ChinAI #140: 2020 China Computer Vision Talent Survey Report

Greetings from a world where…

My dissertation is signed, sealed, and submitted (but not yet defended)….now, to face the gauntlet that is the academic job market

…Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors). As always, the searchable archive of all past issues is here.

Feature Translation: 2020 China Computer Vision Talent Survey Report

Context: Deloitte together with Extreme Mart (Shenzhen-based AI developer ecosystem) and China Society of Image and Graphics, jointly released this report back in February 2021. It’s only available in simplified Chinese, for full download link, see here.

How they got their data: 12,000 computer vision (CV) students, employees, and researchers browsed and visited the survey site —> 3,169 questionnaires collected in the week that it was available —> screened down to 1,578 high-quality questionnaires (864 students, 635 employed in industry, and 79 teaching or researching in universities/institutes). These questionnaires were supplemented by in-depth interviews with 23 CV talents as well as 11 companies about their talent needs in the CV-related industrial chain. Keep in mind this is not a representative sample of the total population of CV talents in China. Nonetheless, it’s a rich source of data.

This week’s translation is jiqizhineng’s summary of the report’s main findings.

Key Takeaways: 

  • Why care about the CV talent base? Among many AI-related domains, computer vision (CV) is the largest application direction in the Chinese market, accounting for 35% of China’s AI market applications. However, according to this report, the current supply-demand ratio of CV talent is only 0.09, which is extremely scarce.

  • More than 70% of CV talents are concentrated in first-tier and “new first-tier cities”  (cities like Nanjing, Wuhan, Hangzhou, etc. that are getting close to the first-tier cities of BJ, SH, GZ, SZ in terms of attractiveness). In addition, 90 percent of CV talents pick first-tier cities as the cities they intend to advance their careers in the future. There’s very limited brain drain: Only .27% of respondents said they intended to advance their careers in Hong Kong, Macau, Taiwan, and international cities.

  • How about their academic backgrounds? While half of CV talents studied computer science, more than 40 percent studied non-CS majors, such as electrical engineering and mathematics. Here’s a very interesting data point: Among those surveyed, almost 7 percent studied the new AI major. Longtime readers may remember that momentum to set up this new major accelerated in 2018 (ChinAI #7), with the 35 universities first offering the major in April 2019.

  • What about their software development habits? Specifically, the survey asked CV talents about what types of machine learning frameworks that they have used in their research or work. Image below shows the responses divided by the CV students (left) and employees in the CV field (right). Pytorch and Tensorflow lead the way in terms of popular usage, whereas just 6.5% of CV personnel have used frameworks developed by Chinese firms and organizations. Again, another illuminating data point that relates to past ChinAI coverage of China’s efforts to build its own AI open source software (ChinAI #22, ChinAI #133)

  • Still a lot of issues with talent supply, including the speed of advance in the field itself. Colleges and universities are offering 1-2 general courses in computer vision (58 percent of those surveyed), but they can’t match the needs of students in certain subfields that are developing quickly (e.g. image matting, target tracking). The article states, “Although the number of computer vision talents in China has reached 200,000, talents who can truly meet the demands of industry and society and reach the target level are still scarce.”

A few other intriguing findings:

  • In terms of factors that CV talents considered when deciding where to live, government talent policies ranked 3rd (after salary levels and employment opportunities). Note: these are likely provincial- and local- level talent policies.

  • Average annual salary of CV algorithm researchers in 2020 was about 330,000 RMB. There are other types of CV positions, including software engineers and AI product managers. But for the highest earning positions, algorithm researchers, many companies or research institutes require candidates to publish in the top conferences in the field of computer vision (CVPR, ICCV, ECCV, etc.) and machine learning (NeurIPS, ICML, etc.)

  • There is a looooooot more to digest in the report, as this is only the translation of jiqizhineng’s readout.

FULL TRANSLATION: 2020 China Computer Vision Talent Survey Report, jiqizhineng’s readout

ChinAI Links (Four to Forward)

Must-read: Cooperative AI — machines must learn to find common ground

Allan Dafoe, director at GovAI, and collaborators from DeepMind, Microsoft, and UToronto, published a Nature commentary on Cooperative AI (see thread below for summary). Builds on their NeurIPS workshop on Cooperative AI , and they are launching a Cooperative AI Foundation to continue efforts in this area.

Must-read: Writing through the Cracks

In the always-excellent Chinese Storytellers, Yi-Ling Liu brings together a group of writers based in China who “steer away from grand narratives, and pay attention to the personal, the particular and the human.” Xuandi Wang talks about visiting hippie communes in Fuzhou; Qin Chen shares her experience reporting on LGBTQ issues; Jaime Chu explores subcultures in China, Huang Chenkuang writes a monthly column on the life stories of ordinary folks in Beijing.

Should-read: NBR Asia Policy 16.2

My essay on China’s growing role in international standards-setting organizations is featured in this issue of Asia Policy, published by The National Bureau of Asian Research. It’s part of a roundtable on China’s rising influence in IT and innovation featuring really great insights from Emily de La Bruyère and Elsa Kania, and many others.

Should-listen: BBC The Real Story on EU AI Regulations

Recent edition of this BBC podcast features good debate and discussion on EU’s AI regulations. I come on at the end to talk about AI regulations and evolving discussions of AI ethics in the Chinese context.

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a Predoctoral Fellow at Stanford’s Center for International Security and Cooperation, sponsored by Stanford’s Institute for Human-Centered Artificial Intelligence.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at or on Twitter at @jjding99