The machine will pick up on rare and uncommon things if the data is there
Not trying to antagonize you as I agree with you regarding the processing of data, but lets consider a few situations:
Patients who are malingering or hypochondriacs. They could be well versed with certain symptoms and provide the computer data that could be consistent with the diagnosis of their choosing.
Patients with psychiatric problems that can for the most part only be diagnosed by gathering subjective data.
True real life example to demonstrate: the homeless male who presented to the ER during my 3rd year of medical school shortly after the Christmas holidays who was threatening to kill himself.
Issues: the homeless are known to have a higher incidence of mental health disorders, the holidays are notorious for higher rates of suicide.
I walked in and preceded to ask him what his plan was to kill himself. After a few pensive seconds he looked up at me and states that he want to use the oxygen tubing (while pointing across the room to it) to hang himself.
The ER resident I discussed the case with quickly informs me that this was pretty common behavior amongst homeless patients in the winter to punch their ticket for a free meal and night or 2 stay in a warm bed in the hospital.
I am not sure how a computer could reconcile these scenarios better than a human clinically.
The extremely rare entities that you have mentioned (that the computing power that Watson could provide a clue into) are just that, rare.
A more important frontier in my mind is disease prevention by better understanding / manipulation of the human genome. This is a cleaner point to intervene with technology than trying to find and solve problems on the back end in humans.
If computing power does evolve to the level you describe, they just might come to the conclusion that we are not worth the effort of helping (ironically in much the same way intellectuals have no real use for God).
This post was edited on 3/1 at 10:32 pm