The Eye of Watson Turns Red? (post 6)

Congratulations to Watson!
His victory is a major milestone for achievement in computing.

One of the first computer programs I wrote was a program to solve sudoku puzzles.  I wrote it in scheme while finishing my math degree at IU.  It was shortly after that I decided to stop doing sudoku and return to crossword puzzles.  I figured that if I could write a program to solve them, then sudoku couldn't be that cognitively challenging.  I could not begin to imagine how to write a program to solve crosswords.  I told myself that if anyone ever did, I would be deeply impressed.  That was in 2004 - turns out such a program had been written in 1999 - but it used the internet, so that's cheating (typing in a crossword clue into google can almost immediately get you the answer).  I still don't know if a crossword solver has been written that does not use the internet, but I think Watson's achievement would trump that (and no, Watson did not use the internet).

To my mind, the most impressive aspect of what Watson did was the interpreting of the clues to form relevant search strategies.  That demonstrates an advanced degree of language understanding.

Computer scientists long ago realized that the most difficult cognitive task we perform is holding conversations with other intelligence beings.  It was Alan Turing (a personal hero of mine) who declared the ultimate challenge for a programmer is to design a machine capable of conversing with a human without the human realizing that she is talking to a machine.  This is known as the Turing Test - and it has never been passed in a meaningful way.  Even Watson, who can so elegantly say, "What is shoe?" could not talk to you for any significant period of time without you realizing that something was weird ("Why do you keep asking questions?  Stop it!").

It is inevitable that the Turing Test will be passed; it is only a matter of when.  We have been notoriously bad at predicting the speed (and order) in which computing milestones would be reached.  Purportedly, Marvin Minsky, the former head of the AI program at MIT, told DARPA in the 60's that he could produce fully autonomous robotic soldiers within ten years (see McCorduck, Pamela (2004), Machines Who Think)  They're still waiting.

For now, it seems safe to say that it isn't going to happen soon, and it is and will continue to be the internet - and its ability to help us to converse with one another - that remains the most compelling and inspiring technological achievement of our era.

Oh, and no, the eye of Watson did not turn red.  We are safe.
But watch this...