Thursday, February 17, 2011
Supercomputer Watson crowned king in Jeopardy - The IBM Challenge
Lees mijn uitgebreide artikel over Watson & Jeopardy uit NRC Handelsblad: http://benniemols.blogspot.com/2011/01/wie-is-de-beste-quizspeler-computer-of.html
In a three day match, IBM's supercomputer Watson has beaten the two best human players in the history of America's most popular tv-quiz Jeopardy.
Watch here part 1 of the third and final match:
Watch here part 2 of the third and final match:
The end result:
IBM's Watson: 77.147 dollar
Ken Jennings: 24.000 dollar
Brad Rutter: 21.600 dollar
On top of this, Watson won 1 million dollar (donated by IBM to the charities World Visionair and worldCommunityGrid.org), Ken Jennings earned an extra 300.000 dollar (half of which donated to a charity) and Brad Rutter an extra 200.000 dollar (half of which donated to a charity).
Watson answered very fast, gave a lot of correct answers and some occasional wrong answers. In one case it made a definitely non-human mistake: Brad Rutter first gave a wrong answer. Next, Watson repeated the same wrong answer. No human player would ever have done that. But this is because Watson does not hear what the other players say.
All in all, Watson outperformed the best human players. Watson's electronic brain solves Jeopardy questions very differently from humans, but who cares. Watson used brute force, humans rely on their superior pattern recognition and highly developed linguistic abilities.
For a computer to win in Jeopardy, is much more difficult than to win in chess, as IBM's supercomputer Deep Blue did in 1997 against world champion Gary Kasparov. Natural language is highly ambiguous, which makes interpretation for a computer difficult. But now IBM has shown that computers are able to do just that.
The scientific research that has led to Watson is much more useful for the world than the research that has led in the 80's and 90's to better than human chess computer. Watson's victory shows that computers are ready to answer questions asked in complicated and often ambiguous natural language.
In the coming years this will lead to new applications of Question-Answering-software for businesses, health organisations, governmental organisations, non-governmental organisations and general consumers.