Author: Robert Hyatt
Date: 19:29:34 12/31/03
Go up one level in this thread
On December 31, 2003 at 19:21:35, Ed Trice wrote: >Think about what a 2900 rating signifies: basically, you could spank a mere 2500 >rated player indefinitely, never incurring a loss yourself. Nope. Think carefully. 400 rating points mean you will win 15 of every 16 games. But you _will_ lose one or draw two if you are only 400 points above your opponent. > >As it would play against 2600 or 2700 caliber players, losses would occur. >Against 2800 rated opponents (pretty tough to find) a 2900 program would still >finish ahead at the end of a long match. > >Before a program could make an honest claim at being 2900, I think the team >would have to scrape up some money and get Gary Kasparov to play a 40 game match >with it. The terms would have to be steep to get the most out of Gary, such as >is he loses he could only play checkers from now on, and if he wins he gets to >have Pamela Anderson for a year plus about $20 million in spending money. > > >>All the top programs now are staggering compared to some years back, and >>Shredder 8, and DFritz 8, seem about to be first and second. Even if something >>else jumps up, like Ruffian or one of the known ones with a great new upgrade, >>how much can we really expect something to outshine the others any more? >> We already see that some of the previous greatest programmers, don't make the >>very top anymore, and there is a decent margin between Shredder, Fritz, and all >>the others. And it has never yet been seen that a program or human has passed >>2850. So it will probably not happen! Certainly not 2900! >> How could it happen? It would mean that simply no machine or human has a clue >>as to what to do about that particular program. >>Let's say that everything was done without all the bug problems, and the best of >>what is possible, all in one program. The most it might reach might be >>....according to ssdf....... 2900. Top humans will still understand things quite >>a bit better than any machine, but will only need to be very careful and wise. >>So machines will only be super, beyond human comprehension, in atleast 10 years >>from now, if other computer potentialities are unleashed, which have not yet >>been explored. Until then, they cannot surpass 2900, or human understanding. >>So there's hardly any room left for a new program to be REALLY unimaginable, as >>had been in the past. >>OR? >>S.Taylor
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.