Better than the guidelines doctors use, in fact. That's the conclusion of a new study written up in Science magazine. Here's how they tested it: A UK epidemiologist and his colleagues tested four machine-learning algorithms against the doctors' guidelines, to see how well they could predict which patients would have heart troubles based on 2005 medical records. Sure enough, the artificial intelligence programs did a better job of predicting who actually had heart attacks and other problems within the next 10 years.
What it means: If computers can teach themselves, they may be able to account for a lot of random factors with people's health, more than doctors can keep track of themselves. If these programs had actually been used, the study's authors said, 355 people's lives could have been saved.
Yes, but: Given how much trouble doctors have had with electronic medical records, are they really going to be enthusiastic about adding even more technology to their practices?
