I recently read an article in the New York Times about a woman who hada poison-ivy-like rash — but it wasn’t poison ivy, but an uncommonreaction to undercooked Shiitake mushrooms. How would we want amedical diagnostics application to behave in a puzzling case likethis? Presented with a set of symptoms and a database of medicalknowledge, its first suggestion would probably be “poison ivy,” alongwith the human MDs. But unlike humans, computers aren’t blinded bythe obvious. Once poison ivy is rejected, the next step is to look atthe answers with lower confidence ratings. That’s where you wouldexpect to see alternative explanations. How far do you have to gobefore you get to the correct diagnosis? I don’t know, but we haven’tyet finished.
Great article about the future of machine learning.