Author: Komputer Korner
Date: 10:05:09 08/15/98
Go up one level in this thread
On August 11, 1998 at 18:25:01, Robert Henry Durrett wrote: >So far, we talk about a chess engine being used to "solve" a given chess >position, where there is no human intervention in the "solving" process. We not >talk at all about the human being the primary "solver" and the software >"assisting" the human. > >An example of the latter would be where the human gets an idea for a PLAN [and >then tries to see if the objective(s) of that plan can be achieved. > >Another example would be where the human dreams up a "fantasy position" and then >tries to find ways to cause that position [or key aspects of that position] to >be brought about in the face of strong opposition. > >Why not program computers to do much of what the "human" is doing in the above >example? In other words, create software which will try out "typical" plans or >automatically produce ideal "fantasy positions" and then search for ways to >approach them" Or, if this is too ambitious, then just let the human enter the >"fantasy position" and have the software search for a way to get there from the >given start position? Similarly, let the human select from a displayed list of >"typical plans" and then let the software offer some kind of assistance relative >to the chosen plan? > >Is this too "far out"? > >The bottom line is: "Is there any software out there designed specifically to >assist humans in the human's evaluation of a position, or can such software be >created?" Would it be practical to try to do so? Certain programs do create plans and 2 of the most successful computer programs, Fritz 5 and Nimzo 98 which are preprocessors do this very successfully. They look at the position at ply=0 or some low ply number and then adjust their algorithms to fit their plan in the position. It seems that strict preprocessing (ply =0) doesn't work out for the best because deeper search will find refutations of the initial plans. However a combination of low ply level preprocessing (do some minimum search before adjusting your plans) works well judging by the success of Fritz 5 and Nimzo 98 on the SSDF list. Granted the programs do not let the user know what is going on but the programmer could create log files to explain what the program is thinking about. Whether you can convince them to do that is another matter. -- Komputer Korner
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.