Author: Rolf Tueschen
Date: 17:09:33 07/31/02
Go up one level in this thread
On July 31, 2002 at 18:56:50, Dann Corbit wrote: >On July 31, 2002 at 17:38:54, Rolf Tueschen wrote: > >>On July 31, 2002 at 16:58:16, Dann Corbit wrote: >> >>>On July 31, 2002 at 16:22:44, Rolf Tueschen wrote: >>> >>>>On July 31, 2002 at 15:26:38, James Swafford wrote: >>>> >>>>>On July 31, 2002 at 13:52:41, Fernando Villegas wrote: >>>>> >>>>>>Dann: >>>>>>Am I wrong or there is a contradiction between these two sentences by you?: >>>>>> >>>>>>"But I am not ready to concede >>>>>>equal ability until it is mathematically demonstrated...." >>>>>> >>>>>>And then: >>>>>> >>>>>>"I am not arguing that computers are NOT GM ability either. Their true strength >>>>>>may be over 2600 on good hardware..". >>>>> >>>>>He's not contradicting himself. They may be, they not be... he just wants >>>>>"proof in the numbers". >>>> >>>>And how could there be proof - if computers do never participate in average >>>>competitions of tournament chess? The inflation of Elo-numbers has no valid >>>>basis. Without competition human chessplayers did not even _begin_ preparations >>>>to develop methods of anti-computerchess. Then it would be obvious that a >>>>general GM-level on tournament time control is a big myst. >>> >>>It is also possible that even with special preparation and anticomputer tactics >>>the GMs will still not prove superior. >> >>The experts deny for the moment. >> >> >>>Remember that we are not asking: "Is the >>>computer the equal of Kramnik or Anand?" We are asking (rather): "Is the >>>computer playing at GM level ability?" which is an absurdly easier bar to jump >>>over. >> >>No! Remember what I wrote about _real_ preparation for anti machine chess. >>Humans tend to optimalize in group transfer processes. So if Kramnik finds >>something successful, even you could exploit it - - _because_ you play human >>chess. Machines can't do this although operators and advisors _know_ exactly >>Kramnik's discovery. The difference! > >Do you really imagine that these chess players have never played against Hiarcs >before? I imagine that all of them have tried it. If they haven't then whose >fault is it that they were beaten? Anyone with money in their pocket can go and >get a copy. Now, that is not what is the problem. I do think that chessplayers own programs. But this is not influencing a "human attempt or united efforts" to work out anti computerchess strategies. There is a difference between general and theoretical guidelines and concrete patterns and lines against concrete opponents. > >>>>(BTW the actually good performance of Hiarcs in Argentina is a good proof for my >>>>statements. Or is anyone here present who wants to declare that the masters >>>>there had trained on the _specific_ machine before? Or that these masters had a >>>>specific incentive to learn alternative chess for machines the whole year over? >>>>Of course not.) >>>> >>>>So - without competition with _motivated_ and _trained_ GM our machines can't >>>>develop GM play. Comp vs comp chess can't constitute GM chess. That is why SSDF >>>>Elo numbers are based on circular logic without GM chess validation. >>> >>>I don't think this particular argument holds. Some GMs from China or India may >>>come to Europe to play against a group of European GMs they have never played >>>against. >> >>Objection. False category! Chinese players do know European players very well >>and vice versa. That's the "scandal" with completely unknown and newborn >>programs! This is simply not possible in human chess. > >Not really with computer chess either, since you can buy the programs if you >like. Even the updated versions do not change much in play. So, here we have again the difference between theoretical guidelines and concrete measures. If you accept that chess is always about concrete positions and patterns then it's obvious that new versions always require new research. But note well that this is already very specific for computerchess. Often the new version (especially designed for the concrete event) is without gamerecords worldwide. I think most of the fun of these show events comes through the hidden preparation by the programmers and operators and on the other side complete ignorance but GM status. This is kind of immature joy, BTW part of human chess too, just look at the long simul and exhibition tradition. When I played Dr. Euwe or Pfleger I had all the chances to show what I could achieve but by definition it wasn't good enough. Against Euwe I played the Wch opening of Euwe's own matches and simply was beaten with the difference of exactly one pawn in the endgame, when I thought that I had a drawn game and had proposed