To win the AI race, open America's doors

To win the AI race, open America's doors
© Getty Images

Policymakers’ preoccupation with “winning the AI race” is reaching a fever pitch. Recent projections state that artificial intelligence will add 16 percent, or $13 trillion, to global economic output by 2030. Economists worry how big the U.S. share will be. National security experts decry that China or Russia will use AI technology to do everything from building more sophisticated weapons to controlling our elections.

They’re right: winning the AI race is critical. So why is the United States giving other nations a head start?


Most experts tend to focus on three essential pieces of the AI puzzle: high-quality data, sophisticated algorithms and computational power. This misses the point. What America needs are more qualified people who can create this infrastructure wherever they go. We need to win the international race for AI talent, beginning with a more open immigration system and resources to help immigrants build a life in the United States.

The stakes are high. Russia’s Vladimir Putin declared in 2017 that “whoever becomes the leader in [AI] will become the ruler of the world.” In 2017, China announced its “New Generation Artificial Intelligence Plan,” which envisions catching up to the United States by 2020 and becoming the world leader in AI technology and applications by 2030. While America still has the lead in most metrics, China is closing the gap on the number of AI companies, patents and published research papers.

To his credit, President TrumpDonald TrumpFormer New York state Senate candidate charged in riot Trump called acting attorney general almost daily to push election voter fraud claim: report GOP senator clashes with radio caller who wants identity of cop who shot Babbitt MORE recognizes AI’s importance to American competitiveness. However, his recently issued “Executive Order on Maintaining American Leadership in Artificial Intelligence” is predicated on the conventional wisdom that “to turn these [AI] ideas into reality, we need infrastructure … that means data, models, and computational resources.”

The Information Age is not about infrastructure; it’s about people. Uniquely talented people develop the high-performance computing resources and machine-learning algorithms that run AI projects. Unfortunately, the global demand for AI talent falls far short of supply. A recent study by a Canadian start-up, for example, assessed that the total supply of AI experts around the world was only 22,000. The Chinese tech giant Tencent put the figure much higher, between 200,000 to 300,000, but against a market demand for millions of AI researchers and practitioners. Microsoft alone has 2,000 open job listings for AI-related positions. American tech firms clearly realize they are in a cutthroat race for talent, paying from $300,000 to $500,000 a year to AI specialists.

Yet the United States is making it harder for highly-skilled immigrants to fill these jobs. The number of petitions for H-1B visas fell by 15 percent in 2018, the first decline in five years. Immigration experts attribute the decline to regulatory and administrative changes, including a significant increase in requests for additional evidence and interviews to process visa applications, as well as proposals to eliminate work authorization for spouses of high-skilled workers.

The alternative to attracting specialists from abroad is educating more Americans in AI-related fields — something President Trump’s executive order calls for. This is a fine long-term strategy, but the current system does not have the capacity to groom a sufficient number of home-grown experts. Talent-starved tech companies are luring away university professors responsible for training the next generation of AI talent with skyrocketing private-sector salaries, perpetuating the talent shortage. Nearly 60 percent of U.S. computer science PhDs take industry jobs after graduation, up from 38 percent over the last decade. A 2016 KPMG report found that six organizations employ over half of all deep learning specialists, of which none are universities.

Faced with a shortfall of American talent, America’s top tech firms have decided that they need to move significant parts of their AI R&D abroad. Five of China’s top 10 employers in AI are American companies, including IBM, Intel and Microsoft. That might be acceptable for big tech’s bottom line, but do we want American firms developing China’s AI talent and workforce?

Given the dearth of AI talent globally, the United States must secure a competitive advantage where it matters most. We need a foundation of policies and practices that continue to attract top talent to America, such as the heads of AI at Apple, Facebook, Microsoft and Google’s cloud computing division, all of whom were born outside the United States. At a minimum, Congress should ensure that any foreign AI expert is able to quickly obtain an H-1B visa and that foreign students specializing in AI at American universities are immediately eligible for a path to citizenship.

And if the United States truly wants to gain global strategic advantage in AI, the government should partner with the private sector to identify the top 1,000 AI experts in the world and offer them jobs, green cards, and resettlement packages to move to America.

Since its founding, America has achieved its economic and security interests by attracting and welcoming the smartest, most hardworking people in the world. Winning the AI race should be no different.

Eric Rosenbach, former Pentagon "cyber czar," is co-director of Harvard Kennedy School’s Belfer Center for Science and International Affairs. Follow him on Twitter @ERosenbach.

Aditi Kumar, former principal at Oliver Wyman, is the Belfer Center’s executive director. Follow her on Twitter @aditikmr.