OR WAIT null SECS
There is a lot of narrow AI, but health systems need more general algorithms that are intelligent and can make inferences.
Daniel Durand, M.D., chief innovation officer and chair of radiology at LifeBridge Health, thinks about artificial intelligence (AI) in healthcare as deep learning, rather than intelligence.
At the 2019 World Health Care Congress meeting in Washington, D.C., Durand told Inside Digital Health™ that over the last 10 years, algorithms have been created to iteratively learn and become smart at diagnosing and extracting certain types of data, as long as they are trained to do that by a human.
Durand mentioned IBM Watson, which competed on Jeopardy. He said that many people think that Watson is a high-functioning system, when in reality, it is not.
If Watson was as intelligent as it appeared, it would be able to answer medical student-level questions in the clinic today.
The reason it didn’t turn out that way? Durand hasn’t seen any general AI — the kind with true intelligence and that can make inferences.
He has, however, seen narrow AI — which is AI that learns how to do certain tasks extremely well.
To be successful with AI, Durand said the industry must ask, “How do we stack up enough of these algorithms in the right way that they augment the team, that they help the MA, that they help the nurse, that they help the physician?”
Doing it this way can also help save jobs, which many professionals in the field fear, Durand said. The algorithms will expedite the process and make workflows more efficient.
“And so, AI holds the promise of being able to make our workforce much more efficient so that we can deliver more care, or better care, or more efficient care, to the same number of people,” he said.
Get the best insights in digital health directly to your inbox.