebook (USA)-33.1Princeton University (USA)-31.5Cornell University (USA)-30.9 Georgia Institute of Technology (USA)-30.1 University of Texas at Austin (USA)-29.9 University of Illinois (USA)-29.4 Columbia University (USA)-29.2Tsinghua University (China)-28.4UCLA (USA)-27.2Federal Institute of Technology (Switzerland) Zurich- 27.0IBM (USA)-25.8 University of Washington (USA)-24.0France National Institute of Information and Automation (France)-23.2 Lobsang Federal Institute of Technology (Switzerland)-22.3Peking University (China)-21.6University of Toronto (Canada)-21.4Harvard University (USA)-19.2 Duke University (USA)-18.7NYU (USA)-Cambridge University (UK)-15.1 Korea Institute of Science and Technology (South Korea)-14.8 Israel Institute of Technology (Israel)-14.6 University of California, San Diego (USA)-14.6 University of Wisconsin-Madison (USA)-14.4 Amazon (USA)-14.3 University of Massachusetts Amster (USA) University College London (UK)-Montreal Institute of Learning algorithms (Canada)-13.5 University of Southern California (USA)-13.5 University of Pennsylvania (USA)-13.3 Seoul University (South Korea)-12.7 Johns Hopkins University (USA)-12.6 Japanese Institute of Science and Chemistry (Japan)-12.3

Top 20 regions in artificial intelligence research in 2019

United States-1260.2 European Economic area + Switzerland-431.5 China-184.5 Canada-80.3Japan-49.4Korea-46.8Israel-43.3Australia-27.00.Singapore-13.2Russia-10.6China-5.3Kingdom of Saudi Arabia-5.0United Arab Emirates-2.3Iran -2.2 South Africa-1.0 Chile-1.0 Malaysia-0.7 Turkey-0.6 New Zealand-0.5

Top 20 countries in artificial intelligence research in 2019

United States-1260.2 China-184.5 UK-126.1 France-94.3Canada-80.3Germany-64.5Switzerland-59.3Japan-49.4Korea-46.8Israel-43.3Australia-27.0India-17.1 Netherlands-15.3Singapore-13.2Denmark-12.2Italy-11.5Sweden -11.3 Russia-10.6 Finland-9.6 Austria-7.4

Top 20 universities in artificial intelligence research in the United States in 2019

1. Stanford University-82.3

two。 MIT-69.8

3. Carnegie Mellon University-67.7

4. University of Berkeley-54.0

5. Princeton University-31.5

6. Cornell University-30.9

7. Georgia Institute of Technology-30.1

8. University of Texas at Austin-29.9

9. University of Illinois-29.4

10. Columbia University-29.2

11. University of California, Los Angeles-27.2

twelve。 University of Washington-24

13. Harvard University-19.2

14. Duke University-18.7

15. New York University-17.7

16. University of California, San Diego-14.6

17. University of Wisconsin-Madison-14.4

18. University of Massachusetts Amherst-13.8

19. University of Southern California-13.5

20. University of Pennsylvania-13.3

The top 20 universities in artificial intelligence research in 2019

1. Stanford University (USA)-82.3

two。 MIT (USA)-69.8

3. Carnegie Mellon University (USA)-67.7

4. University of California, Berkeley (USA)-54.0

5. University of Oxford (UK)-37.7

6. Princeton University (USA)-31.5

7. Cornell University (USA)-30.9

8. Georgia Institute of Technology (USA)-30.1

9. University of Texas at Austin (USA)-29.9

10. University of Illinois (USA)-29.4

11. Columbia University (USA)-29.2

twelve。 Tsinghua University (China)-28.4

13. University of California, Los Angeles (USA)-27.2

14. Federal Institute of Technology Zurich (Switzerland)-27.0

15. University of Washington (USA)-24.0

16. French National Institute of Information and Automation (France)-23.2

17. Lobsang Federal Institute of Technology (Switzerland)-22.3

18. Peking University (China)-21.6

19. University of Toronto (Canada)-21.4

20. Harvard University (USA)-19.2

Top 20 companies in artificial intelligence research in 2019

USA-167.3 USA-51.9Facebook (USA)-33.1IBM (USA)-25.8 Amazon (USA)-14.3 Tencent (China)-8.8 Alibaba (China)-7.5 Bosch (Germany)-7.2Uber (USA)-7.1Intel (USA)-6.9Toyota (Japan)-6.0Yandex (Russia) -5.8Baidu (China)-5.5Invidia (USA)-5.2Apple (USA)-4.6Salesforce (USA)-4.2PROWLER.io (UK)-4.2Criteo (France)-3.9Huawei (China)-3.7NEC (Japan)-3.5

