Artificial intelligence (AI) deployments in the Federal government are already making government smarter, based on examples shared during the second of a three-part series on AI launched last month by the House Subcommittee on Information Technology. Federal agency leaders from the Defense Advanced Research Projects Agency (DARPA), the General Services Administration (GSA), National Science Foundation (NSF), and Department of Homeland Security (DHS) discussed how AI is being implemented to improve the mission of their agencies and what is required to ensure the technology continues to be viewed as a problem solver.
Cautioning against overstating its capabilities and suggesting that AI research should begin with a problem that must be solved, John Everett, deputy director, Information Innovation Office at DARPA said, “I think there’s a temptation to think of AI as magic, and capable of solving all our problems.” “When it comes to implementing something that would be effective for the Federal government, I think it’s important to take the perspective of first understanding what the actual problems are and then working our way back on how AI can actually address those problems.”
James Kurose, assistant director of NSF for Computer and Information Science and Engineering noted that AI research required experts to impact mission.
“NSF prioritizes its AI investments through a proven, ‘bottom-up’ philosophy: the best ideas for research come directly from the science and engineering community,” Kurose said.
Improving Government Services
The Federal government is constantly looking for ways to improve efficiency and accuracy, and AI can help with both. GSA is using AI to achieve process automation and data-driven decision making in the Federal acquisition process.
“GSA’s Office of Government-wide Policy (OGP) has developed a new pilot using AI for Prediction of Regulatory Compliance, known as the Solicitation Review Tool (SRT),” said Keith Nakasone, deputy assistant commissioner, acquisition operations, Office of Information Technology Category, GSA. “The SRT AI platform uses natural language processing, text mining, and machine learning algorithms to automatically predict whether Federal solicitations posted on fbo.gov are compliant with Section 508 of the Rehabilitation Act and alert responsible parties of non-compliance so that corrective actions could be taken.”
The AI-powered predictions have an accuracy rate of 95 percent, according to an independent review. In addition to near super-human accuracy, the platform enables the agency to concentrate its human workforce on other mission-critical tasks–such as tackling non-compliant solicitations.
Using AI to Improve Citizens’ Lives
NSF is focused on funding AI-based research that enhances citizens’ lives through improved healthcare and public agency decision making.
Kurose noted that currently an NSF-funded team of researchers at Johns Hopkins University is developing an AI program integrating data from health records to predict factors leading to septic shock. According to Kurose, “the program was able to accurately predict septic shock 85 percent of the time, usually before it harmed any organs.”
NSF-funded researchers are also using AI to address quality of life and livability issues in America’s cities. Kurose shared that, “The Texas Advanced Computing Center, the University of Texas Center for Transportation Research, and the City of Austin have also developed a new deep learning tool that uses raw camera footage from City of Austin cameras coupled with high-performance computing to recognize objects–people, cars, buses, trucks, bicycles, motorcycles, and traffic lights–and characterize how those objects move and interact. This information can then be analyzed and queried by traffic engineers and officials to better determine traffic patterns and potential safety issues.”
Shoring Up Cybersecurity With AI
The Department of Homeland Security (DHS) Science and Technology (S&T) Directorate believes AI can improve cybersecurity and minimize the risk of attacks and data loss. S&T is working with the private sector to test and deploy new cybersecurity solutions.
“A good example of S&T’s work involves demonstration of Telephony Denial of Service (TDoS) protection for a major U.S. bank with a significant impact on its contact center that processes close to 11 million calls per week,” said Douglas Maughan, division director, Cyber Security Division S&T Directorate, DHS. “The machine learning-based policy engine blocks more than 120,000 calls per month based on voice firewall policies including harassing callers, robocalls, and potential fraudulent calls. It also blocks two to three phone-based attacks each month (computer-generation of calls into 1-800 toll-free destinations in an attempt to collect a portion of the connection or per-minute charges associated with the call). This same technology can be used by 911 call centers to defend against denial of service attacks.”
Coming out of the second subcommittee hearing, it’s clear that AI is on track to become a game-changer for more efficient government operations. The final hearing, slated for April, promises to spark more discussion and hopefully help to lay a path forward for AI adoption across government.