The Federal government is responsible for maintaining some very mature artificial intelligence (AI) models and many agencies are exploring AI pilot projects, with few projects in-between – but COVID has pushed agencies to advance their efforts, Federal officials noted in a discussion on AI.
Speaking at an ATARC webinar on November 5, Chakib Chraibi, chief data scientist at the National Technical Information Service (NTIS), noted the existing landscape of AI and machine learning (ML) in government is divided into large and small projects.
“If you look at the Federal spectrum in terms of ML adoption, there are a lot of small projects where we use RPA [robotics process automation] … and some predictive models. On the other end of spectrum, you have some very advanced analytic capabilities, like JAIC [the Joint Artificial Intelligence Center] and DHS [the Department of Homeland Security], working on some very complex problems like adversarial attacks and cybersecurity issues,” said Chraibi. “The problem is in the middle. A lot of agencies don’t know what to do next – they’ve started a little bit but they are still trying to figure out their way,” he added.
Chraibi suggested that agencies take advantage of their pilots to apply their models to other problems, noting that the work done in creating the first model can be reused and leveraged to solve the right kinds of questions.
While agencies may currently be on ends of the spectrum, the COVID-19 pandemic has pushed AI forward, with agencies turning to the technology to handle the massive scale of the response. Sanjay Gupta, CTO at the Small Business Administration (SBA), noted that the agency has scaled its workforce up by four times compared to February, and is using AI to create user personas and alert staff to unusual behavior.
“These kinds of tools are in place today, we are using them, and those are some of the things that have helped us overcome the surge in staffing,” said Gupta. “Since we moved to 100 percent telework status, we have been able to– knock on wood— maintain our security posture fairly consistently from when we were all on premises, because of the way we implement our cybersecurity posture. [The tool] doesn’t really care where you’re connected from.”
Agencies that are looking to implement AI solutions may be deterred by the prospect of sorting and cleaning large amounts of agency data, but AI can also be a tool in that task, noted Kathy McNeill, director of AI for the Centers of Excellence effort within the General Service Administration’s (GSA’s) Technology and Transformation Services (TTS) unit.
“From our perspective, we think that agencies should understand sources of data and start to leverage the tools we have today, because they’re much more advanced than 20 years ago, where we had to get all our data in order before we applied the tools,” said McNeill. “Use the tools we have today to sift through the data and understand the state of the data. We find that to be an effective way to help agencies move forward, and I think that helps accelerate the process of introducing machine learning and artificial intelligence.”