Stories

AI is spooky good at predicting health care outcomes

Computers show the inner workings of a blood clot
Photo: BSIP/UIG via Getty Images

Artificial intelligence systems can predict patients’ diagnoses and outcomes with startling accuracy, and we don’t even know how they’re doing it. But we’re going to end up relying on them anyway, according to an article in the Harvard Business Review.

The details: A team of academic researchers recently fed Google’s machine-learning system anonymous data on hundreds of thousands of patients, and the AI was able to make accurate diagnoses and predictions.

  • The machines predicted a patient’s odds of dying with 90% accuracy, within 24 hours of their hospital admission.
  • “These predictions, however, were based on patterns in the data that the researchers could not fully explain,” the article says.

Think about the implications here. How would you feel if you walked into the hospital and your doctor told you, “The computer says you’re going to die. We don’t know how it knows that, but it’s almost always right.”

Key quote: “In some cases, doctors may have a legal obligation to use models that are more accurate than [human] expertise … This won’t take doctors out of the loop entirely, but it will create new opportunities and new dangers as the technology evolves and becomes more powerful.”