Today is the last day of the Jeopardy tournament between IBM’s Watson Supercomputer and Jeopardy nerds Ken Jennings and Brad Rutters. On a technical level this is a competition between 16 Terabyte RAM and 90 3.5 Ghz CPUs resulting in about 80 trillion computations per second and two human brains, each one being equipped with about 50-100 billion neurons.
At the moment it looks like Watson is going to win the competition. This illustrates, once more after Gary Kasparov vs. Deep Blue, (1) that computing power increases exponentially, a fact also known as Moore’s law, and (2) that as the more computers rely on learning algorithms, the more they resemble real human beings.
I collected some links on the topic:
- Engadget coverage including videos form the match and interviews with David Gondek, Watson’s engineer
- Jeopardy Archive: Game 1, Part 1 and Part 2
- Noam Chomsky talks about IBM’s Watson
- Ray Kurzweil’s predictions about the increase in computing power
- A paper by Hans Moravec, predicting that “[a]t the present rate, computers suitable for humanlike robots will appear in the 2020s” (Journal of Evolution and Technology, Vol. 1, 1998)
Update 17th February: And the winner is: IBM’s Watson!