As you’ve probably heard, IBM is currently repurposing its Jeopardy-winning super computer, Watson, to preform medical diagnosis. One of Watson’s great advantages over human doctors is the ability to read and digest vast volumes of medical literature that no person would ever have the time to get through.
On a related topic, I recently heard an interview with Ben Goldacre which contained the following disconcerting fact:
“Half of all clinical trials for the treatments we use today have never been published, and trials with positive results are twice as likely to be published.”
As if the implications of this statistic aren’t bad enough—namely that drugs are systematically made to look more effective than they actually are due to passive censorship of failed studies—the arrival of algorithms like Watson on the scene makes the need to release all trials even more imperative. Algorithms dine on data, and the more data they consume the better they will perform. When the algorithm in question is as critical as medical diagnosis, we should be feeding it every piece of information we can possibly find. Quite literally someone’s life could depend on it.