Skip to Content
MIT Technology Review

AI can spot your eye disease with 94.5 percent accuracy

Even more impressive, it can explain its choices.

Explain yourself: The black box has long been a challenge in artificial intelligence. This refers to the tendency of algorithms to spit out results without explaining what went into them—and it can make weeding out bias difficult.

The news: In a paper released in Nature Medicine yesterday, DeepMind researchers described an AI system that can identify more than 50 diseases, refer them to a specialist, and, most important, indicate which portion of a medical scan prompted the diagnosis.

Why it matters: Explainability is crucial if AI is to see increased use in medicine. “Doctors and patients don’t want just a black box answer, they want to know why,” Ramesh Raskar, an associate professor at MIT, told Stat. “There is a standard of care, and if the AI technique doesn’t follow that standard of care, people are going to be uncomfortable with it.”