Economist Dieter Ernst worries that "Technology warfare based on crude techno-nationalism is threatening to destroy AI’s global knowledge-sharing culture...
The U.S.-China trade trade war is not achieving much, and when it comes specifically to the artificial intelligence market, it is doing little more than impeding innovation and damaging the prospects of everyone involved in developing AI. EE Times sat down (virtually) with economist Dieter Ernst, who recently completed a report on China’s AI Chips, “Competing in Artificial Intelligence Chips: China’s Challenge Amid Technology War.” The 70-page report was published by the Centre for International Governance Innovation (CIGI).
Ernst, a long-time China watcher and expert on the global semiconductor industry, is a senior fellow at CIGI (Waterloo, Canada) and at the East-West Center (Honolulu).
We’ve known Ernst for a while, as a constructive voice on the U.S.-China trade/technology relationship. Ernst isn’t one of those scholars who pontificates from the ivory tower. He compiled his latest report literally on foot, through field studies conducted in China in 2019.
China’s AI Superiority: Myth or Reality?
Our conversation meandered from changes in attitudes among Chinese researchers and companies toward foreign researchers like Ernst himself to the fallout of intensifying US tech restrictions on China.
Throughout, we were mindful that the next steps taken by the U.S. and Chinese governments will have far-reaching consequences for the global tech community.
Here are excerpts from our chat.
Changes in Chinese attitudes
EE Times: I have noticed in the past 18 months the increasing difficulty of setting up interviews with Chinese companies — especially those developing AI chips. Some bluntly said that publicity in the Western press only invites trouble. “Now is not the time,” they often say, “to draw unnecessary attention from the U.S. government.”
What about you?
Dieter Ernst: Compared to my earlier field research in China (which goes back many years), I was surprised that lately companies and government agencies were much more reluctant to accept interviews. Initially, there was resistance against a foreigner participating. This shows that the intensifying US-China trade and technology war has made objective field research much more difficult.
EE Times: That’s not good. How did you overcome that hurdle?
Ernst: It probably helped that I gave invited talks on the topic of this research to leading universities in Beijing, Shanghai and Hangzhou, in April 2019 and, most recently, in November 2019. These talks were well attended and provided some unique insights during Q&A sessions and informal discussions afterwards.
In all my discussions, Chinese participants were very concerned that US technology export and visa restrictions would stifle knowledge exchange critical to innovation in the highly globalized AI research community.
China’s AI community caught off guard
EE Times: What other things surprised you?
Ernst: I knew that China has long factored in the risk of future US technology restrictions and it has searched for ways to develop alternative sources of supply. During the interviews, I was surprised to find that both the government and key players in China’s AI industry were caught somewhat flat-footed. While various scenarios have been war-gamed, key players (including those on the government’s A-List) appear to have been unprepared for the sudden escalation of the trade conflict into a technology war.
EE Times: What specific areas of AI do you think caught China off guard?
Ernst: Advanced specialized AI chips that provide increased computing power and storage, while decreasing energy consumption. Time and again I was told that companies that have access to leading-edge AI chips are essentially in the fast lane, where improvements continue to be rapid and mutually reinforcing. China has relied almost solely on the United States to import such advanced AI chips [for training]. The US-China technology war has abruptly disrupted China’s access to these critical sources of AI success.
EE Times: In reading your report, one of the things that surprised me about China’s AI development is what you described as “limited interaction” between domestic industry and public research institutions. Why do you think this happened?
Ernst: … [Historically] China’s innovation system is constrained by multiple disconnects.
EE Times: Disconnects of…?
Ernst: Disconnects between research institutes and industry, between “civilian” and “defense” industries, between central government and regional governments; and between different models of innovation strategy.
EE Times: How do you explain such disparities?
Ernst: To a large degree this reflects the fact that China was late in joining the global innovation race in high-tech industries.
In addition, China had to cope with the institutional heritage of the Soviet planning system and the hugely disruptive effect of the “Cultural Revolution.”
As a result, R&D remained locked into different layers of public administration (both central and regional), while enterprises were assumed to be pure “production units” without adequate research and engineering capabilities and no role to play in marketing and strategic planning.
Despite many efforts of “market reform and organizational change,” there is still a long way to go to enhance knowledge exchange between industry and public research institutions.
EE Times: And then, there is a well-known, wrong-headed China’s patent policy.
