AI Rivals, Unlikely Partners: How US-China Collaboration Defies the Geopolitical Divide

While Washington and Beijing trade regulatory blows, their AI researchers are quietly building a bridge across the Pacific—one paper at a time.

In 2023, 141 out of 5,290 NeurIPS papers (3%) involved US-China collaboration. The pattern continued in 2024 with 134 out of 4,497 papers showing similar cross-border partnerships.

Google’s transformer architecture appeared in 292 Chinese-authored papers, while Meta’s Llama models were cited in 106. Alibaba’s Qwen model found its way into 63 US-authored papers.

George Washington University’s Jeffrey Ding noted: “Whether policymakers on both sides like it or not, the US and Chinese AI ecosystems are inextricably enmeshed—and both benefit from the arrangement.”

OpenAI’s Codex model automated 80% of the paper analysis process via Python scripts, streamlining the study of these technical interdependencies without overhyping its role in the broader geopolitical context.