Poor management of artificial intelligence projects in the Department of Defense could erode the United States’ competitive advantage in the emerging technology, the Defense Department’s watchdog warned in a July 1 report.
The DoD inspector general suggested the Joint Artificial Intelligence Center, established to facilitate the adoption of artificial intelligence tools across the department, take several steps to improve project management, including determining a standard definition of artificial intelligence, improving data sharing and developing a process to accurately track artificial intelligence programs. The JAIC missed a March 2020 deadline to release a governance framework. It still plans to do so, according to the report, but that date is redacted in the report.
The inspector general started the audit to determine the gaps and weaknesses in the department’s enterprise-wide AI governance, the responsibility of the JAIC. After starting its audit, the DoD IG determined the organization had not yet developed an department-wide AI governance framework.
“If the DoD does not develop an AI governance framework in a timely manner, there is an increased risk that the DoD will lose its opportunity to become a strong, technologically-advanced Department, which is essential for protecting U.S. service members; safeguarding U.S. citizens; defending allies and partners; and improving the affordability, effectiveness, and speed of our operations,” the inspector general wrote.
The fiscal 2019 National Defense Authorization Act directed the department to establish a standard definition of artificial intelligence by August last year, but the watchdog found that as of March 2020 the department had not created one. The DoD CIO’s office told the inspector general that it didn’t think a department-wide definition wouldn’t be helpful in part because it “does not align to any specific technical, operational, or programmatic requirement,” according to the report.
The inspector general however found that each service had differing definitions that “created a challenge in determining what is and is not considered an AI project.” The IG detailed miscommunication between different officials at research labs on whether certain projects were actually AI projects.
For example, the report cites a missile project at the Air Force Research Laboratory that a technical manager identified as an AI project because of autonomous technology and algorithms, while the project manager said the project was an autonomy project using predefined flight formations.
“Without a clear and standard AI definition, the DoD’s AI oversight and governance could be applied inconsistently across the DoD,” the inspector general warned.
The IG report also found the JAIC needs a process to maintain insight into artificial intelligence activity going on across the Defense Department. The DoD CIO office told the inspector general in August last year that it has required DoD offices to report AI projects in its annual Information Technology/Cyberspace Activities budget exhibit and it will establish a biannual AI portfolio review. The first review was scheduled for “mid-2020,” the report said.
“An AI inventory management process for identifying and developing a baseline of AI projects is necessary to maintain awareness of the types and number of AI projects across the DoD,” the inspector general wrote.
JAIC also should promote collaboration on AI projects across the services. It found that the Marine Corps and Army are both working on a project to identify service members who are likely at risk of suicide, but are not collaborating. If the services did collaborate, they may develop a technology that “might be suitable for any Military Service.” Other benefits identified were cost savings and transparency.
A data and tool repository would also reduce the cost of AI projects and improve secure data sharing, the IG recommended. The DoD CIO office, where the JAIC is housed, agreed and added that it is working on creating the Joint Common Foundation. The JCF will be a massive collaborative environment that spanning multiple classification levels that will allow for sharing tools, algorithms, source code and models, along with several other capabilities.
Collaboration between the JAIC and other DoD components is a goal of the organization. The JAIC’s former director, now-retired Lt. Gen. Jack Shanahan, said that the JAIC needed to prove its value so service leaders and combatant commanders would seek out the JAIC.
“We have to have the combatant commanders in the service knocking at our door saying: ‘We can’t afford [for] you to go away, we need you to provide too desperately to cut you,’ ” Shanahan said before retiring in June.
The JAIC has so far worked on disaster relief and predictive maintenance projects since its inception two years ago. This year, it has started its first lethality project, which Shanahan said would start by solving low level problems, like calls for fires.
Lawmakers on Capitol Hill are also trying to boost the JAIC’s stature. The House version of the fiscal 2021 NDAA would elevate the JAIC out of the DoD CIO’s office and have it report directly to the deputy secretary of defense.
“Our ability to apply AI and other emerging technologies faster than our adversaries will allow us to maintain our competitive edge over Russia and China,” said Rep. Elise Stefanik, R-N.Y., ranking member of the House Armed Services Committee’s Subcommittee on Intelligence and Emerging Threats and Capabilities.
All Rights Reserved for Andrew Eversden