Ernst: In fact, China’s patent policies are still primarily focused on pushing up the number of patent applications, while little attention seems to be paid to what happens with these patents once they are registered. Most important, no significant efforts are made to identify, develop, maintain, and improve the quality of patents that could achieve high citations.
No more global knowledge sourcing?
EE Times: So, in your opinion, what’s the biggest challenge faced by China in its AI development?
Ernst: The fragmentation of China’s innovation system highlights a fundamental conundrum of China’s approach to AI development. Before the outbreak of the US-China technology war, Chinese AI firms innovated in areas that reflected their comparative advantage. Exploiting their huge database through their vast pool of low-cost university graduates, they focused on competing in China’s rapidly growing mass markets for AI applications.
This strategy was made possible by China’s deep integration into international trade and global production networks, which has provided ample opportunities for global knowledge sourcing. To the degree that Chinese companies were able to rely on foreign technology, they could grow and prosper without investing in in-house basic and applied research. With rising US technology export restrictions, it has become much harder to reap such gains.
China vs. US: Tech Warfare Hurts All
Is massive data China’s AI strength?
EE Times: After we published our initial story about your China AI Chip report, I heard back from a couple of Chinese executives who agree that China was misled by Lee Kai-Fu’s book. Lee, a former Apple, Microsoft and Google executive who is now a venture capitalist, suggested in his oft-quoted book that China should leverage its “big data treasure trove to get ahead in mass markets for lower-cost AI applications.”
Obviously, this strategy, followed by many companies, put China on the wrong path, since China is left with AI applications that use outdated neural networks. Meanwhile, the field of fundamental AI research and development, still evolving, is rapidly advancing elsewhere in the world.
Aside from Lee’s dubious theory, I’m often told that even though China’s government is collecting massive surveillance data, it isn’t necessarily sharing the data with private companies. Or the government is not necessarily labeling such massive data. Have you heard anything similar?
Ernst: In our informal interviews, both your observations were mentioned quite frequently. In the medium-term, China seems to face a double whammy:
Together, this constitutes quite an explosive mix that may over time stifle AI innovation as well as AI standardization and governance in China.
EE Times: In reading your report, I was a little surprised to learn that AI investments by digital platform vendors in China (such as Baidu and Alibaba) are also driven by AI applications. Why is that?
Ernst: The report focuses primarily on research in AI chips. To understand where China really stands on AI, we indeed need detailed studies of priority areas of applied and fundamental AI research conducted by the BATs (Baidu, Alibaba, Tencent, etc), Huawei, Lenovo, etc.
The Alibaba DAMO (Discovery, Adventure, Momentum and Outlook) Research Academy does substantial research in AI algorithms (most of it applied research). But the main focus is on applying AI algorithms to a broad portfolio of AI applications in voice recognition, natural language understanding, data mining, smart cities, smart robotics, and manufacturing IoT.
Baidu’s R&D focus is on autonomous cars and transportation systems, natural language processing (NLP) frameworks, and high-performance computing.
EE Time: So what are the most important messages for U.S. industry from your report? What are the implications for industry and government in both countries?
Ernst: It is time to accept that the US, the most powerful country in the world, can no longer single-handedly dictate the pace of innovation in AI and in IT at large. It is in the interest of US industry that the government returns to a policy that promotes rather than disrupts the rule of law in international trade, in order to regain stability, predictability and a more equitable distribution of gains from trade. It will take quite some time to repair the tremendous damage done by current US policies.
There is no doubt that Commerce and USTR (US Trade Representative) should fight against unfair practices in IPR, trade secrets, and government procurement, wherever they occur, including in China. But most important, both the US government and the private sector need to join forces and develop and implement a national strategy to upgrade the country’s innovation system in order to cope with the challenge of China’s innovation policy from a position of strength.
China in turn needs to reconsider the notion that the country can only progress in AI if it pursues a zero-sum competition policy in its relationship with the United States and other advanced countries. China should provide safeguards to foreign companies against forced technology transfer through policies such as compulsory licensing, cyber security standards and certification, and restrictive government procurement policies.
In short, technology warfare based on crude techno-nationalism is threatening to destroy AI’s global knowledge-sharing culture. This is true both for the “America First” doctrine and for China’s attempts to claim sovereignty over its cyberspace through the Great Firewall.