President Donald Trump signed an executive order (EO) on December 3, providing guidance for federal agency adoption of artificial intelligence (AI) for government decision-making in a manner that protects the privacy and civil rights.
Emphasizing that ongoing adoption and acceptance of AI will depend significantly on public trust, the EO charges the Office of Management and Budget with charting a roadmap for policy guidance by May 2021 for how agencies should use AI technologies in all areas excluding national security and defense. The policy guidance should build upon and expand existent applicable policies addressing information technology design, development, and acquisition.
In order to standardize practices across agencies, the EO sets out nine principles to be applied when designing, acquiring, and using AI. These principles include that AI use must be lawful and respectful of the nation’s values; purposeful and performance-driven; accurate, reliable, and effective; safe, secure, and resilient; understandable; responsible and traceable; regularly monitored; transparent; and accountable.
The EO also directs interagency bodies, in coordination with the Chief Information Officers Council, to identify, provide guidance on, and make publicly available information on non-classified and non-sensitive use cases of AI by agencies, which may provide new insights into how federal agencies currently deploy AI technology as well as potential new growth areas as agencies work to bring their use into compliance with the new guidance. The push for new AI development also could lead to an expansion of federal personnel with AI expertise.
The EO is another measure in the Trump administration’s efforts to define policies for AI, and follows the Office of Management and Budget guidance for federal agency regulation of AI applications in private sector, issued on November 17. With few departures from the draft guidance published earlier this year, the core of the recent OMB guidance is to ensure that agency-implemented regulations do not “hamper AI innovation and growth.” To that end, the OMB guidance urges agencies to first assess the effects in order to avoid “regulatory and non-regulatory actions that needlessly hamper AI innovation and growth.” The OMB guidance also highlights the need for a top-down standardization “to address inconsistent, burdensome, and duplicative State laws’ that might interfere with a national market, but to avoid national regulatory action in instances where uniformity is not essential.
The OMB guidance comes 21 months after President Trump issued an executive order (Maintaining American Leadership in Artificial Intelligence), which marked the launch of the American AI Initiative and sought to accelerate AI development and regulation to secure the United States’ place as a global leader in AI technologies.
Congress wants to boost the prominence of the Pentagon’s AI center
Congress signaled its confidence in the Pentagon’s young artificial intelligence office through a series of measures that increase its standing in the agency, including giving its director acquisition authority.
The annual defense policy bill called the fiscal 2021 National Defense Authorization Act, would alter the reporting structure of the Joint Artificial Intelligence Center, raising the office to report directly to the deputy secretary of defense, instead of the department’s chief information officer. The bill, which still needs President Donald Trump’s approval, establishes a board of advisers to give the center strategic advice and technical expertise on AI matters.
The measures to bolster the importance of the Joint Artificial Intelligence Center come as the organization pivots from focusing on artificial intelligence projects to identifying and solving problems within the services using AI. The JAIC was established in 2018 to increase the adoption of AI across the Pentagon.
Until now, the office hasn’t had acquisition authority. The NDAA would authorize a maximum of $75 million for the JAIC director for the “development, acquisition and sustainment of artificial intelligence technologies, services and capabilities through fiscal year 2025.” This year the JAIC has repeatedly mentioned the challenges that the current acquisition process causes. Its contract awards usually relied on the General Services Administration or Defense Innovation Unit, an entity also meant to speed up the acquisition process.
More autonomy over acquisitions could streamline the process to get the services AI technology faster, said Lindsey Sheppard, a fellow at the Center for Strategic and International Studies.
“Finding available contracting vehicles can be a big challenge for technology development efforts,” she said in an email. “You may have identified a mission need and have a great solution, but no available contract vehicle. And it can take years to get something in place. This NDAA would clear that roadblock by giving the JAIC its own acquisition authorities to get technology in the door.”
Air Force Lt. Gen. Jack Shanahan, former director of the JAIC until he retired in June, called for the authority back in May. At the time, he said the lack of acquisition authorities was slowing the agency down when it needed to go faster.
Martijn Rasser, a senior fellow at the Center for a New American Security, told C4ISRNET that the new authority would help bring nontraditional contractors into the fold.
“This will enable JAIC to speed up the process and gives them the opportunity to level the playing field for small and nontraditional tech companies, which is key to ensuring DoD has access to the broadest array of AI solutions as possible,” Rasser said.
Elevating the JAIC to a direct report of the deputy secretary is an important step in recognizing the office’s importance, experts said, especially as tech priorities could change under a new presidential administration.
“The JAIC reporting directly to the deputy secretary of defense says that regardless of how that shuffling comes out in the next few months, AI will still be a significant priority and its place on the org chart reflects that,” Sheppard said.
The NDAA also would direct the Defense secretary to establish a board of advisers for the JAIC on technical issues, ethical challenges, and workforce issues related to AI use. The board, appointed by the secretary and made up of industry and academic experts, also would guide long-term AI studies and strategies. It would meet at least once a quarter and submit a report annually summarizing its work.
This has been an important year for the JAIC as it started its first warfighting initiative and pivoted to play a role in the DoD’s COVID-19 response. The center also had a change in leadership this year after Marine Corps Lt. Gen. Michael Groen took over from Shanahan.
Meanwhile, the JAIC also rolled out “JAIC 2.0,″ a realignment of its AI programs, called national mission initiatives, to better match warfighting needs and identify challenges in the services where AI can help.
“What we want to do is seek out problems,” Groen said last month. “If in JAIC 1.0, we built technologies and then tried to find a market for them, [then] in JAIC 2.0, we’re going to be problem-pull. We’re going to build the relationships across the department to help us understand where the most compelling problems are so then we can pull our technology development and enablement in that direction.”
Special Operations strives to use the power of artificial intelligence
US Special Operations Command hopes to increasingly use artificial intelligence and machine learning in all aspects of warfare, its commander said.
Army Gen. Richard D. Clarke spoke virtually today with Hudson Institute scholars.
Clarke noted that Project Maven jump-started the employment of AI. Project Maven was initially executed to automate the processing and exploitation of full-motion video collected by intelligence, instead of relying on humans to sort through all of it.
With AI’s ability to shift quickly through terabytes of data to find relevant pieces of intelligence, it allows the human to make faster and better-informed decisions, he said.
AI can also be incredibly effective at monitoring the information environment, he said.
During a recent visit with a special operations commander in Afghanistan, Clarke noted that the commander said influencing the population in a positive way can mean the difference between winning and losing.
Socom has been using AI for logistics, and the maintenance piece in particular, for more than two years now, he said. It saves money in terms of, for example, predicting engine life or failure on a tank or aircraft. And it allows better use of those assets.
AI-powered health care can predict injuries or point to treatments to get operators in the fight more quickly, he mentioned.
In the realm of mission command, AI will power the Joint All-Domain Command and Control system, which will allow commanders to better communicate and make decisions, he said.
While Socom is forging ahead quickly with AI, Clarke mentioned that his organization is also working closely with the military services and organizations like the Joint Artificial Intelligence Center, as well as with industry, allies, and partners.
Clarke emphasized that it’s important that commanders set the tone and set the conditions to allow innovation and encourage people to come up with great ideas.
Humans are more important than the hardware, he said. “It’s the talented people that we have to help foster. You’ve got to invest the human capital into this space.”
Looking to the future, Clarke said he is optimistic that AI will be successfully leveraged by the Defense Department to maintain the lead against peer competitors China and Russia. It will require updating policy and infrastructure, using cloud computing, and having the right people who are enabled with the right leadership.
