Quote:
Originally Posted by suzy_q2010
Following a computerized algorithm might find 99.9% of the horses. But zebras do sneak onto the bridge. How good will the computer be at finding them? Will the computer order an EKG if the patient tells it his indigestion is acting up?
|
This seems like an oversimplified view of what "algorithms" are already capable of. It is not just about following simple rules. For example, the paper in my previous post used Markov decision processes and dynamic decision networks. A properly-done AI would have so much larger knowledge base than an individual doctor, that it could handle the case you described just as well.
It is not a surprise that a doctor knew more about what it could be compared to your dad. But it doesn't follow that he would do better than an AI. Also, if less of the human doctors' and nurses' time was required to perform these tests and follow-ups, more of these tests could be done. Currently, if there is a small enough chance that the problem is something more serious, a doctor might feel that checking for it will reduce valuable time available for other patients.
There is nothing special about gut feelings - they are simply educated guesses based on the facts.
Quote:
Originally Posted by suzy_q2010
Also, how would a computer "practice" psychiatry, I wonder? 
|
While it's true that the brain is the most complex organ and people have the best understanding of other people's thinking, and therefore the highest levels of AI would be needed, the true AI doesn't seem to be required for a lot of work done by the psychiatrists, even the psychotherapy. Through the use of things like the EEG-fMRI, huge amounts of data are recorded, and the computers are best suited to making sense of it. I think targeted treatments (not necessarily with drugs) based on such tools are likely to be much more effective than psychotherapy. It could even be that it's just the social contact aspect of therapy sessions that helps the most (and MDs are not needed for that).