By the middle of the next decade, soldiers will be working with technology so closely that they will seem to be a single system, says Javier Garcia, a neuroscientist at Army Research Laboratory (ARL) in Adelphi, Maryland. To achieve such seamless integration, artificial intelligence agents must understand and predict what humans are doing, then react accordingly. To enable that to happen, Garcia is shooting magnetic pulses into brains.
Those magnetic pulses target a specific part of the brain and mimic stroke damage there. By observing how the brain recovers, Garcia can learn rules that govern how it reroutes information around compromised areas. AI agents that understand those rerouting rules might be able to work around these problems as they arise. “If we could build a predictive model, the agent could be aware of the state of the human—and then maybe there could be some intervention,” Garcia says.
The route to recovery
More than a century of research, including neurological studies of people who have suffered brain damage, have mapped functions to specific areas of the brain. But scientists also know that injury to one area of the brain can affect functioning in another, an effect known as diaschisis. This is because each brain area is also part of larger brain networks.
Garcia wants to understand how the brain reorganizes those networks to compensate for damage. To find out, he and his team simulated the effects of a stroke by placing magnets on each subject’s skull, and sending a pulse into their head every second for 15 minutes. This repetitive transcranial magnetic stimulation (rTMS) temporarily quieted the neural activity of the targeted area and disrupted brain networks. Subjects then lay in a functional MRI (fMRI) scanner and kept track of images of spinning pinwheels, while Garcia and the team collected data that monitored their brain activity.
By comparing how well the subjects did on subsequent runs of this tracking task, the researchers could visualize how the brain quickly learned to compensate for the simulated damage. The networks returned to normal in about 15 minutes, the team reported last year in the journal Scientific Reports.
“This shows us the resilience of the brain,” Garcia says. “And maybe it’s something that can be harnessed to maintain or increase the performance of an individual.” If an agent could predict a fall in a person’s attention, he says, it could potentially use some sort of cue to refocus it—not necessarily magnetic stimulation, but perhaps an audio or visual cue.
Best of both worlds
Garcia has wanted to use neurostimulation to study brain function since his graduate work in cognitive science at University of California, Irvine. After earning his Ph.D., Garcia started working with a startup that had a contract with the Army Research Laboratory, the U.S. Army’s corporate research laboratory. After a year as a civilian contractor, he applied for a full-time job, and he has been at the ARL nearly six years.
He finds it a good mix of what he likes best about industry and academia. His research is less product-driven than it would be in the commercial world, but has a more specific goal than publishing academic papers. “It’s more future focused, and I’m still doing stuff that I love,” he says.
To explore careers in the U.S. Army Medical Department (AMEDD), click here.