For those interested, you can watch the games at: http://www.clusterflock.org/2011/02/the-watson-episodes-of-jeopardy.html
Date: Thu, 3 Mar 2011 07:47:27 -0500 From: [email protected] To: [email protected] Subject: Re: [Computer-go] Jeopardy It's the same thing in other games such as chess. Even a very weak program can play a move faster than a human. I test my chess program sometimes at the rate of several games a second in order to get tens of thousands of games in an hour or two. No human could beat a computer at that level (with clocks) because a human needs at least 1/2 second (my estimate) to mentally "process" a move and then physically make it on the board. In time scrambles you can move a little faster than that because you are anticipating the move and using the other players time and have your hand hovering over the piece before the opponent has made his move, but the principle is the same. At some point it comes down to physical reflexes and our "slow" cognitive hardware. Competitions involving humans and machines should probably be done on our terms, since we are the masters and they operate in our world. For example in GO we play at time controls that are comfortable for US, since it does not matter to them. We don't play game in one second and conclude that computer are far superior. I did not see this competition, but it seems odd to allow a machine to respond before a human can even digest the question. The question in written form can be slurped up and digested in some tiny fraction of a second by the computer, which is a big inherent advantage. So in my opinion it's almost meaningless to play Jeopardy against a computer with a setup like this. But Jeopardy is a reflex game as well as a knowledge of trivia game. I am rather curious how they designed the software to answer trivia questions and what the problems were. Don On Thu, Mar 3, 2011 at 2:34 AM, Stefan Kaitschick <[email protected]> wrote: Am 02.03.2011 19:19, schrieb Colin Kern: On Tue, Mar 1, 2011 at 2:29 PM, Michael Alford<[email protected]> wrote: On 2/10/11 4:31 PM, Michael Alford wrote: A reminder for anyone interested: The supercomputer Watson will appear on the Jeopardy program this coming Feb 14-16. Michael Watson won on the Jeopardy program, however, following up: http://www.foxnews.com/politics/2011/03/01/congressman-upsets-ibm-supercomputer-watson/ Intersting. I'd love to see a video of the match. While Watson's ability to answer questions is impressive, and I big step forward in AI, it's pretty clear that the main reason Watson was able to beat Jennings and Rutter was its ability to beat them on the buzzer. You can see in the footage that both human contestants are attempting to buzz in for a huge portion of the questions that Watson answered. I think that if Watson were pitted against them in a written test, it would be a much closer game. I'd even feel comfortable putting some money on the humans. Colin _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go It really came down to Watson beeing a world class buzzer beater. This speed test was ofcourse ridiculous in itself. But it was further skewed by the the fact that Watson got the full question in writing at the beginning of the question. At the very least the questions would have had to be supplied to Watson word for word at the same time the humans heard the words. Stefan _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
