ChinAI #215: Chinese orgs analyze ChatGPT team member backgrounds
Plus, my China Leadership Monitor piece on measuring China's tech self-reliance drive
Greetings from a world where…
has anyone else been listening exclusively to Zach Bryan music for the past month
…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: Analyzing the research power behind ChatGPT
Context: A sign of how OpenAI’s ChatGPT has become a phenomenon: my mom is sending me articles about this topic, including this week’s AItechtalk[AI科技评论] piece (original Chinese) on the backgrounds of ChatGPT’s 87 contributors. It summarizes key insights from a lengthy report by Zhipu Research and AMiner on the demographics, career trajectories, and educational backgrounds of these 87 OpenAI researchers (link to full report pdf).
*Recall, in ChinAI #152, we noted how data collected by AMiner, a platform co-launched by Tsinghua Professor Jie Tang, which scans academic publications and news sights, could be used to identify the most influential developments and key competition results in China’s AI ecosystem.
Key Takeaways: The report analyzed the ChatGPT team by age, educational and professional qualifications, and ethnic background
Average age was only 32 years old, with 34 percent of contributors in the 20-29 age range
Most of the personnel graduated from the usual suspects: Stanford University (14), UC Berkeley (10), and MIT (7)
The report also highlighted 9 ethnic Chinese contributors, five of whom graduated from Chinese universities (3 from Tsinghua and 1 each from Peking University/Hong Kong University and Huazhong Science and Technology University). All five came to the U.S. for further study.
This article also engages with a broader debate playing out in China over Chinese tech giants’ level of investment in AI research labs, which we covered in ChinAI #187. One of the key flashpoints of this debate was when Alibaba shut down its AI lab back in January 2021.
This article intervenes in that debate by arguing that ChatGPT’s success should inspire Chinese tech companies to re-invest in their AI labs: “In recent years, marginalization and hesitation have caused AI research institutes and scientific research talents in major Chinese tech companies to face survival problems. However, it is believed that under the impact of ChatGPT, AI talents will return to the public eye and will usher in a new round of shuffling and competition.”
As one example of China’s AI landscape shifting more back to the OpenAI model after ChatGPT’s success, the piece points to IDEA Research Institute — a research lab founded by ex-Microsoft exec Harry Shum, which recently shifted its focus to building ChatGPT-like models at the end of last year.
Thanks Mom for the article recommendation!
***FULL TRANSLATION: Analyzing the research power behind ChatGPT
ChinAI Links (Four to Forward)
Should-read: Measuring China’s Technological Self-Reliance Drive
I have a new piece out in China Leadership Monitor. The main takeaway:
Much ink has been spilled over whether China will be successful in its self-reliance drive and its Made in China 2025 goals. Much less attention has been paid to a more foundational question: what counts as success? This article argues that China’s measures of technological dependence are more ambiguous than other scientific and technological indicators, such as R&D spending or patents. This argument is supported by evidence from three areas: confusion over the definition of indigenous innovation, the expanding gray zone between domestic and foreign companies in high-technology sectors, and a close examination of China’s efforts to achieve self-sufficiency in the information-technology domain.
One implication of this argument is that the figures any analyst employs to make claims about China’s technological self-sufficiency can be molded into very different interpretations. Indeed, our claims regarding China’s technological independence may say more about us and our own biases and predispositions than about anything that constitutes an objective, clear-eyed analysis of China’s technological self-sufficiency.
I’m grateful to Minxin Pei at Claremont McKenna College for the opportunity to write on this subject.
Should-read: Censorship of Online Encyclopedias: Implications for NLP Models
In talking with people about large language models in China, one article I kept coming back to is this 2021 publication in the ACM Conference on Fairness, Accountability, and Transparency by Eddie Yang and Molly Roberts. They show that censored Chinese-language datasets can impact downstream AI applications (e.g., word embeddings trained on Baidu Baike result in different associations between adjectives and concepts such as democracy.
Should-read: CSET translation of an Israeli gov document on national security reviews of foreign investments
Edited by Ben Murphy and Rita Konaev, CSET translated a Hebrew-language Israeli government document that details how it manages national security reviews of foreign investments. Very cool to see this type of translation work!
Should-read: Give the Drummer Some
Jack Stilgoe, in Aeon, offers a beautiful meditation on drum machines, AI, and the possibilities of music.
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