On February 11, 2019, the President of the United States signed an Executive Order (EO) directing the federal government to promote artificial intelligence (AI). President Trump’s EO aims to stimulate basic research in AI, reduce barriers to innovation, train AI technologists, and protect America’s advantage in AI such as it is. The EO also directs U.S. agencies to look for ways to make federal data more accessible and AI-friendly, i.e., to “prioritize improvements to access and quality of AI data and models.”
I applaud the White House for joining the previous Administration, and governments across the world, in recognizing the importance of AI to contemporary society and the need to establish federal and international policy. I also appreciate the EO’s repeated nods to civil liberties and privacy. Nevertheless, I see at least two issues with the EO: it’s mostly rhetoric, and that rhetoric isn’t ideal.
The EO doesn’t actually accomplish anything of substance. As counterintuitive as this may seem, Presidents are limited in their ability to control federal agencies. Yes, with respect to the military, the President is the Commander in Chief. And yes, the President can remove the heads of non-independent agencies at will. But the President generally cannot force agencies, which after all are creatures of Congress beholden to organic statutes, to adopt particular policies. And indeed, the EO is replete with language such as “consistent with applicable law” and “as appropriate” for this reason.
More fundamentally, the President does not control the national budget. Agencies such as the National Institutes of Health or the National Science Foundation have some programmatic discretion and could in theory channel certain resources toward AI research. But unless the President successfully declares an AI emergency, any significant investment in AI research and development (R&D) will have to come from Congress. That is why the EO directs agencies to “take this priority into account when developing budget proposals.” With a mere proposal, though, there is no guarantee of funding or further action.
Not only is the EO mostly rhetoric, some of its rhetorical choices are far from ideal. The EO makes repeated reference to something called “American AI.” I have no idea what this means. AI is not a specific technology but rather a set of techniques, aimed at approximating some aspect of human or animal cognition. While many innovations in this broad and diverse field originated in the United States, others did not. Arguably, machine learning owes its recent renaissance to advances by French and Canadian computer scientists.
The construct of American AI—and the attendant notion that America is involved in an AI “race” or “competition” with rival nations—has consequences for AI policy and for the development of AI itself. Consider a recent Department of Commerce proposal to restrict the export of machine learning to other nations. The notion that we should restrict not only the export of AI-enabled weapons, but the very flow of ideas around machine learning, only makes sense—really, is only coherent—if we buy into the rhetoric that artificial intelligence represents a high stakes competitive enterprise with winners and losers.
Similarly, consider the EO’s suggestion that agencies prioritize American citizenship in supporting students and faculty to conduct AI research. Setting aside meritocratic or constitutional values, such talk reveals a wide disconnect between a nativist Administration and the realities of higher education. I cannot think of a serious engineering department that is not replete with, and enriched by, faculty and students from abroad. The same is true in industry, which is why our technology companies fear visa restrictions.
I agree that AI should be a national priority. Were I, unlike the President, able to bend federal agencies entirely to my will, there are certain clear priorities I would set. The first would be to direct agencies to accrue additional expertise in AI and cyber-physical systems. I am not alone in observing that government lags behind academy and industry in technologically expertise at its peril.
The second would be to overhaul procurement. No government should purchase AI-enabled technology of any kind without understanding how it works, and without insisting upon mechanisms by which to audit and challenge AI-enabled decision-making. It defies imagination that a federal, state, or municipal governments would purchase and deploy algorithmic products without obligating the vendor to waive trade secret protection in the event of a challenge. We should also recognize the extent to which governments make markets; by insisting on transparency and accountability in the services they procure, our elected and appointed officials can drive industry in an overall more responsible and responsive direction.
Ultimately, however, the real audience for AI policy is the legislature. It is Congress that can allocate funding for AI R&D. It is Congress that can direct agencies to accrue expertise and adopt new technology. And it is Congress that can thicken the social safety net, require safety standards, and, along with the courts, safeguard our civil liberties and values in a time of great technological change. Arguably the most significant feature of the recent EO is to remind us that the ball is in Congress’ court.
Ryan Calo is the Lane Powell and D. Wayne Gittinger Associate Professor at the University of Washington School of Law. He is a faculty co-director (with Batya Friedman and Tadayoshi Kohno) of the University of Washington Tech Policy Lab, a unique, interdisciplinary research unit that spans the School of Law, Information School, and Paul G. Allen School of Computer Science and Engineering. Professor Calo has testified about robotics and other technologies before the full Judiciary and Commerce Committees of the United States Senate and the German Parliament and has organized events on behalf of the National Science Foundation, the National Academy of Sciences, and the Obama White House.
Suggested Citation: Ryan Calo, President Trump and the Myth of American AI, JURIST – Academic Commentary, March 22, 2019, https://www.jurist.org/commentary/2019/03/ryan-calo-ai-executive-order/
This article was prepared for publication by Stephanie Sundier, a JURIST Staff Editor. Please direct any questions or comments to her at commentary@jurist.org