The Diplomat author Mercy Kuo regularly engages subject-matter experts, policy practitioners, and strategic thinkers across the globe for their diverse insights into U.S. Asia policy. This conversation with Dr. William C. Hannas – professor at Georgetown University, lead analyst at the Center for Security and Emerging Technology (CSET), and co-editor with Huey-Meei Chang of “Chinese Power and Artificial Intelligence: Perspectives and Challenges” (Routledge 2023) – is the 421st in “The Trans-Pacific View Insight Series.”
Examine the correlation between Chinese power and artificial intelligence.
Artificial intelligence (AI) is one of a short list of technologies marked for priority support in China and, by some accounts, ranks at the very top. The attention the PRC government pays to AI development is reflected in state-level plans that began in 2015; by the position AI holds in China’s 14th Five-year Plan (2021-2025), where it ranks first among “frontier industries;” and by China’s unabashed goal to be the world’s leader in AI by 2030.
This aspiration is credible. Behind China’s declarations of intent are the volume and quality of Chinese AI research evidenced in peer-reviewed journals, the scale of state- and private-sector investment, the infusion of AI at all levels of education, support from its diaspora population, China’s unmatched ability to field applications, and its willingness to explore alternate paths to advanced AI beyond the large language models (LLM) that typify most global initiatives.
The link between AI and a country’s economic and military power is understood clearly in China. Our greatest concern – and we have many – is the likelihood China’s ruling elite will harness AI to perpetuate itself under the guise of “safety” and alignment with socialist values.
Identify China’s diverse range of targets and programs for acquiring foreign AI proprietary information and talent.
The PRC’s history of looking abroad for inspiration in science and technology (S&T) has been documented in detail, is acknowledged by the principals, and needs no elaboration. Essentially, we took our understanding of China’s transfer practices and looked for AI examples.
Every transfer venue identified earlier, without exception, we found to be used to support China’s AI development. The science ministry’s foreign transfer offices are all doing some form of AI “outreach.” The State Administration of Foreign Expert Affairs, Ministry of Education, and technical ministries have ramped up engagement. All major “talent” programs are recruiting AI specialists. Most China-oriented overseas professional associations have signaled support. Their counterparts in China – the tech “transfer centers” – have adjusted their priorities.
Access to global AI expertise is facilitated by U.S. tech firms, many of which have facilities in China, and by foreign academics individually and through their institutions. Examples are international AI alliances, school-to-school partnerships, co-authored research and, of course, students abroad. That said, one should not disparage China’s indigenous AI research. The two approaches complement each other.
Analyze the link between China’s AI advancements and its neuroscience R&D.
Brain and AI research have inherent commonalities that reward interdisciplinary study. The goal of AI research from its outset in the 1950s has been to replicate human cognitive functions, so even if one discounts the direct role of brain science (how much do birds and airplanes have in common?), AI is still mimicking the human brain’s output – learning, vision, planning, etc. – so the dependencies are there.
The chief difference is Western AI downplays these linkages while Chinese AI embraces them. As late as 2015, “AI” appeared in state notices as a catalyst for brain science only. Later plans re-emphasized this connection and carved out distinct funding categories for brain-inspired (类脑) AI research. The emphasis on BI-AI in China is evidenced in a rising number of related papers, dedicated research facilities (31 at last count), and in a survey we did of Chinese AI scientists that showed 84 percent of the respondents crediting BI-AI with a higher likelihood of success.
China, like the rest of the world, is furiously pursuing LLM research but is not stuck in the same “monoculture” and is shrewdly hedging its bets.
What are the main goals of China’s AI-brain research?
The overarching goal, articulated in China’s 2017 plan and in statements by top scientists, is a “merger” of human and artificial intelligence. This can be interpreted metaphorically as greater interdependence between AI and humans, and literally as the loss of any meaningful distinction between the two forms of cognition. Unlike their Western counterparts, Chinese scientists have fewer qualms about a future dominated by artificial general intelligence (AGI).
The nexus between AI and brain research comprises three areas. There is BI-AI that aims to replicate human cognitive functions by modeling the brain’s structural, functional, and effective linkages. There is “connectomics,” wherein the dependencies are reversed, namely, AI facilitates brain modeling via algorithms that filter out noise or by extrapolating from visible to invisible neuronal pathways. Finally, there are brain-computer interfaces (BCI), in which China is heavily invested and where AI has major roles. Chinese AI researchers agree with this classification.
Our study unearthed thousands of datapoints relating to China’s AI-brain research, personalities, and infrastructure. Xi Jinping himself publicly encouraged the pursuit of connectomics! Can you imagine a Western head of state doing that?
Assess effective ways to measure China’s AI risk and monitor its global developments.
The risk associated with China’s AI development is a subset of the risk one assigns to AI in general, and a measure of how China uses this game-changing technology.
There is no consensus on the former: predictions range from a heaven-on-earth Singularity to human extinction. Assuming some middle ground, our chief concerns are military adaptations, enhanced influence operations and political control (the immortal Chinese Communist Party), production efficiencies that disadvantage competitors (us), ownership of AI standards, CCP values alignment, and – at the far end – privacy usurpation and mind control through BCIs and “affective” computing, potentially at the discrete neuronal level.
Since we are not in the bean-counting business, we are uncertain how these risks can be quantitatively assessed. But we do have thoughts on how Chinese AI can be monitored, which derive from China’s own decades old foreign S&T intelligence operations.
In a nutshell, China has – and we lack – a dedicated open-source collection and analysis capability focused on foreign S&T developments (of which AI is a part). What little capacity we do have is used to enable classified collection, which ironically is the least productive venue for keeping tabs on foreign science, whose gems are “hiding in plain sight.”