There’s hard work ahead to truly capture the full potential of artificial intelligence.
The federal government has laid out a path forward on artificial intelligence that could help our nation retain our technological edge. But these plans must be followed by investment and action to ensure success, as designing and implementing AI is not easy.
On February 11, the President signed an executive order launching the American AI Initiative, which aims to stimulate AI advancement. The Defense Department followed the next day, releasing an unclassified AI strategy that galvanizes the role of the Joint Artificial Intelligence Center. Most recently, the proposed fiscal 2020 White House budget includes hundreds of millions in potential funding to support the JAIC, as well as AI research at both the Energy Department and the National Institute of Standards and Technology. And recently a bipartisan group of Senators proposed more than $2 billion in funding as part of the Artificial Intelligence Initiative Act.
This flurry of activity sets the pace for a busy few months ahead for federal agencies that must now identify opportunities to use AI, and then design and operationalize it. Critical decisions lie ahead for federal leaders to set goals, reallocate resources and prioritize AI investments.
AI is already being used successfully across the federal government including to thwart cyberattacks, assist disaster relief, fight fraud and so much more. For example, AI can help analysts more quickly spot fraud by using old fraud data to shape predictive modeling capabilities For one agency, this tactic has already prevented over $2 million in false payments. For another, automated fraud and identity theft detection helped protect $6 billion in revenue.
However, there’s hard work ahead to move AI out of the lab and truly capture its full potential.
First, we are still in need of a national AI strategy that solidifies how the government, academia and private sector work together toward common goals to maintain U.S. AI leadership. More than 10 nations around the world have already prioritized AI through their own national strategies, with plans published by Canada, China, Japan and several others.
Key pillars of a strategy must include thoughtful public policy to spur AI adoption and reduce barriers, while still encouraging innovation. Currently, existing regulations or proposed export controls intended to protect American intellectual property often have unintended consequences, such as stifling innovation.
Second, there are some tactical steps that federal leaders developing AI implementation plans—collaborating with industry and academia—should consider, including:
- Galvanize your goals and values. Have you identified the value that you want to capture through the investment? Have you identified use cases that make a practical impact on your missions and serving citizens? What questions and answers have you developed to address ethical concerns such as respect, transparency, privacy, and equitable treatment of people?
- Take stock of your capabilities. Do you have the labeled data necessary to implement AI? How does your computing infrastructure—at the enterprise level and at the edge—need to evolve to capture and process the data you require? Do you have the right mix of expertise to accomplish your goals? How do upskill staff to gain AI technical competencies over time?
- Design for operationalization first because AI is hard. There are too many AI use cases that never make it out of the lab environment. Are you building agility into your MLOps (machine learning operations)? Do you have a plan to verify and validate models that are constantly learning? How are you going to manage inconsistent, outdated IT environments, sporadic networks, static and rigid software architectures? Planning for the challenges and risks that will arise in a machine learning and AI implementation
The administration has taken an essential step in retaining America’s technological edge by codifying the principles in the American AI Initiative. Now, federal leaders—along with academia and industry partners—must push ahead with renewed purpose to move from planning for concrete implementation that secures the promise of an AI-enabled future. If we do not put in the hard work to advance and implement AI solutions, we risk falling behind.
Josh Elliot is a principal at Booz Allen and serves as the firm’s director of artificial intelligence solutions.