Featured
As a doctor with the World Health Organization in the 1970s, Larry Brilliant saw firsthand smallpox’s destructive force.
Working out of the organization’s New Delhi office, the young doctor saw case after case of individuals contracting and dying from the disease. The most distinguishing symptom of smallpox are the fluid-filled blisters that can cover a person’s entire body.
But the epidemiologist didn’t just stand still as the disease took more and more lives — he was one of the key players in a worldwide vaccination initiative.
Leading a team of 150,000 health care workers in India, Brilliant and his colleagues spent 20 months searching every house in India to vaccinate every community and create “immunity rings” when specific smallpox cases were found.
Those efforts paid off. Smallpox was officially declared eradicated in 1980 by the World Health Assembly.
“I want you to be inspired about what a group of fanatic, crazy people can do when they are battling against something as horrible as that,” Brilliant said Thursday morning in a fireside chat with Usama Fayyad, executive director of Northeastern University’s Institute for Experiential AI.
Today, Brilliant is taking on a different kind of challenge as the CEO of Evity.AI, a new AI health care technology company. The company is in the process of developing a health care platform that takes advantage of artificial intelligence to assist in patient care, he says.
Brilliant is working directly with the Institute for Experiential AI on the project and was invited to speak Thursday morning to kick off the institute’s State of AI in Precision Health Conference on East Village at Northeastern’s Boston campus.
“[The science fiction writer] Arthur Clarke said any sufficiently advanced technology is indistinguishable from magic. We have two great bits of magic that are going on,” he said. “AI is indistinguishable from magic. … Medical science has advanced so much it is indistinguishable from magic. I thought, ‘Can we put those two things together, narrow the band and increase years of healthy life for everybody in the world?’”
Brilliant has experience working in technology. For one, he worked directly under Google co-founder Sergey Brin and helped lead the company’s now-defunct Google Flu Trends project, which aggregated and synthesized large amounts of Google search data to help estimate flu activity around the world.
That role helped Brilliant see the positive impact large data could have in health care.
“We beat CDC reports by over two weeks with 100% accuracy for two years,” he says. “The example shows that large data, big data, properly done, can have enormous public health consequences.”
But when it comes to using AI in health care, does it actually work or does it tend “to fail miserably,” asked Fayyad.
In a lot of cases, the AI does a little of both, Brilliant says, noting that ChatGPT often gives wrong answers to medical questions and cites medical papers and journals that do not exist.
“I’m trained to make AI hallucinate because in medicine, we can’t have it,” he said. “We are the most risk-averse profession. We are not allowed to kill one person in order to save 100. What works for ChatGPT and generative AI in other parts of the world will not work in medicine as they are today.”
“So, what is the right approach?” Fayyad asked.
Having experts in the loop, Brilliant said, a play on the term “human in the loop,” which means to have humans working hand-in-hand with machines to correct their mistakes and improve efficiencies. But in the medical field, that means hiring medical experts who are qualified to correct those errors.
“It means paying for experts in neurology, cardiology, orthopedic surgery, burden of proof calculations, stochastic process, geonomics,” he said.
These experts will not only be tasked with correcting mistakes, but also feeding and curating the models with high-quality data for the entirety of their lifecycle.
But Evity.AI’s work goes outside of just training and upkeeping the model. They also plan to develop decision support systems that clinicians can turn to help them make better decisions about a patient’s care, he said.
“Building systems like that is the future,” he said. “We hope to be a part of it.”