Rivalry and cooperation: the paradox of global AI research
At a moment when AI is a centerpiece of USChina competition, a deep dive into research output reveals a web of collaboration that belies simple narratives of decoupling. By analyzing more than 5,000 papers from one of AI's premier conferences, Wired finds that partnerships across bordersespecially between American and Chinese institutionsare alive and well.Where cooperation shows up
- Roughly 3% of NeurIPS papers include co-authors from both US and Chinese labsa consistent figure over recent years.
- Foundational work on architectures like transformers and open-source models such as Meta's Llama appear in shared research, signaling mutual influence.
- Chinese models like Alibaba's Qwen increasingly participate in the broader academic conversation.
What this means for industry and policy
This isn't cooperation born of corporate alliances but of academic ecosystemsresearchers flock to where ideas are strongest. Even amid export controls and industrial policy pressure, the flow of knowledge and methodologies persists, which helps accelerate global progress but also complicates strategic competition.
For companies and policymakers, this suggests that technological advantage is not a zero-sum game; ideas traverse boundaries even when capital and markets do not.