Author: Dave Gomboc
Date: 14:26:18 05/11/99
Go up one level in this thread
On May 11, 1999 at 16:21:01, Dann Corbit wrote: >On May 11, 1999 at 14:32:02, Dave Gomboc wrote: > >>On May 07, 1999 at 21:01:06, Dann Corbit wrote: >> >>>On May 07, 1999 at 20:48:48, Thorsten Czub wrote: >>> >>>>it is a different chess program. working different. >>>>and the ssdf guys are not allowed to publish test results. >>>>it is forbidden forbidden forbidden for them. >>>>:-)) >>>I wish the "non-bean counters" would just take their pasting in the mouth like a >>>man. Eventually, I think that approach will win. But a public spanking every >>>now and then might goad them to work harder at it. >> >>I'm not convinced. They knew how to attack in the 1800s, but there's a lot more >>to chess than attacking. If you don't understand the other things, you will get >>'niqued every time. >I don't mean that all-out attack is the secret to chess. Rather, that knowlege >based programs will produce a revoltion (if any do). Fast searchers face the >exponential wall. Even an exponential increase in hardware strength will >produce a linear increase in playing ability and (at some point) will nearly >stop being useful altogether. > >Hardware used to double in strength every 18 months, but that has accelerated to >once per year. Now, if this trend could continue indefinitely (it can't, but if >it could...) we will have a computer that is 1024 times more powerful in 10 >years, and one that is 65536 times as powerful in 16 years. Yet, since each >move has on average about 28 possibilies, the gains in plies will be >considerably less than the gains in power. Hence, to get a radical increase in >playing strength, hardware will not be the best answer, eventually. Instead, >chess knowlege will need to be incorporated into the programs. Chess knowledge is already being incorporated into the programs. It is just as important to not waste time incorporating chess knowledge that the search would have figured out on its own with less cost. Every talks about a branching factor of 30, or 35, or 28 (your number). But really, it's about 2-4 in today's programs. Still exponential, but it's clear that this makes "eventually" even more eventual. :-) Dave
This page took 0.01 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.