For the last couple of days, one of the top articles, with more then a hundred of others, in the health section of Google News was from The Times Online talking about the proficency of the search engine with diagnosing medical conditions. None of the articles, or even the blog posts picking up on this mentions what is in my opinion the most salient factor: the Google AI has achieved medical Expert System status. And even more, it did it in a bottom up fashion, independently from the specific area of medical diagnosis.
In the late 1980s I was working in a company carrying out AI development, and selling AI development environments. A classical example of the time was medical diagnosis, but with a diametrally opposite approach: top-down formalization of the rules used, and applicable to a specific field.
It is not clear from the articles if the tests conducted by the doctors were done in the newer vertical search engine Google Co-op’s health section, or by just plugging in the syptomps in the general one. YMMV: if you want to have Google diagnose your symptoms independently from doctors the conclusions should not be acted upon without consulting a professional at the end.
Just as with other aims of traditional AI, this will disappear from the radar screen of the reseachers, as something not worthy of their time anymore. It happened with chess, it is about to happen with speech recognition, and vision. The next step in more detailedly looking at what Google can or can’t do should come from a more thorough analysis of its results in the field of medical diagnosis, and from applying the same interrogation techniques to other fields of industrial, geological, geometrical, etc. expert systems.