From: Huge@nowhere.much.invalid   
      
   On 2017-04-04, RS Wood  wrote:   
   > Huge  writes:   
   >   
   >> On 2017-04-04, JAB  wrote:   
   >>> Siddhartha Mukherjee wrote a recent piece in New Yorker called "What   
   >>> happens when diagnosis is automated?"   
   >>> Annals of Medicine  April 3, 2017 Issue   
   >>> A.I. Versus M.D.   
   >>> What happens when diagnosis is automated?   
   >   
   >> They did well. According to my dermatologist friend, the correct diagnosis   
   >> rate by visual inspection is about 50:50.   
   >   
   > Interesting how the toughest nut to crack with AI has been natural   
   > language.   
      
   Not really a surprise, though. Sure, recognising the speech is hard enough   
   (although speech to text converters seem to be quite good these days) but   
   understanding idiomatic language is very, very hard since it requires   
   huge knowledge of context. "I sailed down the river in a boat". Is the boat   
   in the river or the river in the boat? You and I "know" this because we   
   know what boats are and how they work, but a natural language system has   
   to be taught all of that.   
      
   This is quite a good read;   
      
   https://en.wikipedia.org/wiki/Cyc   
      
   --   
   Today is Setting Orange, the 22nd day of Discord in the YOLD 3183   
                     I don't have an attitude problem.   
       If you have a problem with my attitude, that's your problem.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)