Interview, Conference Presentation
China & AI: What we get wrong | Jeffrey Ding (2020)
- Jeffrey Ding, lead China researcher at the Center for the Governance of AI (Oxford), argues that Western discourse on China's AI development is often driven by projections of fear or fantasy rather than grounded empirical analysis.
- China's 2017 National AI Development Plan is not a sudden emergence but part of a long historical trajectory of strategic planning in science and technology, connected to earlier initiatives like "Internet Plus."
- China's AI ecosystem is not monolithic; local governments, conflicting bureaucratic actors, and private companies often pursue strategies that diverge from or compete with central government directives.
- The perception that China has surpassed the U.S. in AI capabilities is a myth; detailed analysis shows the U.S. retains advantages in fundamental research, open-source software, and specific talent pools.
- Discussions regarding AI safety and ethics do exist in China, evidenced by publications from Tencent, the Chinese Academy of Information and Communications Technology, and leading academics like Zhou Zhihua (Nanjing University).
- Zhou Zhihua and philosopher Zhao Tingyang have publicly expressed concerns about "strong AI" (AGI) and the risks associated with uncontrolled superintelligence, drawing from both domestic and translated Western literature.
- Misperceptions regarding the "Social Credit System" often exaggerate its current reliance on machine learning; it currently functions primarily as a low-tech system of blacklists and credit scoring rather than a ubiquitous AI surveillance tool.
- Facial recognition technology is actively used in Chinese provinces outside Xinjiang to count and monitor ethnic minorities, triggering alarms when specific thresholds are met, which would face significant pushback in Western democracies.
- The analogy of an AI "arms race" like the Cold War is flawed because AI capabilities are not easily countable (unlike ICBMs) and involve system-wide economic and military integration rather than discrete weapon systems.
- A "new tech cold war" narrative may become a self-fulfilling prophecy, accelerating decoupling and reducing the effectiveness of technical standards bodies where current cooperation between U.S. and Chinese firms is already occurring.
- China's semiconductor industry has lagged for decades due to the Cultural Revolution erasing a generation of technical expertise and the high barrier to entry requiring massive capital and iterative learning that favors incumbents like TSMC.
- Technological progress in semiconductors is approaching physical limits (e.g., 7nm vs. 15nm efficiency gains diminishing), which may eventually create an opportunity for Chinese firms to catch up by focusing on cost and efficiency rather than pure fabrication speed.
- The concept of "Tianxia" (All Under Heaven), discussed by Chinese philosophers, envisions a form of global governance to solve collective action problems, though it carries ambiguous connotations regarding a potential Chinese-led world order.
- The "9-9-6" work culture (9 AM to 9 PM, 6 days a week) in China's tech sector is increasingly facing backlash from workers and is questioned for its sustainability and actual productivity gains compared to Western models.
- Neural machine translation has reached a point where it can significantly augment research capabilities, though high-stakes accuracy still requires manual correction of idioms and context-specific nuances.
- Becoming a "China specialist" requires language investment; reading 60-40% of relevant materials in Mandarin is essential for accessing the volume of domestic news and policy analysis unavailable in English.
- Organizations like New America (Digi China initiative) and the Center for Security and Emerging Technology (CSET) are key hubs for translation and analysis of China's technical standards and policy frameworks.
- Short-term study abroad in China can induce an "enlightened nationalism" effect, where students become more proud of their home country and less tolerant of potential conflict with their host nation, while long-term immersion remains unstudied.
- The U.S. Commerce Department's export control rules often lump disparate technologies under the broad category of "AI," illustrating a systemic "technology abstraction problem" that hinders precise policy formulation.
- AI capabilities should be analyzed along the value chain: foundational layers (open-source software), technology layers (algorithms), and application layers (hardware products), where the U.S. currently holds a strategic advantage in the foundational layer.
- Techno-nationalism assumptions often fail to account for global innovation networks where benefits do not strictly stay within nation-state borders, and multinational corporations often have interests distinct from their home governments.
- The "industrial internet" (e.g., connecting manufacturing devices for predictive maintenance) is identified as a potential Chinese equivalent to the 19th-century "American system of manufacturers," serving as a key driver of future economic power.
- Cross-company alliances, such as between Chinese firm CASIC Cloud and Siemens, occur frequently despite geopolitical tensions, highlighting the complexity of "de-coupling" narratives.
- China's disinformation tactics differ from Russia's; China generally seeks stability and the status quo, whereas Russia may actively seek chaos to weaken U.S. governance, making China less likely to engage in disruptive election interference.
- The "Community of Shared Future for Mankind" is viewed by Ding primarily as a diplomatic slogan rather than a guiding principle that negates China's pursuit of relative gains in strategic domains.
- China's interest in space and science fiction (e.g., The Wandering Earth) provides a unique cultural medium for exploring long-term AI scenarios, often framing AI challenges as global cooperation opportunities rather than purely adversarial ones.
- The most pressing gaps in U.S. understanding involve the domestic Chinese public's perceptions of technology (techno-nationalism) and the specific socio-technical dynamics of AI applications, rather than just high-level national statistics.