Author: stuart taylor
Date: 15:17:51 08/31/00
Go up one level in this thread
On August 31, 2000 at 16:09:53, Uri Blass wrote: >On August 31, 2000 at 15:59:44, Dann Corbit wrote: > >>On August 31, 2000 at 15:41:39, stuart taylor wrote: >> >>>If you had D.O. 200 plies brute force for every move, I think there can be no >>>question about it, that GM will always lose! If one time it was a draw, the GM >>>could be justly proud of himself, even if he was world champion. >>>I can't imagine how much mhz that would be, though. >>>S.Taylor (maybe 1 with 6-700 naughts). >> >>A trillion terahertz computer could come nowhere close to 200 plies. Probably >>closer to 20. Consider this little list: >>White(1): perft 2 >>total moves=400 time=0.00 >>White(1): perft 3 >>total moves=8902 time=0.01 >>White(1): perft 4 >>total moves=197281 time=0.29 >>White(1): perft 5 >>total moves=4865609 time=6.60 >>White(1): perft 6 >>total moves=119060324 time=164.04 >> >>Draw yourself a graph. Imagine what time looks like at 20. That search would >>play infallible chess, but most real searches don't work like that. They >>examine the square root of the node counts. So work out about what it will look >>like at 20, and take the square root. You will still find that the square root >>of a truly ridiculous number is still a ridiculous number. >> >>With massive pruning, it might get deeper, but then it would be open to errors >>like null move zugzwang situations, etc. >> >>Dann Corbit makes a prophecy: >>"Computers will *never* (and I do mean never, ever, ever no matter how many >>years forward -- millions of years, billions of years, trillions of years) fully >>examine 200 plies forward at tournament time controls of 40/2." > >If we assume that they get twice faster every 18 monthes they need less than 900 >years to examine 200 plies forward. > >In the case that they get twice faster every 18 monthes they can see 2 plies >deeper in 9 years(assuming a branching factor of 8). > >They can see 200 plies deeper than today in only 900 years(assuming a branching >factor of 8). > >I do not believe in this assumption but the question is when do you expect the >improvement in hardware to stop. > >Uri I think the speed race will end in 5-10 years. 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.