Would it not be fair (if admittedly more than a little dismissive) to characterise this proto-AI development by IBM as a "Smart Stethescope"?

I don't want to wear out my welcome here, but I do think that we are perceptably advancing along the development accelleration curve predicated by the Singularity theory. How long before the device described in the video Stephen linkled to is up-rated to gather additional test materials (saliva, urine and faeces being the traditional sources) from the patient and use them to test for an order of magnitude greater variety of potential diagnoses? I would not be surprised if that improvement process isn't itself already well advanced before the initial model enters widespread usage and patient familiarity.

For these and similar reasons, I think it important that discussions about potentialities of such technologies make an effort to distinguish between similar-seeming technology to the extent such speculations permit. As well, even the most extensive refinement of capability doesn't entirely rule out the possibility for mechanical (or other) error influencing the process (whether medical or some other) and, just as we do in the present, an independent (if not necessarily physically separate) mechanism to check results will still be necessary (just as doctors send diagnostic samples to separate labs for analysis today). And, finally, diagnosis and treatment will remain fully separate functions from physical examination and sample analysis functions, for precisely the same reasons they do today, to reduce the potential for catastrophic error (or smiley-tiling over the solar system :)).

Actually, this is a "smart stethoscope." :-)