America’s Got AI Talent: US’ Big Lead in AI Research Is Built on Importing Researchers

Understanding the current state of international competition in artificial intelligence (AI) requires a grounded assessment of one of AI’s key inputs: research talent. Bolstering national capacity in AI and associated fields requires cultivating researchers who can develop new machine learning (ML) techniques and make key theoretical breakthroughs. 

We are interested in the human intelligence behind artificial intelligence. Attracting and retaining this talent sits at the heart of some of the most contentious aspects of industrial, innovation, and immigration policy. 

But sound and balanced policy decisions require good data—a resource that has been woefully lacking in debates over national AI capabilities. This is why MacroPolo has released the “Global AI Talent Tracker,” a data-driven assessment of the global balance and flow of top-tier AI researchers. 

Tracking the Top Tier

Building on our research from 2018 and 2019, we used authors of papers accepted at what many industry insiders consider the top research conference: the Conference on Neural Information Processing Systems, commonly known as NeurIPS. By gathering extensive data across numerous parameters, we have constructed what we believe to be one of the most comprehensive looks at top-tier AI researchers (approximately the top 20%). For more on the study, see our detailed Methodology.

We focus on examining the top-tier AI researchers because this is the cohort that has the ability to lead the way on important new areas of research, as well as to apply AI to highly complex real-world problems. 

With the dataset we sought to answer three key questions: 1) Where do top-tier AI researchers work today? 2) What path did they take to get there? 3) Where do top-tier AI researchers come from? Answering these questions yields insight into not only the current balance of AI research capabilities, but also maps the international talent flows that underpin those capabilities. We begin with key takeaways from the data, followed by a look at the implications for public policy.

Key Takeaways

1. The United States has a substantial lead over all other countries in top-tier AI researchers, and that lead has been built on attracting international students and researchers to work at US institutions.
2. China is the largest source of top-tier AI researchers, but a majority of these Chinese researchers leave China to study, work, and live in the United States.
3. Over half of all top-tier AI researchers are immigrants or foreign nationals, researchers working in a country different from where they received their undergraduate degree.

Takeaway #1: The United States has a substantial lead over all other countries in top-tier AI researchers, and that lead has been built on attracting international students and researchers to work at US institutions.

American institutions employ nearly 60% of all top-tier researchers, approximately six times as much as the next leading country (China, 10.6%) or region (Europe, 10.2%). 

But just as notable is the source of this talent: two-thirds of the top-tier research papers at US institutions are produced by scientists who received undergraduate education in other countries. Without researchers from abroad, America’s lead on talent would likely be considerably diminished.

Of the international researchers who eventually work in US-based companies and institutions, a majority originally came to the United States to attend graduate school. The largest share of these international researchers is from China.

Figure 1: Undergraduate countries of Top-Tier AI Researchers at US InstitutionsNote: Sample population is researchers at institutions whose headquarters is in the United States. This includes current graduate students. 

Takeaway #2: China is the largest source of top-tier AI researchers, but a majority of these Chinese researchers leave China to study, work, and live in the United States.

China stands out as the largest global source for the top-tier AI talent pool, with nearly one-third of these researchers globally completing their undergraduate education in China. Of all universities globally, Tsinghua University leads by a wide margin in producing undergraduates who eventually go on to conduct top-tier AI research. 

But following undergraduate education, China experiences a brain drain of this talent pool. Only 34% of these Chinese researchers are currently in China, while approximately 56% are in the United States. After completing graduate studies in the United States, a full 88% of those Chinese researchers chose to stay and work in the country, while only 10% headed back to China. (This sample includes a combination of recent graduates, mid-career researchers, and veteran researchers to reflect average stay-rates across all these groups.) 

Figure 2: Current Country of Top-Tier AI Researchers from ChinaNote: Sample population is top-tier AI researchers who completed their undergraduate degree in China. Current country is based on the geographic location of the researchers, not their institution’s headquarters. 

Takeaway #3: Over half of all top-tier AI researchers are immigrants or foreign nationals, researchers working in a different country from where they received their undergraduate degree.

A full 53% of top-tier AI researchers do not work in the same country where they received their undergraduate training. This group represents a highly skilled but mobile pool of cutting-edge research talent, one that countries around the world are competing to attract.

Among this population of top-tier immigrant researchers, China and India make up over half. Top-tier AI researchers who receive their undergraduate degree in the United States are highly unlikely to go abroad for work or further study, with just one in ten US undergraduates currently working outside the United States.

Figure 3. Country of origin of immigrant and foreign national AI researchersNote: Sample population is researchers who currently study or work in a country different from where they received their undergraduate degree.

How Policy Might Affect Talent Flows

Growing US-China technology tensions, coupled with the economic fallout from Covid-19, has led to a series of proposed changes to US immigration policy that directly impact international researchers in general, and Chinese STEM researchers in particular. 

These proposals include the potential suspension of the OPT program, which allows foreign students to work in the United States for 1-3 years following graduation, as well as legislation that would ban visas for Chinese graduate students in any STEM field.

While these policy proposals are being deliberated, we can use our dataset to test some hypotheticals and see how they might affect the US lead in top-tier AI research. For the purposes of this exercise, we assume that all affected researchers would return to their country of origin. (In our sample, this is represented by where they received their undergraduate degrees.). 

1) If visas were rescinded for all international students…

…then that would reduce the number of top-tier AI students in the United States by more than 60% and reduce America’s overall pool of AI researchers by nearly 20%. China and India would be the two countries that receive the largest numbers of these returnees.

2) If visas were rescinded only for Chinese students… 

…then that would reduce the number of top-tier AI students in the United States by 32%, and it would more than triple the number of top-tier AI students in China. (Chinese students account for almost half of the top international AI students in the United States.) 

3) If visas and immigration status were rescinded for all researchers who originated from China and now study or work in America…

…then America would lose 29% of its top-tier AI researchers, while China would more than double its number of these researchers. Taken together, this intervention would have a dramatic effect on the current balance of international AI research capabilities, as it would shrink the current US lead over China in top-tier AI researchers by two-thirds.

Conclusion

The above data and hypotheticals illuminate a system of international talent movement on which significant progress in AI depends. It also shows how the current playing field tilts strongly in favor of the United States. 

But that US lead could be disrupted by policy decisions that put a halt to, or even reverse, these global flows. Such decisions could have the unintended consequence of providing more opportunities for other countries—both US allies and competitors—to absorb this highly mobile pool of AI talent.

Future AI breakthroughs depend on the efforts of these researchers. But where those breakthroughs happen depends increasingly on the decisions of policymakers.