ChinAI Newsletter #22: An Open Source AI Strategy - China's New White Paper on AI Open Source Software
|Jeffrey Ding||Aug 6, 2018|
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.
New White Paper on AI Open Source Software (AOSS - get used to this acronym, you'll be seeing it a lot in the translation)
This week has one big translation: we're diving deep into a new White Paper on China's Development of AI Open Source Software (中国人工智能开源软件发展白皮书) published by a new body called the China AI Open Source Software Development League, which was established on March 15, 2018, to support the China Electronics Standardization Institute (CESI) under the Ministry of Industry and Information Technology.
Here's the background context:
- 166pg. White Paper written in a very similar style to a White Paper on AI Standards which was also led by CESI, which I've translated in past issues (remember you can see all past issues via the archive link in the welcome message) and written about here.
- The League and CESI convened 60+ contributors for the White Paper including heavy-hitters like Peking University, Chinese Academy of Sciences, Jingdong, Webank, Ant Financial, Alibaba, Baidu, and Huawei
- Five main sections: Summary, Current Status of AOSS Development, Analysis of the AOSS Ecology, Recommendations for the Development of China's AOSS, and lastly a huge section on Cases of AOSS Application Areas and Scenarios (which includes more than 60 pages worth of details on how AOSS is being applied to real-world problems)
- In the full translation, I've included a Table of Contents and bolded the sections I've taken a stab at, mainly: Analysis of the current landscape of AOSS, Recommendations for Development of China's AOSS, and one case study of an AOSS application area
Main Takeaways from the sections I translated this week:
- There was some fluff and repetition in this White Paper, but overall this was an effort that required some serious intellectual capital
- Leading countries in AOSS are the U.S., China, and other developed countries. U.S. is home base for groups/main developers of 66% of world's AOSS, China is at 13%, and other countries are at 21%, per the White Paper. There's an explanation of the methodology for getting to these figures in the translation.
- Maintainers of AOSS are divided into three groups: companies (e.g. Google), research organizations (e.g. UC Berkeley), and open source development groups/individuals (e.g. Apache software foundation); Baidu is the main Chinese institution mentioned in this section
- Outlines three phases for the development of China's AOSS:
1. Directly Adopt, Partly Participate - adopt and learn from currently existing AOSS and contribute to sections of current AOSS
2. Emphasize Breakthroughs, Lead Locally - the emphasis here is on independently researching and developing AOSS local functions and modules, particularly in areas where Chinese developers may have an advantage, including Chinese-language data processing, speech recognition, and knowledge mapping
3. Independently Lead, Widely Used - ultimate goal is to lead the trend for AOSS development, which includes an interesting emphasis on technical systems that can evaluate and assess AOSS programs
- White Paper also constructs a 3-D model for China's AOSS development based around three aspects of “construct an open ecology,” “build up an external environment,” and “drive forward industry applications."
For those interested in the more technical aspects of how AI is actually implemented (including details like using Mobilenet V2 and Shufflenet to compress deep learning models so as to implement real-time target detection schemes in intelligent traffic video analysis), check out appendix 3 at the bottom of the translation.
This Week's ChinAI Links
A must-read from Elsa Kania on why open debate about the ethics of tech in the U.S. is actually a strength of the U.S. innovation system, in contrast to the relationship between China’s AI giants and the communist party: “The United States must recognize its own enduring advantages, including the dynamism and inclusivity of its innovation ecosystems. The open and intense debates over AI ethics should not be dismissed as a potential disadvantage—but rather recognized as integral to America’s values and vitality as a democracy.”
Logic is a magazine about technology and society that is seeking pitches on an issue about China, find out more here.
Interesting South China Morning Post article on applications of AI to Chinese foreign policy, with comments from the Chinese Ministry of Foreign Affairs
Ran across another cool translation platform this week called China Heritage which has some great translations of Chinese literature.
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 email@example.com or on Twitter at @jjding99