Author: Robert Hyatt
Date: 09:46:03 09/09/02
Go up one level in this thread
On September 09, 2002 at 12:25:03, Rolf Tueschen wrote: >On September 09, 2002 at 11:03:11, Robert Hyatt wrote: > >>On September 09, 2002 at 10:12:44, Rolf Tueschen wrote: >> >>>On September 09, 2002 at 08:59:25, Robert Hyatt wrote: >>> >>>>I don't agree. I _knew_ Tinsley. (...) >>>>He was convinced beyond a shadow of doubt that The final Chinook was better >>>>than he was, because of the big endgame tables they had constructed. (...) >>> >>>Are you sure that Tinsley meant "better" or better? What is the performance of >>>the machine if it can use the _perfect_ tables? What has it to do with playing >>>checkers? Didn't Tinsley assume with all the rights in the World, that he was >>>still the best player? >> >>Nope. In fact he resigned his official title so that he could play Chinook >>in a man vs machine match. And he was quite clear on the "better" issue. No >>doubt in his mind whatsoever, any more than he doubted that Cray Blitz could >>mount his head on the den wall if it wanted to (in chess). :) >> >>The most impressive thing about Tinsley, besides his incredible reign as WC, >>was his basic honesty and lack of arrogance... > >Good to know! > > >> >>> >>>Tables for endgames, at least in chess, had been calculated to the perfect end. >>>BTW what is the specific achievement of a programmer, having a finite room of >>>data, having access to a super computer, having a few months of computer time >>>free for each round? What is the sense to compare such a perfect automat with a >>>human genius? Since you were part of the branch as such, what gave you the >>>scientifical kick out of it? I mean could we compare it with the creation of a >>>logarithm table we all had back in school? Where is the creative element? And >>>finally the same question as last year - what is the kick to let a machine >>>participate with such help in human tournaments? The last question just to have >>>it complete the collection. No nitpick meant, honestly. >>> >>>Rolf Tueschen >> >>I don't think there is any intent to "compare" a program using endgame tables >>to a human, calculating on his own, and drawing any conclusions whatsoever, >>other than "which is the stronger player, period." Man/Axe were soundly beaten >>by the chainsaw. Man/horse were soundly beaten by the automobile. Yet no one >>seriously considers the automobile to be superior to the horse in any type of >>"equal" comparison because they can't be compared. > >First, your position is the minimum that should be taught for the sake of fair >play. But I would be content if the following could be included. Again, this is >not out of lack of respect for the class of programmers, it's IMO a trivial >analysis. A computer shouldn't be compared to automobiles. Automobiles are >surely 'faster' than horses but why is it better? Because it has more power. So >far so good. This is fair and then the end of such a race. > >Now let's see what is new with computers. (You know that I am not a chess >programmer, so please let's talk about the ideas as such.) > >I remember some older statements where experts tried to teach me that a computer >by definition "read" or used databases. I couldn't follow, because I saw a >difference in just "reading" the perfect results for a position and playing >chess with calculating the position with the actually possible depth. >For the latter I accept man vs machine competition, but only for this. The >moment perfect solutions come into the game, it is no longer playing it's simply >- -ok, I leave out the term. Now I wished to know if the differentiation is ok >and accepted in CC. Just copying perfect moves and playing on the base of >calculating... It depends on your definition of "copying and playing". IE humans clearly do this, although they tend to study the variations (sometimes) so that the side-analysis is understood if the opponent goes astray. But they clearly remember and play analysis that they did not create, and in all but IM/GM-level chess, they often play analysis they don't even understand. I don't see how that is so different from a computer. But to continue... In the case of endgame tables, that is simply perfect information. I use some of that myself, however, as I have memorized how to mate with king and queen vs king and rook, or king and rook vs king, or king and pawn vs king. I've never played out _each_ variation, I just understand the "plan" for the king and pawn side in a king and pawn ending, without having to do the analysis... Opening books are a different matter. I don't "just copy and play" at all in Crafty. It first analyzes the openings to see how often each move was played (while building the book) and discards those that look fishy. it then tries the openings in games and "learns" whether they were reasonable or had flaws that caused problems. This "feels" like what I do as a chess player. IE when I first started playing with the Evans gambit, I had to try it for a while to see whether I really liked it or not... and I based that on results OTB. Computers can remember better. Humans can generalize better. I think