▌ takes a closer look at ▌

Academic vs. Industry: percentage of total publishing index

Academic share: 77.8%

Industry share: 22.2%

The top 150 words that appear most frequently in the titles of NeurIPS 2019 and ICML 2019

The top 30 countries and regions in the per capita publishing index

1. Switzerland-6.97

two。 Israel-4.88

3. United States-3.85

4. Singapore-2.34

5. Canada-2.17

6. Denmark-2.11

7. UK-1.90

8. Finland-1.75

9. France-1.41

10. Sweden-1.11

11. Australia-1.08

twelve。 Korea-0.91

13. Netherlands-0.89

14. Austria-0.84

15. Germany-0.78

16. Latvia-0.67

17. Belgium-0.44

18. Estonia-0.44

19. Japan-0.39

20. Norway-0.32

21. Cyprus-0.28

twenty-two。 United Arab Emirates-0.26

23. Taiwan, China-0.22

24. Ireland-0.21

25. Italy-0.19

twenty-six。 Saudi Arabia-0.15

twenty-seven。 Greece-0.14

twenty-eight。 China-0.13

twenty-nine。 Czech Republic-0.11

thirty。 New Zealand-0.11

The world's top 40 leading organizations in artificial intelligence research in 2019 (tree map)

Overall, the top 40 organizations contributed 55 per cent of the total publication index, with a total of 1212.3 out of a total of 2200 papers.

Competitiveness in artificial Intelligence Research (Herfindahl Index)

The Herfindahl index is used to measure the relationship between the number of participants and the industry, as well as the degree of competition among participants.

The formula is as follows:

Among them

An H value below 100 indicates that this is a highly competitive industry.

An H value below 1500 indicates a lack of concentration in the industry.

An H value between 1500 and 2500 indicates moderate industry concentration.

An H value higher than 2500 indicates a higher degree of industry concentration.

In this study, the H value was 146.47, indicating a lack of industry concentration. In other words, there is no monopoly in the AI industry in 2019.

▌, who is leading the artificial intelligence industry? ▌

Nowadays, the competition between China and the United States in the field of artificial intelligence is more fierce. This ranking tends to be studied from a more balanced point of view, but before analyzing the issue, let's take a look back at history:

In 2016, two major events occurred in the field of artificial intelligence.

In March, Google's AlphaGo became the first computer program to beat nine-stage go professional Lee se-dol; in October, the Obama administration released a strategy for the future direction and considerations of artificial intelligence, called "preparing for the future of artificial intelligence."

In China, these two events have prompted the government to give priority and significantly increase investment in artificial intelligence.

In July 2017, China set 2030 as a deadline for the development of artificial intelligence: to reach the top level of artificial intelligence economy in 2020, to achieve major new breakthroughs in 2025, and to become a global leader in artificial intelligence in 2030.

Think tanks like CNAS believe that China's artificial intelligence strategy reflects the key principle of the Obama administration's report-it is now China, not the United States, that is using artificial intelligence.

The ranking of the study began in 2017, and the chart below reflects the top 10 countries in the 2017 publication index.

In 2017, the US publication index was 11 times that of China.

By 2019, the gap had narrowed sevenfold (US 1260.2, China 184.5).

In addition, an analysis by the Allen Institute of artificial Intelligence (Allen Institute for Artificial Intelligence) found that the proportion of Chinese authors among papers cited by Top 10 has risen steadily: in 2018, the proportion of Chinese authors was 26.5%, similar to 29% in the United States.

Some people will say that the competitiveness of the United States in the field of artificial intelligence may be weakened in the next decade.

According to the ranking study, the results will depend on three key elements of modern artificial intelligence: algorithms, hardware and training data.

Anyone who wants to occupy a dominant position in the field of artificial intelligence needs to do all three elements well.

At present, the advantage of the United States lies in algorithms and hardware, while China's advantage lies in the huge amount of data.

The authors of the ranking study believe that although it is difficult to draw conclusions, the United States will continue to take the lead in artificial intelligence in the next few years.

▌ research data ▌

The ranking study also released data.

Since the data at the top of AI will not be standardized, the analysis basically depends on manual work (HTML parsing, Python conversion, standardization of a large number of manual names, etc.).

The data download link is as follows:

Http://people.csail.mit.edu/chuvpilo/publications.html

Article source

Medium blog:

Https://medium.com/@chuvpilo/ai-research-rankings-2019-insights-from-neurips-and-icml-leading-ai-conferences-ee6953152c1a

-end-

Https://www.toutiao.com/i6773271868079079944/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.