The federal government is looking to leverage artificial intelligence (AI) to bridge the gap between the general public and valuable information.
During a five-hour hackathon held May 17, the General Services Administration (GSA) called on more than 100 hackers to come up with tangible solutions to make federal-level information more accessible through consumer Intelligent Personal Assistants (IPAs).
Efforts were aimed at the U.S. Federal Artificial Intelligence for Citizen Services community and were the latest offering from its Emerging Citizen Technology program, which is part of the fed’s Technology Transformation Service’s Innovation Portfolio. Earlier this year, it announced an open sourced pilot around the same idea.
Some attendees characterized last week’s gathering as more of a collaboration than a traditional hack, but were generally complimentary.
Participants included officials from Oracle, Microsoft, Amazon Web Services and Google; development types from San Francisco, Chicago and the nation’s capital; and representatives from among the 27 federal agencies taking part in the pilot, including the U.S. Parks Service and the Consumer Financial Protection Bureau (CFPB).
The goal of GSA’s Emerging Citizen Technology effort, the agency has said, is to understand the business cases and impact of technology ranging from autonomous vehicles to the Internet of Things, and to weigh opening federal programs to mobile and stationary AI and IPAs.
Megan Vorland, director of programs at D.C.-based graduate-level accelerator Dcode42, told Government Technology the hack was really “a lot more one-on-one meetings or one-on-many meetings on ‘What should the solution look like, what are your challenges?’”
In some cases, she said, it helped federal agencies narrow their focus from potentially offering weather data, for example, to one day releasing more pressing information on natural disasters like hurricanes. In other instances, she said officials learned the value of having good use cases; an up-to-date road map for where they hope to go; and, perhaps especially, good data to make available.
“I think even if the chatbot is not specifically created to x-y-z as laid out, it’s on the radar of all these public agencies, which is good news for them and good news for the American public," Vorland said. "Because it only gets cooler from here; the chatbot only gets cooler from here.”
In one collaboration, CFPB officials were able to work virtually with officials from San Francisco-based Coseer, a cognitive computing technology company with a focus on language and “trying to teach English to computers,” CEO Praful Krishna told Government Technology.
On its GitHub pilot webpage, CFPB indicates it is “interested in seeing the opportunity for people to get answers to financial questions and to provide just-in-time financial education to better help them navigate their financial choices.”
CFPB officials presented 50 to 100 pages from their website of frequently asked questions pertaining to automobile loans, Krishna said, and Coseer demonstrated how those could be imported to its platform and to its chatbot, which had been prepped beforehand.
“And we demonstrated it to them on like how a discovery of knowledge hidden in 50 pages can just become a matter of typing the right question and getting the answer then and there. The whole idea was that it’s actually that simple,” said Krishna, who pointed out the collaboration is expected to lead to a follow-up meeting.
Justin Herman, who heads the Emerging Citizen Technology program office, said on Thursday, May 18, in tweets that the event had been the “largest #AI for Citizen Services hackathon in the U.S.,” and advised readers to “forget ‘the future’ of public services: we now have prototypes for efficient, self-service programs in development now.”
Tad Anderson, venture capital and startups manager for Amazon Web Services Worldwide Public Sector, said via email the company was invited to participate as an industry collaborator and “provided subject matter expertise related to Federal AI/Virtual Assistant technology," and how it could be leveraged to build "more advanced ways of providing information to the public.”
Many agencies in attendance, he said, gave overviews of what types of public data they have, how those are accessed, and in what format they exist — focusing on frequently requested information ranging from student loan qualifications to parks information to applying for a passport.
“While I can’t comment on GSA’s ability to build and share solutions that incorporate public data in AI services, the event was a great platform to highlight potential solutions for agencies to leverage, as well as share best practices and common challenges to keep agencies exploring ways to use new technology to meet their missions,” Anderson said via email.
Even if any contract agreements were reached at the hackathon, Krishna estimated a working connection between federal-level data streams and IPAs or chatbots would take at least a year due to production constraints. But, he said, the overall mood was one of intense interest.
“We’re already seeing relationships come out. Now where do the relationships lead is something that’s yet to be seen," Krishna added. "Cautiously optimistic is probably a cliché, but that’s where people are: very intrigued, very interested and [they] have that sense of urgency that this needs to be done in the next couple of years.”
GSA officials could not be reached for comment by press time.