ChinAI #202: CUHK as Cornerstone of China's Computer Vision Scene (part 2/2)
Finishing up a longform article on CUHK and the history of computer vision in China
Greetings from a world where…
it’s not a question but a lesson learned in time
…As always, the searchable archive of all past issues is here. 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).
Feature Translation: CUHK and China’s Computer Vision Ecosystem (part 2)
Context: Last week, we covered the first half of an excellent Leiphone article (link to original Chinese) on Tang Xiaoou, Chinese University of Hong Kong (CUHK) Professor and SenseTime co-founder, as the keystone figure in China’s computer vision ecosystem. We learned about Tang’s rocky start to his academic career at CUHK and his efforts to recruit students from mainland China (starting with his alma mater USTC) to join his MMLab. How do we get from there to SenseTime and the take-off of China’s computer vision industry? Let’s go through the second half of this longform article.
Key Takeaways: Tang builds another HK-mainland bridge via Microsoft’s Beijing-based lab, Microsoft Research Asia (MSRA)
After Tang co-authors papers with MSRA researchers, he starts to regularly visit the Beijing lab. During this period, he recruits Shuicheng Yan, an intern at MSRA at the time. When MSRA offers Tang the opportunity to lead their visual computing team, Tang and Yan essentially switch places: Tang accepts MSRA’s offer; Yan goes to CUHK as post-doc to study facial recognition.
Yan marks the first student from one of China’s top two universities (Tsinghua and Peking) to join MMLab, but he is not the last. For instance, Xu Chunjing was a classmate of Yan’s at Peking; he went to MMLab as an RA and then PhD student; today, he directs Huawei Noah’s Ark Lab’s computer vision lab. Soon enough, MMLab is referred to as “a subdivision of Tsinghua and Peking.”
Tang would spend more than half of his time trying to recruit talented people. As one of his former students, Dahua Lin, jokingly tells Leiphone, “Throughout the development of his entire career, Dr. Tang is really an excellent HR (human resources person).”
MMLab’s reputation takes off alongside the deep learning wave
Around 2005, people at top computer vision conferences knew about MMLab but there was still a big gap between it and the top MIT and Stanford labs. Things change in 2009 with two developments: 1) MMLab and MSRA co-author the CVPR best paper (first time awarded to an Asian institution); 2) after receiving his PhD at MIT, Wang Xiaogong returns to MMLab and decides to go all-in on deep learning, which was a nascent direction at the time.
Conditions for making deep learning work were poor. Wang and two doctoral students wrote code in C++ to run on a personal computer’s CPU; frameworks that standardized basic tasks like Caffe had not yet emerged. Tang had taken a position at the Shenzhen Institute of Advanced Technology (Chinese Academy of Sciences) and used their funds to support MMLab’s research.
Those early, tough years of investment paid off after the AlexNet breakthrough ushers in the deep learning wave. Per the article: “From 2011 to 2013, CUHK’s MMLab published a total of 14 research papers based on deep learning in ICCV and CVPR, accounting for half the total number of deep learning papers accepted by the two top conferences worldwide (29 papers).”
And the rest is history:
SenseTime, co-founded by Tang, Dahua Lin, and Wang Xiaogang (among others), is established in October 2014. Justin Niu, partner at IDG Capital, flies to Hong Kong to visit Tang and quickly invests in a rare angel round in the tens of millions.
Many of MMLab’s former students return to the fold as researchers after their studies. Most of the youngsters who came out of MMLab have received all the top awards in their fields (IEEE Fellows, AAAI Fellows, etc.), served as the head of computer vision for many Chinese enterprises, and became the mainstays of AI's take-off.
FULL TRANSLATION: The person who built the Chinese University of Hong Kong into China’s AI Vision Whampoa Academy
ChinAI Links (Four to Forward)
Must-read: Chinese AI Governance In Transition: Past, Present and Future of Chinese AI Regulation
I’m late to highlight Julia Chen’s excellent chapter on Chinese AI regulations, published in July 2022 in Reframing AI Governance: Perspectives from Asia, a project put together by Konrad-Adenauer-Stiftung and the Digital Futures Lab. A lot of good details about the impact of academics, netizens, and media outlets in promoting AI governance in China.
Should-attend: CSET and AMPLYFi webinar on China’s AI Workforce
Join in-person or virtually for a discussion on China’s demand for AI talent: “To bridge the Chinese AI workforce data gap, CSET partnered with AMPLYFi to develop a novel dataset of Chinese job postings and used it to analyze AI workforce demand in China. Join us for a seminar to discuss the initial findings of this research. CSET Research Analysts Dahlia Peterson and Luke Koslosky will be discussing their work on this topic with moderator Oliver Hayman, AMPLYFI’s Head of Professional Services.”
Should-read: Biden’s Unprecedented Semiconductor Bet
Matt Sheehan, for Carnegie Endowment for International Peace, has produced an informative distillation of the key issues to watch related to the U.S. government’s new restrictions on China’s access to advanced chips and semiconductor manufacturing equipment. His concluding point is important: “In the next decade, the impacts of these restrictions are deeply uncertain.”
My take: be wary of people who confidently predict what will happen decades in the future. As these past two issues of ChinAI should remind us, deep learning changed everything just a decade ago.
Should-attend: Panel on the CCP's 20th Party Congress and China's political, international, and economic future
This upcoming Friday, I’ll be speaking on a panel hosted by the Sigur Center for Asian Studies at GW, along side Bruce Dickson and Iza Ding. Two Dings, One Panel!
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 an Assistant Professor of Political Science at George Washington University.
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 chinainewsletter@gmail.com or on Twitter at @jjding99