Author: Rolf Tueschen
Date: 14:08:04 05/20/02
Go up one level in this thread
On May 20, 2002 at 13:04:06, Albert Silver wrote: >On May 20, 2002 at 12:13:51, Rolf Tueschen wrote: > >>On May 20, 2002 at 11:17:47, Timothy J. Frohlick wrote: >> >>>Dear Mr. Tueschen, >>> >>>I can't give you percentages for engines without EGTBs or Opening Books. I do >>>think that the knowledge of endgames and openings is crucial in human chess >>>progression. A chess master has openings stored in his/her memory. What is >>>wrong with the chess program having the same information? >>> >>>The specialized knowledge of today's chess programs is what makes them more fun >>>to play. It is no fun playing a machine that computes an opening line for three >>>minutes that is on record as being a lousy line. I prefer all the added >>>information. >>> >>>Computers have no intelligence on their own. They simulate intelligence. You >>>know that Rolf. >>> >>>Tim Frohlick >> >>To prevent that we begin talking at cross purposes, let me please add, Tim, that >>I like playing the programs too with all that integrated. The question about >>strength of the engine is often confused with wrong comparisons. Here are some >>of them I met in earlier discussions. >> >>- Humans learn theory by heart so why books are wrong in computer programs? >>- The design of a computer program was always a combination of engine and book. >>- Also human players learn by heart without necessarily understanding each move. >> >>All these arguments are false. But it's not so at first sight. And therefore we >>discuss all the time. > >Allow me to refresh your memory. Yes, please. > >>Then the point learning by heart without understanding. Well, that's an easy >>one. This is how weaker amateurs must play chess. Still it makes fun, as I know. >>Masters would not be masters if they played chess like this. Masters and their >>big brothers write the theory weaker amateurs then must learn by heart. >> >>Of course I know the simulating thing, Tim, but I cannot understand why "we", >>computerchess people, programmers and their programs should try to simulate >>being GM without respecting the normal FIDE rules of chess! Why human >>chessplayers can't read out of books during a game of chess too? Because, I got >>the answer, opening books are not books, they are integral constituent of a >>machine. Ahar... > >Yes, Phil Innes stuck by this utterly ridiculous argument too. You both argued >that every article in the FIDE rulebook must be followed to the letter by chess >programs. Does it matter to you if I did never write such things as to the LETTER? Respecting the normal rules of chess. I am not a lawyer. >You singled out the one on consulting previously stored information >and it was shown you had two problems with this: It's new for me that I should have had two problems... > >1) If you follow the rules, then you must truly follow all the rules at which >point you will find there are MANY that machines do not follow. Why? Because the >rules were written for humans. There is a section for specific computer-related >rules. It's a real pity to read such a point here. Point 1. So, above you mock that I pointed to each and every letter of the rules, which I didn't do, and now you begin to get me with seemingly contradictions with myself? You must find a decision for which side you want to lecture. Usually it would be even worse if the machine wouldn't follow more than one rule. In your style it sounds as if that could become a problem for my argumentation, which is clearly false logic from your side. As I already pointed out to Dr. Hyatt, FIDE hasn't defined what a computer should be. But also surely the actual computers do not follow the rules of FIDE as I have quoted them here today. > >2) The book is not the only area of previously stored information. Other than >the obvious EGTBs you have the entire hashtables which are nothing more than >information written into memory by the program to be organized and consulted. So >if books are illegal as you contend then so is the engine itself. What is the purpose of your argument? That if we would follow my arguments we would end in ridiculous pain? Isn't it more so, that you found more critical points for the programmers to look upon if they want to participate in human tournaments? Let me repeat my main point, perhaps it could bring some harmony back. I do not attack computerchess just out of joy, but I wrote all articles in the perception that it would be good if computerchess could participate in human tournaments without playing unfair on the human players. What chess programs should contain for the interested player or client, that is a totally differend story. I myself would not be interested in buying a program without a book. Couldn't we find at least a consense about our mutual good intentions? > >You also argued that the books served no purpose other than to save the programs >from falling into known bad positions. It was then shown that humans are no >different. You contended that humans *understand* the moves. Ok, if the programs >must *understand* the moves then it is hopeless. Is this all coming back to you? Perhaps not? At least