Author: Dann Corbit
Date: 09:51:45 12/08/98
Go up one level in this thread
On December 08, 1998 at 03:58:24, Mark Young wrote: [snip] >>I would like to see Deep Blue verses Kasparov at one day per move. At that >>scale, Deep Blue would see so far ahead, it would start to see things >>strategically. 256 million positions per second times 86,400 seconds would be >>2.2e13 positions examined. Now that would be an awesome game. > >It would be a trashing of Deep Blue. I would bet the farm on Kasparov. Even at 1 >billion position a second Deep Blue will only see a very few plys past what it >could see at 3 min a move. Where a strong human player will find the correct >line of play, and not be hit with the tactics they could not work out at 3 min a >move. No chance for any computer at that time control IMO. And would show how >much more improvment there is to go in computer chess. You are assuming that they don't use singular extensions or related techniques to peek far ahead in certain circumstances... Besides which, IBM has the resources to form 6 piece engame tablebase files. In fact, Deep Blue was a modified database server machine. Even at that, if they just used the ordinary Alpha/Beta they would probably get at least 16 plies on tough positions, and a lot more on easy ones. So in all cases, the machine can see 8 moves ahead on offense and defense. When the board got sparse, it would see very deeply indeed. Of course, I also think Deep Blue would lose, but I'll bet it would be a great game. Wouldn't you love to see it? [snip]
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.