Despite escalating geopolitical tensions, the United States and China maintain surprisingly robust collaboration in artificial intelligence research. A recent analysis of over 5,000 papers presented at the NeurIPS conference reveals that roughly 3% involve joint work between US and Chinese institutions. This suggests that despite political posturing, the two nations recognize mutual benefits from shared advancements in this critical field.
The Extent of Cooperation
The level of collaboration isn’t just a few isolated incidents. Roughly 141 out of 5,290 papers (3%) featured authors from both US and Chinese organizations, with similar rates (around 3%) observed in the previous year as well. The exchange goes beyond merely co-authoring: algorithms and models developed in one country are rapidly adapted and integrated into research across the Pacific. For example, the widely used transformer architecture, originally from Google, appears in 292 papers with Chinese authors, while Meta’s Llama models are present in 106. Conversely, China’s Qwen large language model features in 63 papers including US researchers.
Why Collaboration Persists
The persistence of this collaboration is not accidental. Many Chinese researchers receive training in the US, forming lasting professional relationships. As George Washington University’s Jeffrey Ding notes, both countries benefit from this arrangement, regardless of political pressure. This reality undercuts narratives of complete decoupling in AI.
“The US and Chinese AI ecosystems are inextricably enmeshed—and both benefit from the arrangement.”
—Jeffrey Ding, George Washington University
Automation in Research
The analysis itself demonstrates the growing role of AI in AI research. The study leveraged OpenAI’s Codex to analyze thousands of papers, automating a task that would have been impractical manually. This highlights the potential for AI to accelerate scientific discovery while also raising questions about the reliability of such automated tools. Researchers must verify results carefully, as AI models can make unexpected errors.
Broader Implications
This collaboration occurs at a time when both US and Chinese policymakers are increasing investments in AI, often framed in terms of national competition. The continued interdependence suggests that despite rhetoric, neither country can afford to completely isolate itself from the other’s progress. The findings serve as a reminder that in the race for AI dominance, cooperation remains a significant factor.
In conclusion, while geopolitical tensions dominate headlines, US-China collaboration in AI continues to thrive. This reality underscores the interconnected nature of the global scientific community and the mutual benefits of shared innovation, despite political pressures.