not in my logic, Albert. You are in difficulties with your logic above. I say that human players understand chess (at least as GM or master). Now, your answer ist, well, then all is lost, because the programs simply would never "understand". But this was not my argument. The notion 'understand' is uninteresting. The point is the 'oneself' or "on its own". Will you deny that a GM would do it on his own? And do you agree that we have a lot of position no machine will ever look through or foresee in the next couple of years? And you would still not think that this could be judged as a real disadvantage in relation to 'strength'? >You countered the engines must at least be able to reproduce the moves they >play, showing they could find the moves on their own. That argument too is >deeply flawed. It was more of a help for purposes of the debate. It was not a direct proposition for the programmers. But let's see why it should be flawed. I know it's not. > Examples abound, but I'll repeat the one I used: The Marshall >Gambit. Prior to Marshall's introduction of the gambit, no one had found the >idea as he played it. Thus any player that suddenly began playing it cannot >claim they would have found it on their own. They certainly hadn't until then. >The Gambit was taken apart by Capablanca in a historic game and the gambit fell >into disgrace for at least 30 years (!). No one played it until some Russian >analysts found the sequence of ...c6 instead of Nf6 as played by Marshall >himself. Soon enough it gained credibility and perhaps the moves were >*understood*, but that certainly doesn't mean that all the players who play it >today, even top players, would suddenly be able to find both 9...d5! and the >necessary ...c6 on their own had they no prior knowledge of this opening. You >certainly know the story of Columbus's egg. There is nothing wrong with this, >since why should every mathematician have to find and rediscover on their own >over 2000 years of theory, as opposed to learning what has already been >discovered and building from there. That is how we advance. Now, this is a flawed example, yes! The Marshall is a typical tactical line. It's called a Gambit! Now, in such lines you will always have historical and changing sway. Until - the final solution will be found. But the Marshall is of so much interest because we couldn't find clarity. Albert, the late master Diemer gave his whole life to find and defend his Gambit. May he rest now in peace. But here other than in the Marshall the experts see permanent disadvantages due to the material loss and no ersatzgain for White. But who knows?? In 20 years there will be a rebirth of the Diemergambit and its relatives. Your argument has more flaws. You write: Thats how WE advance. Fine! But a computer program has not won the status yet to become like US. Your report about the Marshall has a top flaw. THe idea that the 30 years had anything to do with - comparable to science - a necessary advancing of chess knowledge. This however is false! It had nothing to do with such aspects. What it took was simply a player or analyst who was willing to put enough heart blood into the material. The whole analysis isn't complex enough to explain the 30 years! It's more so, that chessplayers often are like other human beings, they follow the mainstream. Real researchers are not so typical! Now, after that having said, where is your argument pro books in computers? Do you want to doubt that computers won't find certain solutions? Or do you want to reclaim that perhaps in 30 years they could find it, so we should assume that the were able - in principal??? Here would follow a clear nay. Today there is still a very principal horizon problem. Period. > > >> >>For me the development of computerchess took a wrong course. For me a >>self-learning system playing chess could be a better symbol of AI than the >>package which is simply not following the FIDE rules of chess. I'm talking about >>games between human players and comps. What were the reasons for the programmers >>to take the forbidden short cut? > >Finally, what is a chess program? Is it AI? No, although many have imagined it >would be such when it was in its infancy, it isn't AI. Is it just the engine >then? No, the program is the sum of its parts enabling it to play the best chess >moves possible from the first move to the last. There is no *forbidden >shortcut*. They are machines, programs, human constructs to give us the best >possible opponent when we have none. They are giant calculators. They are toys. >They don't compete, they are not animate. Competition is a concept of living >creatures. If I (and this is still ALL from those previous posts BTW) propose a >footrace between myself and an automatic motorcycle (just to remove the human >element), am I *competing* against it. On my end, perhaps, but the bike is >certainly not *competing* against me. It gets turned on and then shut off. >Computers are the same. > > Albert But here I fully agree. Well said. So, would you reject my ideas of a partication of comps in human tournaments? Rolf Tueschen > >> >>Rolf Tueschen
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.