ChinAI #87: Chinese Academy of Sciences 2019 AI Development White Paper
Another day another white paper, daylight comes I'm on my way
|Jeffrey Ding||Mar 23|| 3|
Welcome to the ChinAI Newsletter!
Greetings from a land in which the residents seek to derive meaning and purpose amidst a stream of endless technocratic white papers…
…as always, the archive of all past issues is here and 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: 2019 AI Development White Paper
Context: Published last month by the Key Laboratory of Big Data Mining and Knowledge Management, Chinese Academy of Sciences, this White Paper provides a good overview-style update on China’s AI ecosystem divided into four parts:
Key technological breakthroughs: slide 6 has some good stats on Chinese submissions at top computer vision conferences as well as mentions of key papers (e.g. “Densenet” architecture that won the best paper award at one of the aforementioned conferences). Interestingly, this is framed as a Tsinghua University team accomplishment, even though the primary author was a postdoc fellow at Cornell (he is now an assistant professor at Tsinghua).
AI-empowered industry verticals: there’s a nice table that assesses various verticals by the degree of diffusion of AI technologies (slide 19), with security, finance, retail, transport, and medicine ranking the highest across a range of interesting variables (e.g. data cleanliness, maturation of data storage processes)
AI Open Innovation Platforms: some cool case studies here — I highlighted two slides on Tencent Miying’s auxiliary diagnosis platform (34-35), which makes some really strong claims about Miying’s capabilities and also provides an illustrative graphic of Miying’s industrial ecosystem and the interaction pathways with medical software providers and hospitals. Side note: I’m not that convinced about these “open innovation platforms” associated with companies. Isn’t Miying still a proprietary framework and really the only thing that’s different here is data sharing? Open-source toolkits for neural machine translation, like THUMT developed by Tsinghua, seem more important for open innovation platforms that actually help a bunch of other companies build on top of open source code (slide 9).
List of world’s top AI companies: obviously very arbitrary but interesting to see what researchers from this CAS key lab highlight and their selection criteria (45-46).
For those interested in digging more into the Google slide deck, the above screenshot gives a good picture of my approach. Whenever I replace text directly in the PPT deck, it's my attempt at a direct translation (all the text in the table). Whenever I comment on a slide, those are paraphrased summaries. Only slides that are marked with a comment were translated/analyzed in-depth because I thought there was interesting, new information, so skip the others unless you read Mandarin or something catches your eye in particular:
ChinAI Links (Four to Forward)
Have recently started following the China Neican 内参 newsletter edited by Yun Jiang and Adam Ni, two experienced China researchers who provide weekly briefs of commentary, analysis, and policy recommendations on a range of China-related topics, such as geopolitical competition, trade dependence, technology competition, foreign interference, regional security, and human rights. As both Adam and Yun have advised the Australian government on these issued, it’s particularly refreshing to get a non-U.S. centric view. I particularly enjoyed the recent March 15 newsletter which highlighted how Chinese netizens avoided censorship of an interview article with a Wuhan doctor via creative “translations” of the article into emoji and oracle bone versions.
A sobering analysis about the impact of high-tech surveillance on social distancing in China. “Much of China’s success so far in containing the virus’s spread outside Hubei has depended on mobilising legions of people to man checkpoints armed with clipboards and thermometer guns, or to go door-to-door making note of sniffles…For now, China’s digital monitoring methods for covid-19 are a hodgepodge of disjointed efforts by city and provincial governments…”
Should-read: (somewhat) Friendly Explanation of DenseNet paper
A helpful explainer, by Mukul Khanna, of the Densenet paper mentioned above.
Should-read: The Education China Hands Need, But Most Do Not Get
Tanner Greer, for The Scholar’s Stage Forum, argues for a text-based approach to the study of China’s Communist Party politics — drawing lessons from Simon Leys 1990 review of Laszlo Ladany’s book on the Chinese Communist Party. Ladany published a newsletter, called China News Analysis, that “was drawn exclusively from official Chinese sources (press and radio).” H/t to Ben Garfinnkel for pointing me to this post.
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 researcher at GovAI/Future of Humanity Institute.
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 email@example.com or on Twitter at @jjding99