Author: KarinsDad
Date: 12:53:16 07/02/99
Go up one level in this thread
On July 02, 1999 at 14:27:19, Sarah Bird wrote: [snip] >> >>I think that poppycock is a fairly strong word. Fritz 6 running on a >>multi-processor OS is much more than just a slightly improved version of Fritz >>5. So, assuming that Fritz 7 is just a slightly improved version of Fritz 6 and >>not a major rewrite is incorrect. The consumers have no idea how many changes >>and to what extent there are between one version and the next, so they have no >>idea how much is different. >> >Is it alledged to be the object of the tournament having either software or >hardware available to the everyday person in the commercial marketplace. Even if >that were the case and it was written for and played on a PII 450 not everyone >can afford a PII 450. The hardware it is being used on therefore is available >(if you can afford it). The same goes for the software if you have enough money >you could buy it regardless of experimental or not. > >>Why do you think that Fritz 7 should be allowed to play on a 256 processor >>system where for it, next year's tournament is equivalent to about G800 (13+ >>hours for his side to move) when compared to Fritz 6's current 8 processors at >>G25? >> >As noted above, the hardware is available it is purly and simply a money issue >what is a lot to some, may not be a lot to someone else. I believe the computer >should be able to make use of any processing power available, including custom >built. > >>Where do you draw the line? >Why draw a line, computers playing in these tournaments will improve software >possibly hardware to a point where better products will be available to the mass >marketplace. The reason to draw the line is not one of whether we can build the better mousetrap. We know that we can. We know that if you put Fritz 6 on a 256 processor system that it will outperform Fritz 6 on an 8 processor system (on average assuming that there is no bug or sideeffect which causes the 256 processor system to play worse). The idea is totally one of comparison. People claim that Kasparov will be a different person in a year. Better, stronger, more theoretical novelties, etc. However, there is no proof of that AND Kasparov can only play one game against Fritz at a time. So, he cannot pull his hundred new theoretical novelties out in a given game. He can only pull out one novelty and then only if the game heads towards a position where he has a novelty to pull out. Mostly, at the superGM level, it is NOT a matter of pulling out a novelty against a program. The program doesn't get pysched out. Kasparov just has to play consistently good chess throughout an entire game to win. The program on the other hand CAN be placed on a 256 processor and can be 32x faster than today. The human cannot even be 2x faster. He may know a little more, but chances are that for any given position or even game variation, he will know about the same next year as he does this year. Once you get to the playing ability of Kasparov, it's not about outplaying your opponent. It's about springing theoretical novelties on him and taking extremely small mistakes that he makes and taking advantage of them. However, half of this (the psychological half) is taken away when playing against a computer since there are no real "theorectical novelties" against a computer. Effectively, you are saying that the Fritz TEAM won the right to play whatever they want in the Giant section of the tournament next year. I am saying that they won the right to play Fritz 6 on the same hardware in the Giant section next year. I compare using a different program next year to a chess coach having the World Junior champion as a student, but the next year he has a better student, so the coach won the right to send whomever he wishes the next year in order to achieve the best results as opposed to the student winning the right to go back the following year. Can you understand the difference? I thought it was clear with the steroid example. It's real easy when talking about computer chess to not compare apples with apples, but with any fruit we can find. Although humans and programs are SO different, it is easy to categorize them as the same. Using improved hardware next year is like a stock car race where one contestant wins the right to race a certain race next year, but brings in a downhill only track to run on. The playing field would not be level. The same applies WITH RESPECT TO THIS TOURNAMENT ONLY for the right to play in the Giant section next year. The Chessbase people have the right to play Fritz 7, 8, 9, etc. on any hardware they want in other tournaments, but only Fritz 6 on the given hardware has won the right to play next year. >> >>It really is starting to get similar to having a human racing a motorcycle now >>that multi-processing systems are being used. Within 5 years, computers will not be competing in any human tournaments anymore, even 40/2 1 tournaments. >> >I welcome that point in time. I see nothing wrong with having the best >chessplayers in the world being made from silicon. As i see this as a natural >evolution anyway. Your statement reads like "let's not help them", they need no >help. It will happen one day it is purely a matter of time before chess is >solved. Also once that day were reached it don't mean the end of chess, the >human mental capacity wouldn't be able to remember the endless variations >available. However it will help human play a better game. I did not mean to imply that I do not want the best program to be silicon. I went off on a tangent here. > >>So, yes, Chessbase will be able to build a better mousetrap by next year and >>probably win the Giants section of the tournament. Big deal. It only proves that man is ingenious enough to build a machine to do something that only humans >>could previously do. That's been done before. >> >Isn't that the purpose, ultimately isn't that what we all want. No. I could care less if Chessbase is able to win in the Giants section or any other tournament for that matter. In fact, I am actually more interested in my own program and could care less about the commercial ones (with the exception of how my own program plays against the commercial programs). > >>Your comment of "The same as the GM can improve his/her opening/ middlegame/ >>endgame knowledge" is somewhat naive. After studying for many months for a major tournament and a lot of hard individual work, superGMs may pick up 10-20 elo rating ability (not actual points). >> >Totally disagee. Naive !! are not new variations of opening played year after >year, have not computers already taught us more about the endgame. Did I say by >how much he/she could improve their performance, no. Anyone would readily agree >that study will prove beneficial by how much is a seperate issue. My point is not that we are making improvement. My point is that the SLIGHT amount in which superGMs will improve by next year after a lot of hard work will be miniscule compared with the amount chess programs will improve mostly due to multi-processing and bigger badder hardware. > >>But improvements on a program could consist of setting up automatic tests, >>putting it on better hardware, and minor programmng tweaks which leverage the >>work of many many people and may not really consist of a lot of hard work from >>one individual. But, the program can improve 40 to 80 elo within a year by doing this (granted, major improvements such as going to a multi-processor system do take a lot of hard work). >> >I hope by next year Fritz will be 150 elo better, but again I see nothing wrong >with it. There is nothing wrong with that. > >>So, once again, comparing chess improvement between people and programs does not take into account the effort involved, but rather just the results. As long as we can get a program to beat the superGMs, that is all that matters, not the >>method in which we do it. >> >I would say I generally agree with that. The object for me is to produce >programs capable of beating humans at any time control. See above for my >opinions on the hardware and software implications and/or effect there of. > >>Fritz 7 running on a new set of hardware is NOT Fritz 6 running on the current >>hardware (or would you like to dispute the results of the SSDF for the last 10 >>years). So, only Fritz 6 qualified. You could run Junior 18 and say, "Oh well, >>it's running the same GUI, so Junior really qualified.". Again, where do you >>draw the line? One change or a thousand changes, it's still a different program. >> >So, it will be a different Kasparov and different Anand I don't see the deal >with it being a different program, who cares and why. Would anyone really be >that interested in whether Fritz 6 is able to beat Kasparov next year when >chessbase are sitting on Fritz 7 who isn't allowed to play. Let's for argument >sake say last year Fritz 5.0 qualified to play in this years event. With Fritz >5.32 known to be better (commercial) and other programs being tested would we be >that interested in the results. If it won, you would be interested. But, this gets back to my point in my earlier post. People do not CARE about fairness. They care about money and program strength, etc. That is the crux of why you and I have opposing views here. You care about program strength. I care more about equity and fairness. You do not see where fairness comes into the equation. I do not see why people have to always be on that surfboard of bigger, better, faster, badder. C'est la vie. > >>Upgrading Fritz 6 to Fritz 7 is like putting an athlete on steroids and saying >>"Well, it's the same person, it doesn't matter if we boosted his performance. >>It's only the results that count.". > >No, steroids are dangerous to human health. There is a big difference to the >life of a person than a computer frying. >> >>I guess you and I just have a different idea of fairness. >This is what it boilsdown too, since I don't see fairness as an issue. Fair enough. KarinsDad :) >> >>KarinsDad :) >Sarah :-)
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.