ChinAI Newsletter #28: Chips and Compute - Alternative Perspectives and City Rankings!

Welcome to the ChinAI Newsletter!

I’m experimenting with a switch to Substack over Mailchimp for distributing the newsletter. One of Substack’s cool features is a full archive of past ChinAI issues (Mailchimp only offers the past 20). This is more in line with the vision of ChinAI as a weekly-updated community library. Let’s say you want to learn more about how AI is influencing China’s privacy landscape. You can now go to the full archive of past ChinAI issues, scan the past issues and dive into issues #9, #16, and #19 which focus on privacy and data protection issues. My aim is to keep ChinAI free as long as possible, but full disclosure: as we hopefully expand the translation team, Substack does offer nifty paid subscription features for compensating them. The design is a little more barebones than the Mailchimp version (no more bright blue buttons, all links to translations and articles are bolded and underlined). Please give me feedback on this trial switch to Substack.

China’s Most Famous Economist: Dangerous to Develop the Chip Industry at All Costs

Re-raising this week and pushing our chips into another issue focused on China’s chip industry and the linkage between compute and AI applications. In contrast to last week’s epic 34-page/10,000-word screed on how China could create an “independent” chip industry, this week’s three bite-sized translations provide some alternative perspectives. The first is by Wu Jinglian, arguably China’s most famous economist. He’s known as “Market Wu” for his outspoken liberal views on economic reform. This was also published in the wake of the ZTE sanctions in April and directly rebuts the translation from last week.

The value of this type of piece is that it serves to remind us that just because the Chinese gov. announces a massive chip fund or flashy AI plan, there is still a vigorous debate happening over whether a statist, planned approach to strategic technologies is actually advantageous. Wu argues that the characteristics of the chip industry, esp. the need for constant technical iterations and upgrading, mean that the “develop chips at all costs” and “Two Bombs, One Satellite” approaches (which Chinese public discourse trumpeted post-ZTE) are doomed to fail. In the past, central and local government investments in the chip industry have been detrimental because the easy money attracts a large number of “spoilsmen.” Link to full translation of Professor Wu’s article:

Full Translation: “Wu Jinglian talks about the ZTE incident: it is dangerous to develop the chip industry at all costs”

How Many More Years Will it Take for China to have a Proper Place in the Global Chip Market?

That’s the question posed by a Weibo topic - a prominent blogger 破破的桥 (Peking University graduate who focuses on public opinion on the web) provided the top response. An excerpt:

“Well now mobile chips are very mature and there is not much to do. Where can the money be invested more efficiently? One is the iOT field. Although the profits are relatively thin, the giants don’t have the desire to enter the industry. For example, Intel has cut down its entire iOT department. The other is the field of AI chips. These two areas are highly customizable, demanding, and R&D is just beginning. There is no monopoly giant, and it is quite diverse. If you are really have interests in and dreams for chips, don't research general-purpose chips, but instead join these two areas as much as possible. Doing so may not match your ideals, but it will save a lot of time and money.”

Link to full translation of this blogger’s somewhat cynical response: How Many More Years Will it Take for China to Have a Proper Place in the Global Chip Market

Computing Power Rankings: Chinese Cities Edition

Last couple translations on chips were from April, this article is on a forthcoming “2018 China AI Computing Power Development Report” by Inspur and IDC Research which was previewed at a conference on September 12.

In the first tier: Hangzhou, Beijing, Shenzhen, Shanghai and Hefei rank as the top five in cities for AI computing power (measured by demand for servers/trillion GDP)

In the second tier: Chengdu, Chongqing, Wuhan, Guangzhou, and Guiyang.

Don’t get too distracted by the power rankings - dive deeper into the full translation - it references the OpenAI article on the doubling of computing power, contains some nice insights from expert interviews, and provides a good overview of landscape of AI commercialization in different cities (the ChinAI story is not just a Beijing/Shanghai/Shenzhen/Guangzhou story!!!)

Full Translation: 2018 China City Rankings in Computing Power: Hefei and Guiyang Soar as Dark Horses in the Midwest Region

This Week's ChinAI Links

Chinese Idiom of the Week:  欲速而不达 (yu4rsu4er2bu4da2) – if you rush something quickly, then you will actually be farther from your goal, “haste makes waste” - from the Analects.

Am behind on my reading for this week but will catch up next week. In the meantime, re-upping this Chublic Opinion piece on the trade war - was original source of two of the chip articles from this week.

Also Sinocism, the OG China newsletter by Bill Bishop, has a two-week free trial for his paid newsletter - the “presidential daily brief for China hands.” Check it out here.

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 jeffrey.ding@magd.ox.ac.uk or on Twitter at @jjding99