ChinAI #162: The Misfires — How BAT All Stumbled in Medical AI
Wendy Liu translates the first article in the Leiphone series on medical AI
Greetings from a world where…
75 Washington Post reporters worked 9+ months to provide the definitive account of what happened before, during, and after the January 6th attack on the Capitol
…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: The Misfires — How BAT All Stumbled in Medical AI
Thanks to Wenmiao “Wendy” Liu for contributing this week’s feature translation. She’s a product delivery manager at Kyros.AI, an innovative AI platform for education management, and former ML geophysicist at Schlumberger Oilfield Services. Previously, she translated the July AI Development Monthly Report by SciToutiao (ChinAI #152)
CONTEXT: This is the first article in the “Medical Weak Points of the Giants” series by Leiphone, which covers how large AI companies have struggled with medical AI applications in China. Last week was IBM. Baidu, Alibaba, and Tencent are up this week.
Back in 2017, the BAT companies all laid out ambitious medical imaging strategies. Baidu released an AI system for 24-hour medical consultations. Alibaba launched the Doctor You AI system for medical imaging diagnosis. At the time, the company’s VP of health, Ko Yan, said: “Doctor You will soon enter many medical institutions across the country to serve as the best assistant for doctors. We expect medical AI to take away half of doctors’ workload within 10 years.” Tencent launched Miying, an AI medical imaging system for early cancer diagnosis. China’s Ministry of Science and Technology designated Tencent to lead a national open innovation platform in medical imaging.
KEY TAKEAWAYS:
What happened next? The most interesting stories happen after the press release:
Despite plenty of talent and resources, BAT companies are in a “rather embarrassing situation”: they can’t get regulatory approvals for medical products. Specifically, they need a Class III permit (highest risk level of products which covers medical imaging AI products for diagnostic support) from the National Medical Product Administration (NMPA), a Chinese regulatory body.
In contrast, smaller players have received approvals. According to the article, as of 2021, 15 AI products have successfully passed the certification process from the NMPA.
The buzz around the BAT companies’ ambitions in medical AI has died down. Search Alibaba Health’s “Doctor You,” and the latest news will be from 2017 and 2018. English language coverage is also mostly from 2017; see for example, this SCMP article.
It take time to get medical AI software approved:
Beijing Keya Medical, one of the upstarts that succeeded where the BAT didn’t, received approval for a software system that analyzes coronary arteries based on heart imaging scans. Cao Kun, who leads their medical R&D division, shared the timeline of certification with Leiphone:
“In 2016, we started the R&D and obtained the registration test report; In 2017, the prospective clinical trial was completed and submitted for registration; In early 2018, we entered the expedited channel of examination and approval and obtained the EU CE certification; In 2019, we completed a retrospective clinical trial; In January 2020, we obtained the first registration certificate of NMPA's Class III AI medical devices.”
Referencing comments by a vice minister involved in supervision of medical devices at a major AI conference in 2021, the article concludes:
“The signal from the national level is clear: for medical AI, strict regulation will protect the interests of patients and thus eliminate the risks faced by products. This strict evaluation mentality, coupled with the evaluation criteria that have never been set, is a tough waiting period for medical AI companies that live on financing and lack stable cash flows.”
Why have the BAT companies failed to get approvals for medical AI products?
The full article cites a lot of factors, including: shifts in strategic focus, a cold wave of investment in medical imaging AI, leadership issues, lack of domain-specific expertise in getting medical product approvals, etc.
This could change in the future: One employee who works on the regulatory side for a medical AI company added, “Several BAT products have passed through the expedited channel. As far as I know, Tencent has 3 products on the list. Moreover, these products are not in the already crowded lung nodules and coronary artery racetrack. We should hear about them soon.”
FULL TRANSLATION: The Misfires: How BAT All Stumbled in Medical AI
ChinAI Links (Four to Forward)
Should-read: Medtech AI & Software Regulation in China: 5 Things to Know
It seems like coverage of AI regulation in China often fluctuates between two extremes: it’s either a lawless land where anything goes or a government stronghold that stifles any space for private sector initiatives. In trying to give more color to the middle ground in the medical AI space for this ChinAI issue, I benefited greatly from this explainer.
Should-read: Nov 1 issue of latitude(s)
From Karin Fischer’s weekly newsletter about global higher education: A federal-government probe of academic espionage is casting a chill among scientists of Chinese descent, with fears of government scrutiny of their research leading many to cut off critical collaboration with colleagues in China.
Those are the findings of a new study by Jenny J. Lee and Xiaojie Li of the University of Arizona, who surveyed nearly 2,000 professors, postdocs, and graduate students at leading American research universities about the China Initiative.
Forty percent of Chinese or Chinese American scientists reported feeling racially profiled by the U.S. government, according to the survey, which was supported by the Council of 100, a group of prominent Chinese Americans. And half of Chinese scientists — the authors use “Chinese” as shorthand to refer to students and professors of Chinese descent, regardless of nationality — said they felt “considerable fear or anxiety” that they were being “surveilled” by federal authorities. As a result, a quarter of Chinese researchers said they planned to pull back from future projects in China.
Should-read: China Information Operations Newsletter 05
Edited by Hannah Bailey, a researcher at the Programme on Democracy and Technology at Oxford University, the latest issue of the China Information Operations Newsletter links to important topics, including China’s removal of Caixin from media sources approved for domestic republishing, inauthentic Twitter amplification of a conspiracy theory that covid was imported to Wuhan via Maine lobsters, and Xi Jinping’s heavy reliance on propaganda.
Should-read: How the PRC Blocked Out Foreign Tech Products Claiming Security Risks
For IPVM, Charles Rollet reviews the history of how the Chinese government has pursued a “domestication strategy” in video surveillance and other IT domains. Includes translations of webpages, articles, and comments by key players.
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 postdoctoral 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 chinainewsletter@gmail.com or on Twitter at @jjding99