US-China AI Collaboration Defies Rivalry: 3% of NeurIPS Papers Show Joint Research Between Tech Superpowers

In a surprising twist to the narrative of US-China technological rivalry, a comprehensive WIRED analysis of 5,290 AI research papers presented at the prestigious Neural Information Processing Systems (NeurIPS) conference reveals that the world’s two AI superpowers continue to collaborate extensively despite mounting geopolitical tensions.

The numbers tell a compelling story: 141 papers—approximately 3 percent of the total—featured authors from both US and Chinese institutions. This collaboration rate remained remarkably stable, with 134 joint papers out of 4,497 total submissions in 2024 alone.

“This finding challenges the conventional wisdom about US-China technological decoupling,” says Jeffrey Ding, assistant professor at George Washington University and leading tracker of China’s AI landscape. “Whether policymakers on both sides like it or not, the US and Chinese AI ecosystems are inextricably enmeshed—and both benefit from the arrangement.”

The collaboration extends beyond mere institutional partnerships. Cutting-edge innovations flow freely across the Pacific, with American-developed technologies finding their way into Chinese research and vice versa. The transformer architecture, pioneered by Google researchers and now foundational to modern AI, appears in 292 papers with Chinese institutional authors. Meta’s Llama family of models features prominently in 106 Chinese-authored papers, while Alibaba’s increasingly popular Qwen model appears in 63 papers involving US organizations.

These partnerships reflect deeper cultural and educational bonds. Many Chinese-born researchers complete their graduate studies in the United States, establishing professional relationships that persist throughout their careers. Katherine Gorman, spokesperson for NeurIPS, emphasizes this point: “NeurIPS itself is an example of international collaboration and a testament to its importance in our field. Collaborations between students and advisors often continue long after the student has left their university.”

The analysis methodology itself showcases AI’s potential to streamline complex research tasks. Using OpenAI’s Codex model, WIRED’s Will Knight automated the process of analyzing thousands of research papers, demonstrating how AI coding assistants can handle labor-intensive analytical work that would otherwise be prohibitively time-consuming.

This collaborative spirit persists even as US politicians and tech executives increasingly cite China’s technological rise to justify regulatory rollbacks and massive AI investments. The findings serve as a crucial reminder that despite the rhetoric of competition, both nations stand to gain significantly from continued cooperation in advancing artificial intelligence.

As the AI arms race narrative dominates headlines, these numbers reveal a more nuanced reality: the world’s two AI superpowers remain deeply interconnected, with scientific progress benefiting from their continued collaboration rather than suffering from their rivalry.

AI #USChinaRelations #ArtificialIntelligence #NeurIPS #TechCollaboration #AIResearch #GlobalInnovation #MachineLearning #TechnologyPolicy #ResearchPartnerships

China AI collaboration, US-China tech partnership, NeurIPS conference analysis, transformer architecture, Llama models, Qwen AI, international AI research, technological decoupling, AI ecosystem collaboration, cross-Pacific innovation

“US and Chinese AI ecosystems are inextricably enmeshed”
“Both benefit from the arrangement”
“Scientific progress benefits from continued collaboration”
“Cutting-edge innovations flow freely across the Pacific”
“Deep cultural and educational bonds persist”
“Competition narrative masks underlying cooperation”
“Geopolitical tensions haven’t stopped scientific exchange”
“3% of top AI papers show US-China collaboration”
“AI superpowers still have much to gain from working together”

,