Author: Dirk Frickenschmidt
Date: 13:54:41 03/16/98
Go up one level in this thread
Hi Ed, hi Bob, I think Ed pronounced the critical point where book learning is leading us: >If you can recognize the opponent everything is possible. >Scenario.... >Play 200-300 games against a SSDF opponent. >You then have a learned book especially tuned on that opponent. >Save the new book. >Repeat that for every expected SSDF opponent. Save the new books. >Release the program with these optimized opponents books. >Being in AUTO232 recognize the opponent and load the "prepared" book. >I am not in the mood to put energy in that. It's also a clear cheat. >However if you manage you can enter SSDF with 2900. This problem already has occured now in a slightly less drastic from in at least one well known program. What you get when you play this program are obviously pre-played autoplayer-games merged into the book. Togeteher with strong book learning this means sorting out the losses and playing only or nearly only the wins which have been merged into the book before. This leads the original idea of book learning competely ad absurdum. What happens then, Bob, has nothing to do with learning *while* playing. But with remembering autoplay wins which have been played long before. No real games happens any more. It's just replay. Seeing Bonanza once more: same film, same episode. This is something which amuses me as a user without limits. Why should I wait hours for a game which more or less is already in the book? Why not build in messages then like: "Dear user. This game was a win in 63 moves against program x Do you want to see the rest of the moves up to the late middlegame right now or rather wait until the two programs have repeated them themselves?" :-)))) I accept Bob's idea that all would be nice and even *very* interesting if you saw two learners play fresh from the start: clean books so far (containg not a single autoplayer game played before), then trying to specialize game after game one against the other, just like strong humans knowing whom they play and trying soemthing... But in our times of misuse of technical opportunities this scenarion is only a fairytale, I fear, and far from the reality we already see. So my question to Ed, Bob and all of you is again: a) is there a chance for a technical solution of any kind, making the misuse Ed and I described impossible or hard to perform(after defing misuse of course in a more fundamental way than I just did: there will be more than one opinion on this issue!)? b) is there a chance for a testing rule by SSDF which would make sense (again: to a big majority of users and programmers and not just from *one view*? c) is there a chance for any kind of agreement between programmers? Ed doubts this for good reasons. Kind regards from Dirk
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.