a draw. Against Pfleger I exchanged heavy pieces on the open line and then lost all domination in the later complications where I had expected good chances for me, which gives you a good picture for the differences between the vision for the future events or just pattern knowledge in GM format and all this in a more befuddled and limited variation in myself. Both events showed very clearly how far ahead the level of play was from my own. Nothing magical. But usually chessplayers react different. They claim that they lost through a single lost pawn only or by chance in very complicated positions. In real Euwe showed me how strong a GM must be to win technically with one pawn and Pfleger how experienced a GM must be to have the self confidence to always search for the natural moves, while I was hoping for a classical knock-out. Both, 1 pawn and complications couöld lead the amateur to believe that the differences were not that big as they are in reality. If you think that I digressed from computerchess then I can raise objection. Those who feel certain that they (in CC) are already above or at least even or almost as good as GM players, they can't know what GM players really can do. That the practical technique to _win_ a won position is often far more difficult against a computer is not a good argument for those who claim the existence of a balance. Because such technology, that was my point, is just a matter of hard work and inspiration. For the moment nothing a GM should waste too much time for. But here and elsewhere I read that this is the computer's strength. Yes, against non-prepared humans. But even good amateur players are able to drag programs into bad or lost positions. Now it is generally accepted logic to conclude that then machines can't be equal or stronger than GM players. You know, it's not the question of a gneral tenacity that would allow to call machines GM players but the many easy to get winning positions against machines do prove actually that machines are still way below GM play. The best analogy I know of is the difinition as the 'idiot savant'. On the one side a lack of experience like a child or even baby and on the other side the very limitied (in space) knowledge of a noble prize winner. By definition smart GM players make putty out of a chess machine. _But_ not winning with 9 pawns running for a new Queen! No, one pawn or a strong Bishop or the opposition will do the job too! Which doesn't constitute that the difference in strength is just one single pawn or a little chance. I wished that I could write better English (or German) to explain what I mean. > >>Here the factors are: >> 1. Human chess. 2. Internet distribution of games almost daily. >> >> >> >>>The groups may not have time to specially prepare adequately for the >>>opposing groups. But we can still get some data from the result. Preparation >>>may change things if they rematch, but we will still accept that result. >> >>Objection! Of course we would accept show events and their results. But we can't >>accept human tournaments with such alien participation. It's a show in a serious >>event. > >If we hid the players from each other and they saw only the board, then what >does it matter? It's opponent against opponent and all that matters is the >board. The computer has no special magic and neither does the GM. Of course you are right here. After you ripped the GM of all the advantages then you can conclude euality, but this is no longer human chess. It's the abstraction of all motivational and sportive aspects of chess. No! As I said, the moment human players start the nasty way, total war if you want, only then you'll see that machines are really far away from GM play, what the experts already see now. Bullying is the expression if I remember this right. But that doesn't mean that machines won't win here and there a game due to very human fallacies. If that is consolation? I'd always prefer advancements in chess. Rolf Tueschen > >>>Therefore, I think we must accept the Hiarcs result at face value. It is the >>>same situation. >> >>I might come to the acceptance but not on the basis of the "allegedly same >>situation", because it's not the same. The lack of identity is still the >>unsolved problem in computerchess. And my special theses is that the _actual_ >>face value is based on surprise factor and usually forbidden use of books and >>table bases. Viewed from human chess. > >It's not identical. But it's close enough. If something is forbidden or >allowed, then the rules should be followed. If rules were broken, then prizes >should be withdrawn. I suspect that you are only being overly dramatic.
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.