To win the AI race, we need more humans

The United States is trying to develop a national strategy on artificial intelligence (AI) but is ignoring its single greatest advantage: many of the world’s brightest minds would like to live and work here but can’t. Just last night, President Trump met with some of America’s top CEO’s and expanding the supply of high skill visas was top of mind for them. Every year, tens of thousands of international students come to the United States to pursue graduate degrees in computer science fields but only a small fraction are allowed to stay in the country upon completion of their degree. To solidify our lead in the global AI development race, we must embrace our strength as the world’s melting pot for technical researchers and practitioners.

Over the past five to ten years, the pace of AI development has accelerated rapidly, with substantial gains in machine learning, which now powers driverless cars, personal voice assistants, autonomous drones and much more. While the United States has been on the cutting edge of technical research in this regard, China has also been in aggressive pursuit and has laid out plans to become the global leader in AI research by 2030.

{mosads}This matters because AI has many dual-use military applications—meaning advances in civilian technology can easily be applied in military domains. Furthermore, because AI can improve and automate aspects of human decision-making, it has the potential to power productivity growth across a wide domain of industries, acting like a multiplier on the strength of the entire economy. AI thus has significant strategic importance for the long-term outlook of both U.S. foreign and domestic policy.

If ballooning salaries are any indication, there is a significant shortage of AI talent in the United States today. Typical AI specialists can expect to earn between $300,000 — $500,000 at major tech companies; numbers that are significantly higher than those of their peers in other computer-related subfields. Industry experts like Google’s Hal Varian have pointed to the scarcity of AI talent as the primary bottleneck on AI development and application. And what makes AI talent particularly interesting from the perspective of international competition is its zero-sum nature relative to other factors of production. Unlike venture capital or hardware, two other important AI inputs, skilled researchers can’t easily flow across borders or be rapidly scaled up when new opportunities arise. Building up adequate talent pipelines may eventually solve this problem but such a scenario could take many years to develop. In the meantime, the countries with the most AI talent will be the ones with the most AI progress. 

The United States, however, already has a significant advantage in its strong university and education system, which regularly attracts tens of thousands of top graduate students from around the world. In 2015, for instance, the nation had 58,000 graduate students in computer science fields, the overwhelming majority of which (79%) are international students. Unfortunately, only a subset of these students are able to stay in the country long term, due to the small and fixed cap of H-1B visas, the primary immigration pathway available for high-skill workers. For the past 16 years, the H-1B visa limit has been exhausted and the number of applications filed has consistently has been twice that of the available spots.    

China, meanwhile, is well aware of the importance of skilled talent and is working actively to encourage Chinese students and AI researchers to return to China. The Chinese government frequently lures these potential startup founders back home with the promise of free housing, free office space and equity-free financial backing. As a top Chinese venture capitalist told the Financial Times: “We still believe the top talent is there [at US companies and universities] so we are still trying to bring them back.” And indeed, their efforts appear to be working. In the early 2000’s, only 1 in 10 Chinese students returned home after studying abroad, today it’s 8 in 10.

Allowing more international students to live and work in the United States upon completion of their degree, either through an expansion of the H-1B visa program or through the creation of a new technical worker visa program, would be a relatively straightforward and effective method to alleviate the country’s talent shortage around AI. Either of these options would also strengthen our position relative to China’s, given the zero-sum nature of talent acquisition.

Such reforms would have the added bonus of increasing domestic competition around AI development and application. As long as acquiring top AI talent requires hundreds of thousands of dollars per person and an HR staff capable of braving the immigration-visa bureaucracy, it will be incredibly difficult for anyone other than large tech giants like Google, Amazon and Facebook to afford and access top AI talent. Making the path easier for startups and smaller competitors to hire AI practitioners would be an easy way to increase the internal competitiveness of the U.S. tech market as well.

Put simply, cutting-edge AI will be developed somewhere. It’s in the best interests of the United States if that somewhere is here. At least in principle, both political parties recognize the importance of skilled immigration for our international competitiveness. As Congress considers how the nation should approach AI from a strategic perspective, the lowest hanging fruit is simply to allow the high-skill students already here to stay and work in the country that trained them.

Caleb Watney is a technology policy fellow at the nonprofit R Street Institute.