Author: Terry McCracken
Date: 00:59:49 02/08/03
Go up one level in this thread
On February 08, 2003 at 02:27:21, Dann Corbit wrote: >On February 08, 2003 at 02:13:47, Terry McCracken wrote: > >>On February 08, 2003 at 01:58:28, Dann Corbit wrote: >> >>>On February 08, 2003 at 01:52:42, Terry McCracken wrote: >>>>On February 08, 2003 at 01:38:26, Dann Corbit wrote: >>>[snip] >>>>>Hsu would create 65536 chips that are 1024 times faster each. It would be a >>>>>whipping. I don't even think it would be entertaining, unless you can't stand >>>>>Kasparov. >>>>> >>>>That's a bit of an exaggeration! >>>>IMO when Kasparov was at his best, 1992-93 no computer then or now or a Super >>>>Deep Blue could have stopped him! >>>> >>>>Machines are NOT better than the BEST!! >>> >>>I agree that Kasparov is the best human chess player of all time. >>> >>>But a machine that is thousands of times faster than Deep Blue would simply be >>>unbeatable. Let's suppose 2048 times the compute power (for example). That is >>>clearly achievable. >>> >>>That is 11 fold speed improvement. At 50 ELO per doubling, we would add 550 ELO >>>to Deepest Blue. Instead of 2700-2800 approximate performance, it would be >>>3300, and Kasparov would get 5% of the points in a match (on average). Which >>>means if they played 100 games, Kasparov would get 5 points, if he did not tire >>>out. I think probably he would tire and get zero, if he did not score early. >> >>I don't agree, and there are diminishing returns as you keep increasing speed, >>as the tree explodes! So no Super Deep Blue would _not_ be 3300. >> >>Maybe a quantum computer, with a 1000 Q-Bits?!;-) > >The technology to achieve 2048 times the performance is definitely here today. >They could also be less conservative with their algorithms, having an even >larger safety factor. > >>So after awhile you get 1 point or less for every doubling of speed untill it >>makes no difference. > >That is (of course) a real possibility -- the increase of compute power could >trail off in benefit. Unless it gets built, we will never know. > >I would like to see a battle for computer supremacy. IBM verses Hitachi or >something like that. That way, there would be some money to fuel the darn >thing. We might see something that could do trillions of nodes per second. Here I fully agree!:o) Terry
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.