Author: Daniel Pineo
Date: 10:11:52 04/14/05
Go up one level in this thread
On April 13, 2005 at 21:52:48, Mike Byrne wrote: >On April 13, 2005 at 20:07:18, Walter Faxon wrote: > >>A brief article in "New Scientist" about 5 years ago suggested that future >>quantum computers will be able to attack chess by doing the quantum equivalent >>of a full-width search 300 ply or more deep. I googled today to find the >>following material: >> > > >I think the word quantum understates the jump required to do a full width 300 >plyu search. In the near future (next trm years), will continue to see a >doubling of processor speed every two to three years. I also see a patternm of >diminishing returns for the faster and speeds at these higher levels. Doubling >the speed at 1800 might have provided us with 100 elo points (or more), >Doubling the speed at 2300 ELO might have provided 50 elo points or so. Perhaps >doubling the speed at 2800 might provide an ELO of 25 point gain (against >humans). It is possible that computers 10 years from now will not be >convincingly better against the top humans than they are today. Why is that ? >Top human players bring a feel to the game that is very hard to quantify, >program and measure that is beyond calculating abilities of even top programs. > >Naturally if there is a shift in the processor speed doubling every 2 to 3 years >paradigm -- say from every 2 to 3 years to every 2 to 3 days - things may end >upa lot different. But without that shift, my prediction that in the next 40 >years - Chess ELO on a home PC <= Top 10 GM Elo under tournament time controls. >Not too much different from where are today imo. > >I am probably in the minority with this opinion - and feel free to tell me how >wrong I am! > >Waiting for Rebutal Regards, > >Michael Quantum computers are more powerful than that. Basically, they let you store more than one value in a register. In a classical computer, how many bitboards can you put into a 64-bit register at a time? Just one right? In a quantum computer you could stuff 1, 2, 10000, or all 2^64 possible bitboards into the register at the same time. When you perform a function on the register, the function is applied to every number stored in the register simultaneously. Of course, the big drawback is that you're only allowed to pull one bitboard out of the register, and the one you get is random. You can, however, do some fudging with the probabilities. You could take a register containing a board, run a GenMoves function on it, and you would now have the 30 or so successor boards in the register. Run GenMoves again and now you have the 90 or so depth 2 boards in it. Repeat that 100 times and you you'll have all the depth 100 boards in that one register. Then you can run an Evaluate function on the register. etc. Anyway, once they actually build build one big enough to hold a chess board, it should be possible to solve the game of chess.
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.