Author: Vincent Diepeveen
Date: 11:15:15 02/17/99
Go up one level in this thread
On February 16, 1999 at 09:22:59, Robert Hyatt wrote: >On February 15, 1999 at 15:04:53, Komputer Korner wrote: > >>On February 12, 1999 at 19:01:05, Bert Seifriz wrote: >> >>>On February 12, 1999 at 15:26:09, Djordje Vidanovic wrote: >>> >>>>I read Larry Kaufman's review of AlexS on CCC, and I find it inaccurate to the >>>>same extent >>> >>>Yes, it is true, Larry did not catch what actually happened. >>>And we did neither at the beginning. >>>In fact, ALEXS learns nothing. It was CREATED with a learning program. >>>With an AI program the author says. > >This _really_ sounds like hogwash. IE 2-3 months ago I began hearing about >another _new_ approach to ai and learning. The program was called 'voyager' >and supposedly used 'fuzzy logic' and so forth. I had a chance to play it >several games on either FICS or chess.net, and it was very strong. In fact, >it is _crafty_. Front to back crafty. Maybe a few eval tweaks. maybe a minor >search change or two. But a pretty much 100% knock-off of crafty. Sure i directly saw this. Some things are translated to german in voyager, and that's it. It printed at deeper search depths exactly same lines like crafty gave, even the length of the variation was as long. Easy to catch this voyager as a cheater for me, even though he stripped crafty a little and renamed everything to german. bob later confirmed this as being a 100% crafty version, with only some minor changes. >When I hear that "an AI program designed a chess program" I _know_ that is >bullshit. Before an AI program can 'design' a strong chess program, it would >have to _be_ a strong chess program, because the complexity of playing chess >is less than the complexity of designing a program that designs chess programs. >IMHO, this is fishy. This program may well be an original program, I have no >idea. But to purport that it was created by an AI program is really funny. Alexs is currently in fact the dutch nightmare. Alex van Tiggelen joined dutch open some years ago with nightmare renamed to alexs (i am not sure why) with only PARAMETERS tuned by neural network. It played horribly. This year he joined with a nice windows interface and with HAND tuned parameters. So that is simply the (dutch) nightmare program i.m.h.o. and has nothing to do with learning. Alex van Tiggelen confirmed this at dutch open loud and clear. This participation was meant to show that it plays no better with hand tuned patterns... ...however it played way better, so so far again no good news from an automatic tuned parameter program. To quote Alex: "in fact this years participation in fact contradicts the viewpoint that automatic tuning the parameters using a neural network works" > >>>The point is the user would like to have the Artificial >>>Intelligence program himself. Mr van Tiggelen says this is much >>>too expensive and too complicated to use in the current >>>status. This maybe true. > > >or maybe it is 'smoke and mirrors'??? > > > >>>But as stated earlier, this would just be the interesting thing: >>>a stupid program which starts learning and then it draws its >>>own rules! Such a program seems to be decades away! > >check the web for the chess program named "morph". It does just what you >want, but it would probbly take longer to learn how to play decent chess than >it would to learn how to forcast weather. > > > > >> >>That is why I put it on my Krash list. It is no better and a lot worse than some >>freeware programs. I don't care and any other user doesn't either in how a >>program was created. The "proof is in the pudding so to speak". With little or >>no features and weaker in strength than Crafty, why would anybody buy the >>program? Since there is no way to make the program learn anything it is no >>different than any other artificial or commercial chess engine. Anybody can say >>they have an AI program. Indeed all the commercial programmers say that their >>program is very intelligent. So even if ALEXS is based on AI methods that others >>don't use, to the end user that buys the program there is no difference. The end >>user might as well buy CSTAL if he wants a program that plays wildly different >>chess and is a lot stronger than ALEXS. >> >>-- >>Komputer Korner > >I only hope this 'the program was created by an AI program' was just a poor >translation to English. Otherwise, _don't_ buy any ocean-front property from >the guy. > >:)
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.