International Collaboration

Inside leading AI labs one will often find diverse researchers that hail from numerous countries, who also frequently collaborate on research across national borders. But both geopolitical tensions and ethical considerations have put new strains on transnational AI research, potentially jeopardizing the future of collaborative activities.

In the charts below, we present data on two different types of international collaboration:

  1. Researcher-level collaboration: this refers to collaboration between researchers who come from different countries (e.g. when one Google researcher originally from India and one originally from the United States co-author a research paper).
  2. Institution-level collaboration: this refers to collaboration between researchers whose respective institutions are headquartered in different countries (e.g. when a researcher at Microsoft China co-authors a paper with a researcher from Tsinghua University).

Note: All charts below are based on data drawn from papers accepted at NeurIPS 2019. We use the country where a researcher attended undergrad as an imperfect proxy for their country of origin. For more on the study, see the detailed Methodology.

What portion of top-tier AI research papers involve collaboration between researchers who come from different countries?

Nearly two-thirds of these papers involve teams of researchers who completed undergraduate in different countries, highlighting the role that transnational migration plays in fueling top-tier AI research. The most common pairings of researcher-level collaboration are Chinese and American, followed by Indian and American, and then Indian and European.

What portion of top-tier AI research papers involve collaboration between institutions headquartered in different countries?

Almost one-third of papers involve collaboration between researchers whose institutions are headquartered in different countries. An example of this type of collaboration would be a paper co-authored between two researchers, one affiliated with Microsoft Research (US headquarters) and one affiliated with Tsinghua University (China headquarters).

What is the composition of the research teams producing top-tier AI research for US institutions?
Note: “US AI research” is defined as research papers whose authors are all affiliated with institutions headquartered in the United States. “Immigrant researchers” are defined as researchers who completed their undergraduate studies in countries other than the United States, but who currently work for US institutions. “US-origin researchers” are those that completed their undergraduate studies in the United States.

Nearly one-third of top AI research papers at US institutions are produced by teams composed entirely of immigrant researchers. This compares with just 22% of papers produced by teams of exclusively US-origin researchers, and over 45% of papers authored by mixed teams of US-origin and immigrant researchers. In other words, immigrant researchers contribute to roughly 80% of top AI papers from US institutions.