Unveiling Misconceptions and Strategic Insights in the US-China AI Competition

Jeffrey Ding is an Assistant Professor of Political Science at George Washington University. His research agenda covers emerging technologies and international security, the political economy of innovation, and China's scientific and technological capabilities. His book, forthcoming with Princeton University Press in August 2024, investigates how past technological revolutions influenced the rise and fall of great powers, with implications for U.S.-China competition in emerging technologies like AI.. Dr. Ding’s research has been published or is forthcoming in European Journal of International Relations, Foreign Affairs, International Studies Quarterly, Review of International Political Economy, and Security Studies, and his work has been cited in The Washington Post, The Financial Times, and other outlets. He received his PhD in 2021 from the University of Oxford, where he studied as a Rhodes Scholar, and earned his B.A. in 2016 at the University of Iowa.
Jian Athena Ke '26 interviewed Dr. Jeffrey Ding on April 12, 2024.
Photograph and biography courtesy of Dr. Jeffrey Ding.

What is the most prevailing misunderstanding in the mainstream media about the US and China AI competition? And what do you think lies behind such misunderstanding?

One of the primary misconceptions my research addresses is the notion that China has either already surpassed or is on the brink of surpassing the United States in terms of AI prowess. This belief stems from a tendency within the US to consistently overestimate its technological rivals. This pattern echoes historical instances, such as the overestimation of the Soviet Union's capabilities during the Cold War and Japan's technological dominance in the 1980s and 1990s. This trend seems to persist in assessments of China's AI capabilities. 

Many commentators and influential thinkers have likened China's advancements in AI to a "Sputnik moment" for the US—a wake-up call indicating that China is nearing parity or even supremacy in these capabilities. Reports like the final assessment from the National Security Commission on AI have underscored this concern, warning that China could outpace the US within the next decade. However, a comprehensive analysis of China's AI landscape compared to that of the US reveals a significant lead held by the latter. The US is indisputably well-positioned to maintain its status as the preeminent leader in AI. 

How do past instances of technological revolutions, explored in your research, offer insights into the strategic advantages and vulnerabilities that the U.S. and China may encounter in the current AI competition?

A central point I aim to convey is that historical patterns regarding technological revolutions and their impact on the rise and decline of major powers reveal a consistent theme: new technologies primarily influence national power through economic productivity growth, preceding shifts in geopolitical influence, and military capabilities. Historical examples, such as Great Britain's ascent to economic dominance during the first Industrial Revolution before becoming a geopolitical and military hegemon, and the United States' emergence as an economic powerhouse following innovations in the second Industrial Revolution from 1870 to 1914, illustrate this pattern. Thus, the crucial lesson for today's US-China competition lies in recognizing that the most significant impact of technologies like AI may be on economic productivity and leadership, echoing the historical precedence set by prior technological revolutions. 

One of the notable breakthroughs that underscore this point is the advent of AI-driven chatbots like ChatGPT, a sophisticated language model built upon the foundations of GPT-3. This development has significantly accelerated research in natural language processing, which focuses on enabling machines to comprehend and communicate using human-like expressions. This breakthrough sparked a global competition among various countries and research labs to develop large language models comparable to or even surpassing ChatGPT's capabilities. Notably, this competition has been particularly intense in China, where research labs have been striving to catch up with GPT-3, the pioneering model mentioned earlier. 

On average, Chinese labs have demonstrated remarkable agility as "fast followers" in this field, typically taking one to two years to develop their own competitive large language models. Despite this rapid progress, there still exists a performance gap, even in handling Chinese language prompts. Nonetheless, this development underscores the importance of tracking how the competition unfolds within the realm of AI research. 

Considering your analysis of China's scientific and technological capabilities, what specific areas within AI research and development in which Chinese institutions and companies particularly excel, and how does this compare to the strengths of their American counterparts?

Chinese companies have established significant competencies in several key areas within the realm of AI. One notable domain is facial recognition, where there is substantial demand from the government for surveillance systems and identity authentication services. Chinese computer vision companies have benefited from access to government contracts at both the central and local levels, thereby bolstering their expertise and market presence in this field. 

Another area of strength lies in Chinese language natural language processing (NLP). Certain Chinese labs concentrate exclusively on serving the domestic market with their chatbots and language models, enabling them to focus on the nuances and intricacies of the Chinese language. This includes considerations such as idiomatic expressions and specific formats like couplets. By tailoring their products to the particular features of the Chinese language, these companies are better positioned to meet the needs of their primary market, potentially allowing them to excel in serving specific language markets more effectively. 

Your work delves into the political economy of innovation. Could you discuss a political or specific policy or regulatory approach in China and the US that has notably influenced the AI ecosystem and competitive positions?

A significant policy that has impacted both the US and Chinese AI ecosystems is the US's decision in October 2022 to impose export controls on high-end chips crucial for training AI models. These export controls have limited access for Chinese labs to chips often supplied by US companies such as Nvidia. Consequently, this restriction has posed challenges for China's AI ecosystem, leading labs to either seek alternative chip sources or stockpile existing reserves of Nvidia A100 chips. 

However, the policy has also affected the US ecosystem, as companies like Nvidia, which develops these high-end chips, are now losing a valuable customer base, including firms from China. This underscores the interconnectedness of both economies and their AI ecosystems, revealing how policy decisions in one country can have ripple effects across borders and industries. 

The U.S. has restricted China’s access to advanced AI chips.  How has this affected China’s efforts to develop its AI capabilities?

In the initial year following the implementation of export controls from October 2022 to around November 2023, loopholes persisted, allowing third-party resellers to circumvent some restrictions. Many Chinese companies still manage to leverage cloud computing to access compute resources from clusters located in other countries. However, the US government tightened these controls in November 2023, closing several of these loopholes. Despite this, the impact has been somewhat mitigated by major Chinese companies stockpiling chips. Looking ahead, the key question is whether Chinese firms can develop their own advanced AI chips, with Huawei's Ascend processors being a significant indicator. 

For US semiconductor companies like Nvidia, continuous reinvestment in research and development (R&D) is crucial to maintain a competitive edge in chip technology. However, losing a significant portion of revenue from sales to Chinese companies due to export controls could impede this reinvestment. The concern is that without adequate funds for R&D, US companies may fall behind in developing the next generation of chips. Therefore, the US government may need to consider providing adjustment assistance to these companies to offset potential losses incurred from export controls, ensuring they can continue to compete at the forefront of technological innovation in advanced AI chips. 

Jian Athena Ke '26Student Journalist

El contenido de Pixabay se pone a su disposición en los siguientes términos (“Licencia de Pixabay”). En virtud de la Licencia de Pixabay, se le otorga un derecho irrevocable, mundial, no exclusivo y libre de regalías para usar, descargar, copiar, modificar o adaptar el Contenido con fines comerciales o no comerciales., CC0, via Wikimedia Commons


Share this:

Leave a Reply

Your email address will not be published. Required fields are marked *