ChinAI Newsletter #23: Why leave the world's largest facial recognition tech platform (Face++)? Reflections from Face++'s first product manager

Welcome to the ChinAI Newsletter!

These are Jeff Ding's weekly translations of writings on AI policy and strategy from Chinese thinkers. I'll also include general links to all things at the intersection of China and AI. Please share the subscription link if you think this stuff is cool. Here's an archive of all past issues. *Subscribers are welcome to share excerpts from these translations as long as my original translation is cited.

I'm a grad student at the University of Oxford where I'm the China lead for the Governance of AI Program, Future of Humanity Institute.

Additional sections of AI Open Source Software White Paper

Before we dig into this week's main translation, a deeply personal Zhihu (China's Quora equivalent) thread by Face++'s first product manager, we've got some additional sections of last week's White Paper translated --the headline takeaway is watch out for the China Aerospace Science and Industry Corp (CASIC) - the major contractor for China's space industry. 

I've added two more use case examples to appendix 3:
1. First is on intelligent manufacturing of aerospace products, which features CASICloud (an "industrial internet" platform created by CASIC) that combines advanced machines with internet-connected sensors and big-data analytics. Per ChinaDaily, CASICloud has attracted nearly 1.7 million users and collects data from nearly 1.4 million pieces of equipment inside companies registered on the platform.

2. Second is on smart factories where the example use cases feature PetroChina (listed arm of CNPC, one of China's three big national oil companies) and CASIC again. 

White Paper on the Development of China’s Artificial Intelligence Open Source Software

Reflections on the early days of Face++

We've covered Face++ in issue #11 of the newsletter - zooming in on its role in surveillance in Xinjiang. This translation covers the billion-dollar unicorn startup's early days (2013-2015) from the perspective of Face++'s first product manager, Li Sanke (now a product manager at Uxin, a car e-commerce platform). 

He posted a long response to a Zhihu (Chinese equivalent of Quora) Thread that asked people to respond to the question: "Why did you resign from Megvii (Face++)? It's a deeply personal read that starts from him wanting to give advice to a friend who is agonizing over joining Didi Chuxing, Face++, or Jingdong (JD).

It gives insights on:
- Face++'s now-29 year old CEO - Yin Qi
- the Tsinghua clan culture in the company
- Face++'s products and their foray into consumer-facing products before pivoting back to focusing on technology and selling to businesses
- general vibes around Chinese entrepreneurship

There are so many ways one can analyze China's tech scene and it's necessary and important to talk about systems/structures and the larger context (e.g. influence of the Chinese communist party, insidious relationships with surveillance architecture, economics, great power competition); at the same time, I think it's helpful to remind ourselves to see the actual people in ChinAI, the people who are insecure about working in a company filled with Tsinghua grads, people who worry about buying a house in Beijing...

Zhihu Thread: Why did you resign from Megvii (Face++)?

This Week's ChinAI Links

Important piece in the Bulletin of the Atomic Scientists by colleagues at FHI's Governance of AI Program, Jade Leung and Sophie Charlotte-Fischer, on the DoD's new Joint Artificial Intelligence Center - the section on how the U.S. military could lead in setting the standard for preventing accidents and ensuring robustness for highly complex AI applications is especially insightful

Fascinating article about foreign espionage in Silicon Valley, with an emphasis on China’s efforts, though France, Israel, South Korea, and Russia are also mentioned - h/t to ChinAI subscriber Nahua Kang
Two relevant Arxiv links from the last two issues of Jack Clark’s ImportAI newsletter
One on how Microsoft’s XiaoIce chatbot generates poems from an image
The second on how researchers from Tencent and Hong Kong Baptist University are trying to reduce the time to train models without trading off with accuracy

Thank you for reading and engaging.

Shout out to everyone who is commenting on the translations - idea is to build up a community of people interested in this stuff. You can contact me at or on Twitter at @jjding99