The Department of Veterans Affairs’ four new artificial intelligence centers will place a premium on iteration in the AI testing process, National AI Institute Director Gil Alterovitz said Wednesday.
The different locations of the new NAII centers will play a role in that iterative process, Alterovitz said during a webinar produced by Fed Gov Today.
“When we talk about iteration and so forth, we’re trying to bake that into the whole process and system, by having these different NAII, AI centers in different places,” Alterovitz said. “We’re going to have four new AI centers like this and we’re looking forward to moving on that, on the iterative process on that as well.”
Having the AI centers built in different places is beneficial, Alterovitz added, so that the VA is able to test various AIs and continue to move forward in understanding and applying the technology.
These new AI centers, which were announced during a September AI summit, will be established in different Veterans Integrated Service Networks, or regional areas home to different VA medical centers.
Medical centers throughout each VISN were able to apply to support the development of the future AI centers through the end of September, Alterovitz said.
“We’re looking for medical centers to be the primary applicant, although, of course, they can include partnerships with either academic affiliates, VISN and so forth,” Alterovitz said during the September summit. “Our goal is to be able to enable operational use cases within that VISN by leveraging talents that we can have in terms of an NAII center.”
The four established VA institutes that are part of the NAII AI network are located in Washington, D.C., Long Beach, Calif., Kansas City, Mo., and Tampa Fla. Alterovitz said that each center pilots and scales new projects, broadens the talent pool and examines diverse patient populations.
Alterovitz also spoke Wednesday about AI’s future growth, pointing specifically to two angles of general AI growth: the technology’s ability to mimic a human and using patterns in data.
“The combination of those two, that’s really going to be powerful,” Alterovitz said. “That increased data set, that human-like capability and, potentially, embodiment through robotics.”