On Feb 14, 2011 IBM’s “Watson” computer will take on existing jeopardy champions.  Watson will win.  Why?  ‘It’s elementary dear Watson.’  Watson will win because there’s no way that IBM would release their creation on national TV risking a loss.  They know it will win. 

That little bit of human intuition, that knowledge of human knowledge however, is far far beyond Watson.  Watson will parse the knotty Jeopardy clues which will represent a significant move forward in natural language processing.  However, Watson won’t have the first clue as to what any of those facts mean. 

Watson will be able to tell you that a certain movie featuring Keanu Reeves in 1999 is “The Matrix”.  Watson won’t be able to tell you that the movie is a commentary on the philosophy of solipsism, unless of course that fact is written about in Watsons huge database of articles and trivia.

What is truly revolutionary about Watson is that programmers gave up on ‘him’.  They couldn’t program Watson to parse the language of Jeopardy questions.  They couldn’t come up with the ‘recipe’ that was sufficient to capture all the cases.  Instead, they allowed Watson, with careful guidance to teach ‘himself’. 

This paradigmatic shift in programming mimics the very evolutionary mechanisms which brought forth our human minds.  If the technique can be scaled and improved upon, it won’t be long before Watson, or machines like ‘him’, will soon be able to tackle larger questions which we humans still find baffling. 

If you’d like to see the power of machine learning… witness the power of Google Translate available now as a free app.  The translation algorithms were not programmed per se, but the program was allowed to evolve itself to a high level of functioning.