I really enjoyed Katie Hafner's article about ongoing efforts to create computer programs that can do medical diagnostic work, but in some ways I think the focus on the methods and results of the very best diagnosticians in the world is misleading.
Think about something like mass-produced clothing. You could go back to the early days of factory-made suits and ask whether they'd ever equal the quality of the craftsmanship of the best bespoke tailors. The answer is no. Nonetheless, mass production of clothing has been a huge success. Its cheap and it's good enough.
By the same token, the key economic opportunity for computer medical diagnostics is probably with the cases where people currently don't see a doctor at all. We've all been in a situation where we have some failure minor but annoying ailment, and we don't go to see a doctor because going to see a doctor is expensive and inconvenient and the odds are good that it's not a big deal and you'll feel better anyway. If it were extremely cheap and extremely convenient to get that kind of thing checked out via computer, a bunch of good things one happen. One is that all our minor ailments would probably be treated more optimally, allowing for moderately faster recoveries. A second is that some problems that seem minor but are in fact a bigger deal could be caught sooner. A third is that human health care professionals would have less of their calendars filled by people with very minor problems, and thus the system would have more capacity to deploy humans where humans are really needed. Last but by all means least, a fair amount of health care is about peace of mind—people like to feel on top of the situation, and the cheaper and simpler we make it for people to get that result the happier we'll be