The reintroduction of the Artificial Intelligence in Government Act this month is a much needed response to concerns that the United States is lagging far behind both foreign governments and American industry in reaping the promise and perils of artificial intelligence. Sponsored by a bipartisan group of senators, the bill promotes the adoption of artificial intelligence in the federal government, while addressing the potential negative consequences. A companion bill was introduced in the House.
Central to the bill is the creation of an Artificial Intelligence Center of Excellence within the General Services Administration, which will provide the technical expertise, research, and advice to federal agencies on the acquisition and use of artificial intelligence technology, including all of the accompanying “economic, policy, legal, and ethical challenges and implications.” A key part of its mission is to direct and assist the agencies in developing and maintaining governance plans for their use of artificial intelligence. Those plans are intended to promote the innovative use of artificial intelligence for the public benefit, while upholding “civil liberties, privacy, and civil rights” and identifying, assessing, and mitigating any “negative unintended consequences” including discrimination and more.
The bill also directs the Office of Personnel Management to identify the key skills necessary for related positions in the government and the General Services Administration chief to establish an advisory board of agency designees and representatives of outside entities to advise on the issues facing the Artificial Intelligence Center of Excellence. The bill is important precisely because American leadership in this area is vital to our economic success and national security. Like the executive order on artificial intelligence signed by President Trump, the bill is a necessary part of developing a national artificial intelligence strategy and addressing concerns that countries like China already have such a strategy in place.
The impact of artificial intelligence on the global economy and society grows every year, so the federal government needs to get its hands around the benefits and risks of the technology now rather than later. In recognition of that, the bill has been endorsed by Microsoft, Facebook, the Internet Association, and several other organizations that are focused on policy for artificial intelligence, including the Committee for Justice.
Several of the specifics in the bill are worthy of praise. For those of us who are concerned that excessive caution or too much regulation could stifle the promise of artificial intelligence, it is good to see that the bill directs consideration of ways to “reduce barriers” to its use. Also encouraging is the inclusion in the legislation of “assisting agencies in applying the management and use of data in applications of artificial intelligence” under the duties of the Artificial Intelligence Center for Excellence. The greatest advantage the government has in building artificial intelligence applications is access to large data sets. However, as a former artificial intelligence scientist, I am keenly aware that turning this advantage into successful systems is highly dependent on the skillful use of that data.
To be effective, the Artificial Intelligence Center for Excellence and the agency activities it directs must draw upon the wealth of often superior technical and policy expertise outside the federal government. The bill recognizes this by tasking the Artificial Intelligence Center for Excellence with promoting joint initiatives with industry, academia, nonprofits, state and local governments, and the like. The agencies are also required to solicit public feedback through public hearings and online submission of comments in developing their artificial intelligence governance plans.
Because artificial intelligence is a rapidly evolving technology, these governance plans must also evolve to keep up with changes. Accordingly, the legislation requires each agency to update its plan on a yearly basis. The requirement for an artificial intelligence governance plan applies to relevant federal agencies as determined by the Office of Management and Budget director. The director should broadly interpret this language to include virtually all federal agencies because the central purpose of the bill, which is to “advance the innovative use of artificial intelligence” in government, logically includes the consideration of artificial intelligence technology even by the federal agencies that are not currently using it.
The director is also tasked with choosing the members of the advisory board, who are to include the “representatives of public interest groups representing civil liberties, privacy, and civil rights issues.” Because it will be easy to find representatives who believe that artificial intelligence is a substantial threat to these cherished principles, the director should also seek other perspectives, including on how artificial intelligence can be used in positive ways to enhance privacy and decrease discrimination.
The many risks and rewards of federal adoption of artificial intelligence technology must be discussed and addressed if it is to be successful and beneficial to the American public. The sooner the Artificial Intelligence in Government Act becomes law, the sooner those discussions can begin.
Curt Levey is president of the Committee for Justice. He previously worked in the artificial intelligence field and pioneered technology for overcoming the black box nature of neural networks that was hindering their adoption.