both have advantages and disadvantages. And once computers surpass all humans, so what? My car can outrun any sprinter already... yet the olympics still have sprints and long-distance races... > >Back to checkers and Tinsley. I thought that he still was the better player. He >had no perfect tables. And without these tables Chinook would have been much >weaker. So we can conclude that Tinsley was the better player - without the >tables being counted. I think that's trivial. Chinook could not have beaten him without the tables, back then. Today it would be a different issue, since we know how strong computers are, even without the endgame tables at their disposal... Hardware marches forward. Eventually technology will do what clever programming can not, given sufficient time. > >Let's go back to education in school and university. Leaving eidetic people >aside. I know for sure that you would never accept if students cheated with all >kind of hidden help during examinations. You would say that students should be >able to "think" for the correct answers. Looking at the help and then telling >what is written there as perfect answer, this isn't making any sense. No, but if I gave a human a position with white to move, white king at e1, black king at e8, white pawn at e2, and asked "what is happening here?" If he responded "white wins" I would not care whether he knew that if the white king can get in front of the white pawn 1 square with opposition, then white wins, or whether he played the position out and finally concluded black can't stop the pawn. Personally, I know the rule about the king and pawn, and I didn't create that knowledge myself, it came from a book and someone else, but I feel honest in using it in chess games... > >So, this is the ethic we all know and obey to. > >(NB the following is about competition and _not_ tools for training and other >helps for players.) > >How could you explain why it is so difficult to convince chess programmers and >probably checkers programmers too, that the usage of "perfect" databases in >tournaments is absolutely odd tradition and should be regarded as unethical? >What is the basic influence in computer sciences that leads people to believe >that such a technique could be allowed? I did never understand that. For me it's >in the same way clear as it's clear that medical patients are never insulted for >their illness - even if it was caused through prior heavy smoking or some such. >We would never raise any critic in the direct communication with the patient. Of >course we can make conclusions for prevention campaigns against smoking etc. > >So, let me repeat, everywhere where I made my proposals, people from CC reacted >as if I had said the worst I could say in CC. For most it was not even a >question at all, it was immediately viewed as muckraking. And I always thought, >that CC at times could be similar in practice to religious or political sects. > Again, what question are you trying to answer? If it is "can the best computer opponent beat the best human opponent?" then you set the experiment up to answer that. If it is "can the best computer opponent, using the same sort of pseudo-rules as a human has to use, beat the best human?" then it is a different issue. IE if you want to specifically exclude the computer from using an opening book, that's "an experiment". Not a fair one, because the computer still has other advantages, but now you just gave the human an unequal advantage as well because _he_ can use his "opening book" that he has memorized over many years. Again, equality seems impossible. If it is "can a computer beat a human on equal ground?" then the question is unanswerable because humans and computers have nothing whatsoever in common. Except that they can both play the game of chess... >Let me give a final addition in form of an analogy for the automobile example. >Usage of perfect tables is as if the engineers of the automobile's side would >organize the following for the race against horses. The race is held on a steep >street which is in addition carefully frozen!! > >What would you say, if such events were organized? Would you still believe that >it's clear who is better, because automobiles are faster than horses? Or would >you say, no, this way the competition has become a bit unfair? > >NB that I do not want that automobiles should use no fuel for the race, all I >want is that the horse has a fair chance to run without the automobile has a >surplus advantage. > >Tables are such surplus advantages in my eyes and not integrated part of >computer machines as such. For the moment they give the machine an unfair edge. > >Rolf Tueschen Again, trying to equalize the human and anything that is not biologically based seems impossible. For the automobile vs the horse, how much gas can the auto burn to equalize it with the horse? How much displacement in the engine can be used to equalize that to <what> in the horse? What about tires vs hooves? Etc. I simply don't see trying to make such a thing "equal" as being interesting, since it is essentially impossible to do...
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.