Atul Gawande says AI is 'massively far' from taking over medical diagnosis

IBM Watson
Atul Gawande says machines like Watson can't account for the variability in patient symptoms the way a doctor can.

Renowned surgeon Atul Gawande hasn’t quite bought into the hype of artificial intelligence in medicine, in part because diagnosing a human being is highly variable.

Speaking with George Mason University economist Tyler Cowen, Ph.D., Gawande balked at the idea that a patient could obtain an accurate diagnosis simply by relaying their symptoms to a program like IBM’s Watson. Healthcare is “massively far” from that type of scenario, he said on Cowen’s podcast Conversations with Tyler.

"I think it's one of the hardest things," he said.

One reason he doesn’t believe AI can take over medical diagnosis is because patient symptoms are often messy and imprecise, whereas machine learning requires clean data points. Patient symptoms are often "more of a narrative" than "a straight set of data," he said. 

Gawande emphasized that he still sees the value of AI in healthcare as a tool to augment physician decision-making, he just doesn’t envision machines taking over the primary function of doctors.

“I think the puzzle of it is that you need that capability to integrate information coming from the person interpreted and be able to get it into these kinds of systems,” he said.

RELATED: Plenty of buzz for AI in healthcare, but are any systems actually using it?

Providers have pushed back against the hype of AI recently. Several Stanford researchers called on the industry to reexamine its view of AI in medicine to focus on how it can be used as a decision support tool rather than as a replacement for clinicians. Other researchers have also pointed out the challenges associated with accessing clean data points. 

On the other hand, Gawande is high on wearables, which he says can be a “very powerful” tool in helping patients achieve their health goals. Although currently marketed wearables leave a lot to be desired from a clinical standpoint, Gawande said the healthcare industry is “entering a phase” in which doctors can combine existing patient information with data from wearables to refine medications.

At Brigham and Women’s Hospital, where he practices, Gawande said clinicians are using smartphone sensors to track patient recovery following surgery.