Science fiction fans like me have long been fascinated with the idea of a machine that could actually think and behave like a human being, or one that could surpass even the best human being at some human task.  
That’s one reason I was particularly interested in a recent news story that told about Watson, a computing system from IBM that is designed to compete against humans on the well known game show “Jeopardy.”  According to an article in Computerworld, the show’s producers are expected to announce plans to air a show pitting a Jeopardy champion against the machine.  
Watson is designed to understand complex questions and to answer them quickly enough to compete on a rapid-fire show like "Jeopardy," the article said. Remembering a similar competition, a chess match between IBM’s Deep Blue and chess champion Garry Kasparov, in which Kasparov was defeated, one would have to say that our human "Jeopardy" players are in trouble.  
Not necessarily, however. Figuring out chess moves is one thing, but responding to natural language questions from a human speaker is another. IBM noted that to succeed, the software will have to be able to understand nuanced questions, slang and riddles, which would make it “a leap forward in artificial intelligence,” said Computerworld.
Think about a simple statement, such as “You are really cool.” If I say that to you, as well I might, do I mean that your body temperature is very low? Am I saying that you seem to lack emotional warmth? Am I saying that you are somehow “hip,” according to someone’s definition of hip? If so, whose definition are we using? The correct answer, of course, will depend on the context. It will also depend on things like my tone of voice (Am I saying it in a sarcastic way, meaning that I don’t think that at all?), my facial expressions, my body language, and perhaps how well you know me.  
This is where Watson will find the going not so elementary. If "Jeopardy' answers were simply statements of fact, no human could hope to compete against a high-speed computer programmed with the knowledge of hundreds of almanacs, encyclopedias and other references. But when interpreting spoken human language becomes involved, computers have been remarkably ineffective to date. Oh, sure, you can “train” a voice recognition software system to respond to commands that relate to its capabilities, but such systems often run off the rails when another speaker takes over, or when the trainer speaker uses a different term. I’ve actually seen this happen in a live demo of a voice recognition system, much to the embarrassment of the vendor—and these folks were AI professionals!  
So why am I bringing this up at all? It’s because one of the elementary fears spurned by the Computer Age is that the devices will eventually be able to perform all of the tasks now performed by humans—putting all of us out of a job. In this recessionary economy, this type of fear could actually spawn a backlash against automation technology that seems to do just that.  And automation is all the buzzword in today’s insurance technology space.  
The truth is, however, that AI is not ready to replace humans at this point in history, and that despite the best efforts of the brilliant minds at places like IBM, it remains a work in progress.  Will some jobs be made obsolete by automated systems? Certainly they will. But for the moment, at least, humans are required to build, program and maintain these systems.  
"Jeopardy" category: Landmark Dates in Computing. "Jeopardy" answer: When computers will be able to build, program and maintain themselves. "Jeopardy" question: What is no time soon? 
This article can also be found at