VA’s Artificial Intelligence Director Details AI Institute’s Early Efforts

The Veterans Affairs Department’s nascent National Artificial Intelligence Institute is focusing on hammering out policies and streamlining its processes so it can share data with partners in a speedy way. 

After its initial launch late last year, the agency’s inaugural artificial intelligence director and lead of the center Dr. Gil Alterovitz shared few details with Nextgov about its ultimate aims, but at an event in Washington Wednesday, he expanded on the center’s early intentions and efforts. 

“We are at that time in history where human intelligence at some point will intersect with artificial intelligence,” Alterovitz said at the ACT-IAC’s second intelligent automation and AI forum. “And so it's a really special time for us to learn about it.” 

The AI director said the intersection between human intelligence and specific areas where computational artificial intelligence will meet is inching closer to reality by the day. After that, AI “then will soon exponentially continue to grow beyond our capabilities,” he said, and humans will “have a short time to realize how quickly things change.” And through the NAII, Alterovitz now supports a range of initiatives to help VA and others adjust and prepare.

In its early days, the institute’s insiders are looking at various ways to synergize the agency’s versatile, already-existing efforts deploying the tech and working to create policies that “enable AI to be used.” Alterovitz was a member of the writing team that produced the 2019 Update to the National AI Research and Development Strategic Plan, and also supported the creation of ACT-IAC’s newly released AI and IA playbook for the federal government. 

The institute also aims to identify new veteran-focused use cases for AI, and Alterovitz added that insiders are already collaborating with other agencies and engaging with stakeholders and others across industry, academia and nonprofits to gain feedback on budding applications to explore. 

Focused on “defining AI” and honing in on how it can best benefit the VA’s mission, NAII also identified what Alterovitz looks to as five “kind of key capabilities” that the agency aims to build capacities around through strategic research and development. According to Alterovitz, those areas include: deep learning, trustworthy AI, privacy preserving AI, explainable AI and multiscale AI analysis.

He also elaborated on a variety of NAII endeavors that aim to revamp public-private partnerships and ultimately accelerate the development and adoption of AI applications. Alterovitz explained that in one of its most recent AI-driven innovations, the agency collaborated with technology company DeepMind Health, and created an AI system that can forecast a life-threatening kidney disease before it appears. 

“And you know how long it took to make the agreement to share that data? It was like two years,” Alterovitz explained. “And so we're thinking, can we speed that up?”

He has a history of leading speedy projects known as tech sprints, which allow outsiders to test out data in the VA (or other agencies’) formats to develop tools and programs that can lead to new data-driven insights—without waiting long periods to establish partnership agreements. Through the NAII, he said the agency is now launching “parallel pathways for potential partnerships,” which will essentially allow for stepwise agreements where small datasets can be quickly accessed and then potentially used for future tech sprints, pilots and national competitions. 

“And so, you know, we want to speed up some of these CRADAS and [other things] that allow us to share data, so that we can take advantage of some of these AI capabilities,” he said. “We want to build AI capabilities in-house so we can communicate, and be able to learn about some of the data once we have experts from companies and partnerships.”