Cybersecurity & Tech Surveillance & Privacy

Trump’s Artificial Intelligence Strategy: Aspirations Without Teeth

Caleb Watney
Wednesday, February 20, 2019, 9:28 AM

On Feb. 11, the White House released an executive order on “Maintaining American Leadership in Artificial Intelligence” (AI)—the latest attempt to develop a national strategy for AI.

President Trump signs the Executive Order on “Maintaining American Leadership in Artificial Intelligence” (Source: US Mission to the OECD)

Published by The Lawfare Institute
in Cooperation With
Brookings

On Feb. 11, the White House released an executive order on “Maintaining American Leadership in Artificial Intelligence” (AI)—the latest attempt to develop a national strategy for AI. The order envisions the United States taking significant steps to increase research and development efforts while reforming its executive agencies to better compete with the Chinese government’s investments in AI development through its Made in China 2025 plan. Although the order is full of promising language and constructive suggestions for executive agencies, it is unlikely to have much of a long-term effect without further support from Congress.

The executive order has three basic prongs. First, it charges executive agencies to “prioritize AI” across several dimensions. Essentially, if a department handles research and development, it’s encouraged to put work on AI at the top of the queue. If it issues educational grants, the department should emphasize programs that increase apprenticeships and educational opportunities that build science, technology, engineering and mathematics (STEM) skills. If a department regulates products that incorporate AI, department heads should be on the lookout for ways to reduce entry barriers and promote innovation while addressing new challenges. Given the broad scope of this prioritization effort, the order could theoretically have an impact on a wide swath of existing agencies. Research and development funds are concentrated primarily within eight different agencies, but many other agencies regulate parts of the economy on which AI is encroaching.

Second, the order asks all agencies to identify high-priority federal datasets and models that could have useful commercial applications, and to increase the public’s ability to access these resources. This is important because the federal government owns datasets that are difficult to acquire otherwise and that could help U.S. companies train and develop more sophisticated AI models with the information made available. The order outsources to the public, through an upcoming comment period, the question of which federal datasets would be of most value.

Finally, the order promises the creation of several new guidance documents, through agencies such as the National Science and Technology Council and the National Institute of Standards and Technology, which will shape future regulation and cultivation of AI in the United States. Notably, the order does all of this without promising any additional agency funding.

Most of these initiatives are important and, if successfully implemented, could do a great deal to increase the degree to which AI improvements diffuse throughout the economy while promoting a healthy, competitive ecosystem. The most useful policy framework for accomplishing these goals is a relentless focus on reducing the barriers to entry in AI development. Increasing the supply of government datasets available to the public and improving domestic talent pipelines would help in this regard, as the scarcity of talent and of high-quality data are key bottlenecks to AI progress. So it is encouraging to see the administration begin to take on these issues directly.

But take a step back, and the overall structure of the order looks a bit fuzzy. While U.S. agencies should be prioritizing the development of AI technology and a skilled workforce to match, agency heads are already being pulled in many directions at once, and one more document on their desks suggesting prioritization is unlikely to radically change their direction. As with many proposals to increase the quality of U.S. governance, the details are crucial, and aspirational statements without enforcement mechanisms are unlikely to suffice.

To bridge the gap between aspiration and implementation, Congress should approve specific funding to increase AI research and development efforts and to create incentives to ensure that agencies are prioritizing the initiatives they are asked to prioritize. The U.S. must be willing to pay top dollar to recruit and keep the technical talent in federal agencies that can maintain these projects with quality. And there should be regular reporting requirements to ensure that both Congress and the public can track agency progress.

Some of these details are doubtless being outsourced to the new guidance documents that are coming down the pike. But these core structural issues cannot be addressed until Congress steps in with specific budgets and oversight mechanisms.

In addition to the lack of enforcement mechanisms, the executive order represents a missed opportunity to address a cornerstone issue: international talent. While it is critical that the U.S. develop better domestic talent pipelines for AI skills, tremendous talent is slipping away through the current immigration system. Every year, tens of thousands of international students pursue their graduate studies in STEM fields in the United States, but only a fraction are allowed to stay in the country upon completing their degrees. This means that a significant portion of the global AI talent is leaving the United States for competitor nations.

Here, again, Congress needs to step up and either expand the number of existing visas used by these students (primarily the H-1B program) or create a new visa classification for high-skilled tech workers allowing for easier entrance and permanent residence. The Trump administration could certainly make the path easier by reversing some of the president’s previous executive orders that have made it more difficult to acquire H-1B visas and for accompanying spouses to work in the U.S. themselves.

Talent development isn’t an either-or issue; the U.S. national strategy for AI can both build better domestic talent pathways and embrace America’s strength as the world’s melting pot for technical researchers and practitioners. Demand for technical talent is far outstripping supply at the moment, and to remain competitive internationally, the United States needs all hands on deck.

AI promises to be the next true general purpose technology, meaning it could power a whole wave of productivity enhancements across the economy. And it also has dual-use implications, with each breakthrough in commercial applications likely having corresponding military applications. Meanwhile, China is investing heavily in research and development around AI, making reams of government data available to top domestic firms and actively recruiting back Chinese students from American universities because they recognize the strategic implications of leading in AI.
As the geopolitical consequences of technological superiority only continue to increase, the U.S. cannot afford to fall behind due to self-inflicted wounds, whether from a lack of implementation details or from ham-fisted immigration policies. The administration’s national strategy document represents a good first step in ensuring the United States remains competitive. But it will require more work (and attention to detail) from Congress to finish the job.


Caleb Watney is a Technology and Innovation Fellow at the R Street Institute and leads the Institute’s work on emerging technologies, including autonomous vehicles, artificial intelligence, drones and robotics.

Subscribe to Lawfare