|Jul 23, 2018||Public post|
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.
Xilinx Buys Chinese AI Chip Startup DeePhi (深鉴科技) - Should the Chinese Government have Intervened?
That's the question asked and answered by one article from this week's two translations on the larger context behind the Xilinx-DeePhi deal (which I covered in a Twitter thread earlier). Xilinx is a U.S. company that dominates field of FPGA chips, which my China AI Dream report analyzed in detail. At a high-level, think of FPGA chips as more flexible, energy-efficient chips that can run AI applications. The first translation gives a nice overview of some of the reactions by Chinese netizens to the deal ("filled with indignation," complaints that the U.S. had "deposed" one of the Chinese AI chip Big Three — Cambricon, Horizon Robotics, and DeePhi.
It ends up concluding that it was appropriate for the Chinese government to allow the deal to go through, arguing that DeePhi is considered "one of many" and that companies like Cambricon, which make their own specialized chips, will eventually replace FPGAs. The line "let the commercial belong to the commercial" and of course a reference to "win-win cooperation" are included in the article.
Since DeePhi has developed on Xilinx's FPGA framework from the beginning, its purchase doesn't threaten national security or lead to the industry to be monopolized by foreign capital, according to the Guancha.cn author. Not sure I agree with the dismissal of DeePhi's significance and the importance of FPGAs in the future, but nevertheless an interesting take to read in full.
DeePhi "Selling its Body" Reflects the Dilemma of Chinese AI Startups
Kinda striking that media headlines (both from Chinese and English media) seem to frame cross-border acquisitions as if these companies are betraying their countries. These two articles also use this technonationalist rhetoric (The phrase 卖身 in the headline, which I translate to "selling its body," could also be translated as "prostituting itself") but at least also provide some nuance and balanced views in the body of the text.
This piece from SCTN, a news agency that focuses on China's domestic stock market, analyzes the DeePhi case as a forewarning of a larger "M&A tide" in China's AI chip industry and potentially a "Bankruptcy tide" in the overall AI field. Problems faced by Chinese AI chip startups:
Too many of them - hundreds of startup companies involved in AI chips in China but since there are high technical barriers, cycles are long, and the necessary investment levels are high --- currently the strength of talent and funds are too dispersed
Too separated from commercialization/use cases - it's difficult for AI chips to fit neatly different application scenarios, driven by different terminals, customers, and business models.
Foreign companies have nearly monopolized the CPU, GPU, and FPGA market - a large number of Chinese AI chip companies just do secondary development, optimization, and things on the application level on the architecture of these dominant companies
This Week's ChinAI Links
Along the theme of China's AI chip companies facing challenges, h/t to Lewis Shephard for pointing me to this NYT piece by Li Yuan on China's tech startups facing an overall cash crunch
Future of Life Institute has produced a great overview of AI policies by countries around the world.
I'm late on reading this excellent piece by this NYT article by Paul Mozur on how AI is enabling China's surveillance activities.
Authors from Google Brain and YC Research emphasize how the emerging field that synthesizes AI and IA (intelligence augmentation) can bring many benefits. As Tim Hwang, who recommended this piece on a longform articles of the year list, writes:
The field of computing has often been characterized as a struggle between Doug Englebart’s vision of machines as a tool for intelligence augmentation and the full automation contemplated by artificial intelligence. Carter and Nielsen make an argument for a new emerging path: artiﬁcial intelligence augmentation (AIA), the use of AI systems to help develop new methods for intelligence augmentation. It’s a journey that starts with a simple demo and accelerates into a fractally fascinating and complex discussion.
Also wonderful is Distill itself — a beautiful, richly interactive approach to explaining and clarifying the often arcane finer points of machine learning. Two great tastes that taste great together.
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 firstname.lastname@example.org or on Twitter at @jjding99