Author: Graham Laight
Date: 08:32:41 07/17/00
Go up one level in this thread
On July 17, 2000 at 09:44:57, stuart taylor wrote:
>To Dr. Hyatt, So how far do you beleive it is possible to go without tremendous
>speed? If software was maximised the most possible, could 1ghz. ever overtake
>D.B.? or maybe 2 ghz? What is the potential that still hasn't been realised?
>S.Taylor
If I may offer my thoughts on this issue:
Human players only look at a few nodes per second, so computers ought to be able
to play GM chess with the speed of a machine you can buy cheaply at the local
computer factory.
A while ago, I proposed a form of diagram to compare what was happening with
knowledge vs speed in the search space. Unfortunately, the use of these diagrams
hasn't caught on in CCC.
Basically, having knowledge allows you to see further into the future in certain
aspects of the game.
If you were searching (say) 10 deep, full width, with a low knowledge eval
function, your knowledge map would look like this.
ply |-------------------------------------------------------------|
| |
25 | |
| |
20 | |
| |
15 | |
| |
10 |#############################################################|
|#############################################################|
5 |#############################################################|
|#############################################################|
|-------------------------------------------------------------|
Breadth of knowledge
You'd probably miss any traps that were sprung more than 10 ply (5 moves) ahead.
On the other hand, if you were relying on knowledge only, there would be good
overall coverage, but gaps all over the place:
ply |-------------------------------------------------------------|
| # # # # # # |
25 | # # # # # # #|
| # # # # # # # |
20 | # # # # # # # # |
| # # # # # # # # # # # |
15 | # # # # # # # # # # # # # # # |
| # # # # # # # # # # # # # # # # # # # # |
10 |# # # # # # # # # # # # # # # # # # # # ##|
|# # # # # # # # # # # # # # # # # # # # # # # # # # # # # # #|
5 |## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #|
|#############################################################|
|-------------------------------------------------------------|
Breadth of knowledge
Here, you're OK if, and only if, the knowledge you used is relevant well into
the future of the game.
The typical chess computer's knowledge map typically looks like this:
ply |-------------------------------------------------------------|
| # |
25 | # ### |
| # #### # |
20 | ## ###### ### |
| #### ######## ##### |
15 | ###### ########### ####### |
| ########### ############### ########## |
10 |#############################################################|
|#############################################################|
5 |#############################################################|
|#############################################################|
|-------------------------------------------------------------|
Breadth of knowledge
Showing full searches to 10 ply, with "extensions" after that. However, as you
can see, there are massive gaps in the knowledge - almost nothing is known about
how the game will look in 25 plys.
This reflects the prevailing doctrine in computer chess that search depth is
more important than knowledge.
The problem is that chess programmers' have not yet succumbed to the inclination
to manage their knowledge methodically. In the past, I have suggested using
database APIs to store knowledge, so that the best knowledge for the current
situation can be managed, maintained, and retrieved just when it's needed.
Anyway, I hope this contributes to the discussion of whether a current PC can
play at Deep Blue's level...
-g
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.