China and the United States: Unlikely Partners in AI

Despite both rivalry and rising tensions between the United States and China, the two nations have become the world’s leading collaborators in research on artificial intelligence.

The newly released AI Index Report, which tracks AI trends on a host of fronts and is published by the Stanford Institute for Human-Centered Artificial Intelligence, finds that U.S. and Chinese AI researchers teamed up on far more published articles than collaborators between any other two nations.

Overall, U.S.-China collaborations on AI research have quintupled since 2010 and totaled 9,660 papers in 2021—much faster than the increase in collaborations between any other two nations. Collaborations between the United States and United Kingdom, the second most prolific source of cross-border research, increased almost threefold to 3,560 papers.

Read the 2022 AI Index

The startling trend highlights a paradox. Even as China and the U.S. race for leadership in what they view as a strategically important technology, researchers on both sides appear to see benefits in sharing expertise and working together.

“What’s clear is that the amount of collaboration between the United States and China has gone up dramatically, and it has gone up much more than collaborations between any two other countries,” says Raymond Perrault, Distinguished Computer Scientist at SRI International in Menlo Park and co-chair of the AI Index Steering Committee.

To some extent, the surge in U.S.-China research simply reflects the fact that both nations have poured vast resources into artificial intelligence and produce huge amounts of research. On top of that, many Chinese researchers were trained in the United States and retain close professional ties to their American colleagues.

But the practice is consistent with patterns observed during previous technological revolutions in textiles, steel, and chemical engineering. Research by Jeffrey Ding, a postdoctoral fellow at Stanford HAI, has shown that the full economic impact of historic tech advances stemmed less from which nation pioneered a technology than from which ones were best at applying it across a broad range of industries. That dispersion of technology requires sharing information across industries as well as borders, much as the United States catapulted applied British advances in steel machinery to develop manufacturing approaches that catapulted it to economic dominance.

That said, the collaboration in AI comes at a time of growing friction between the United States and China over trade, human rights, and strategic power in the Pacific Rim.  Former President Donald Trump villainized China over its trade practices, and President Joe Biden imposed a diplomatic boycott of the Beijing Olympics over China's human rights abuses. 

Indeed, the volume of published U.S.-China collaborations in AI has declined slightly from its peak in 2019. Perrault says it’s unclear whether the recent dip reflects a lag in data, a temporary disruption, or a more fundamental change. Collaborations have also declined slightly between most other nations, such as those between researchers in the United States, Canada, and Europe.

Globally, AI research has soared over the past decade. The volume of peer-reviewed AI journal articles has more than doubled since 2015, hitting a new record of 172,000 papers in 2021. On top of that, researchers posted 56,000 pre-print articles in repositories specializing in AI.

Chinese researchers have been the most prolific for the past several years, publishing 27.5% of all AI journal articles worldwide. American researchers accounted for 12%. Chinese journal articles also led those of every other nation in citations, an indicator of their scientific importance. And although the United States continues to receive more AI patents than any other nation, China is now filing more than half of all the world’s patent applications in the field.

AI research has also intensified in most other parts of the world, notably in the European Union, Canada, and Japan. The one notable exception is Russia, which has largely gone its own way.

“Russia is way, way smaller in this area,’’ says Perrault. In the United States, 1,200 institutions published roughly 3,000 cross-border AI collaborations in 2021. In China, 500 institutions published 2,000 cross-border projects. By contrast only 60 Russian institutions teamed up on 600 AI projects.

Read the new 2022 AI Index.