OK, you probably know from following Watson on Evri at http://www.evri.com/product/watson-0x505342 or from http://www.latimes.com/news/science/la-na-jeopardy-20110217,0,586302.story?track=rss that Watson won handily. To quote from the LA Times article:
“During the Wednesday finale of the three-day “Jeopardy!” challenge that
pitted all-stars Ken Jennings and Brad Rutter against an IBM
supercomputer, the machine beat the men. Watson finished with $77,147, Jennings
with $24,000 and Rutter with $21,600.”
The natural language processing and the AI part of me rejoiced in this triumph of engineering, and IBM deserves a lot of credit for all this work. But I had a vague sense of disquiet, and I am not sure why. Perhaps it was because Watson was weirdly human today, vulnerable and fallible.
This time, Watson started well but seemed to grow hesitant, not quick enough to the buzzer. Even when Watson does not get to answer, its responses are shown, and several of these responses were off target. Watson seemed to do better in some categories and not others (just like humans?). All three contestants answered the final question correctly ! The good thing is that IBM donates the $1 million prize to 2 charities and Jennings and Rutter are giving away 50% of their winnings — so a good day for computer science and philanthropy!
As for me, I am going to re-read my copy of Joseph’s Weizenbaum’s book. Weizenbaum, author of ELIZA, a 1966 program emulating a Rogerian therapist — see http://en.wikipedia.org/wiki/ELIZA was so struck by people’s reactions to Eliza that he wrote “Computer Power and Human Reason: From Judgment to Calculation”, a very good read. Pip-pip for now.
#watson #jeopardy #ibm