Dr. Watson's Diagnosis

Now that IBM's Watson computer system has won Jeopardy, IBM is moving into computer-aided medical diagnosis. I'm surprised it has taken this long.

Medical diagnosis has been moving towards more and more "evidence-based medicine" and "Algorithmic-based medicine" which tend to somewhat mechanize the diagnostic process. Last month, a physician at Harvard recently told me that medical education is moving from "knowing" to "knowing where to find information". A few years ago, a professor at Harvard told me, "We don't encourage students to know what to do, we encourage them to know what to do next. The idea that a student should be able to leap to a diagnosis is misguided." A pediatric cardiologist who studied Electrical Engineering a year ahead of me at Caltech told me that studying medicine was more about memorization than the problem-solving techniques he'd learned as an engineer. He found that thinking through a problem was handicapping him against his fellow students who were doing a lot more rote learning.

My conclusion was that computers are superb at "memorization" and "algorithmic processing", and a time would come when computers could out-memorize, out-search and out-process many human physicians. Educating medical students to act like computers might be a losing strategy against computers that act like computers quite nicely already.

Now please count me as a skeptic on the general subject or artificial intelligence. So I don't make such arguments lightly. I first heard of computer aided diagnostics when my brother did a medical rotation at Stanford in the late '70s, so this has been a long road. Many customers have asked when our stethoscope will provide automated diagnosis and I've expressed great doubt about automated auscultation. A company tried to offer us such technology a few years ago, then went to a competitor and now seems headed for bankruptcy or a desperate buyout, after spending tens of millions of dollars on research.

I do, however, have faith in IBM's computer scientists. If a company with that depth of intellectual talent has decided to focus seriously on medical diagnostics, we could see some real progress.

This begs the question - how will medical education evolve when students can consult an App on their smartphones and get a REALLY RELIABLE diagnosis? More importantly - how will medicine evolve when patients can do the same?

Controlling the cost of medicine is going to come from a technological discontinuity. Tweaking the current system will be a failure, given the political and economic interests at stake. Watson may just be part of the solution.