Author: Uri Blass
Date: 06:06:27 09/10/02
Go up one level in this thread
On September 10, 2002 at 08:37:42, Rolf Tueschen wrote: >On September 09, 2002 at 20:19:42, martin fierz wrote: > >>On September 09, 2002 at 12:25:03, Rolf Tueschen wrote: >> >>>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. >>> >>>So, this is the ethic we all know and obey to. >> >>this is just not true. > >It is true. But thanks anyway for the many good arguments. > > >>in many exams at my university, we were allowed to bring >>along any material we wanted. the real question you want to have answered by an >>exam is: can the student give you the correct answer? it doesnt matter if he >>looks this up in a book, or figures it out himself. > >Martin, you know that this isn't true. We are now in a debate about definitions. >Depending on the field you are of course allowed to use material of different >well defined sources, but you might agree that e.g. a reader with the questions >and the perfect answers is not included! That makes the difference. It's a >questions of the parts and the whole. And if teachers want to examine your >solving abilities they don't want particularly test your memory. But let me add >also this. Real class, not to speak of genius, must also be able to do it >without any help at all. So, please, do not confuse the truth with the factual >situation in our education system. :) > >Hint: Take the difference between multiple choice questions and open questions. >Here you have the reason why medical experts are trained on such tasks where >they must exclude inappropriate solutions. That is the big part of their later >professional work. The rest is practice. > >But before you get annoyed by too lengthly talking let me repeat my critic >against tables. It's not the added material as such, it's the perfect answer in >the material. This is the scandal. > > > >> a student who knows nothing >>at all will not know what book to bring, or where to look up. >>once you start working seriously, not in artificial exam situations, it is >>always allowed to look up stuff you did earlier, or you read in a book but dont >>remember the details. > >This is perfectly ok. > > > > >> would you go to a doctor who tells you "i will not look up >>in my books if i don't know what you have got?" - i don't think so! > >Right. But not for the reasons you thought. I would consult a different expert >if the first would say that he didn't know, no matter what he promissed to do >afterwards. And not because he might be a weak expert but because my problem >surely looked complicated. So I would always like to hear at least two experts. > >BTW nobody would say what you said. Because you will always have situations >where you don't know _yet_. That is the daily routine then that is required. >That is simply called the process of making diagnoses. I would avoid the quacks >who already have the solution before they could have known the problem. Of >course certain problems can be solved by the authority alone of some guru. > > > >> >>>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? >> >>easy to explain... because your whole argument is completely arbitrary. >>computers and humans arrive at the same ends (playing a game well) by totally >>different means. you want to forbid the computers from doing something they are >>good at. > >The word ___ comes to mind. :) > >>if your arguments were valid, > >Of course they are, no? > > >> you could also say: "wait, a human can >>only look at 3 positions per second. therefore, all computer programs must be >>limited to searching at that speed." you CANNOT compare the way humans and >>programs work. > >Let's stay on topic, Martin. We were talking about perfect answers! Term comes >from the idea that many games will be (perfectly) solved some day. Would you >deny that a human player can also "look" at the perfect solution and then play >it? Where did you read that humans are forbidden to get some implantation? I >only read that it's forbidden to use books or outside help... > >Would you like to see what happens then? I tell you. Until the game isn't >completely solved human players will be superior! > >So, in consequence, let's forget about the implantation of perfect solutions, in >both, machine and human player. > > >> >>besides, for the endgame TB case in checkers, if i really cared, i could use my >>computer to produce a big set of rules which allow it to decide with 99% >>certainty whether a given 6-piece checkers position is a win or a loss or a >>draw. > >Then do that. Not for me, but for a demonstration a still not sufficiently >solved game. If it's solved, no human player would play you anyway. Then your >program would be a good trainer. (Martin, just between the two of us. Why do you >want to shorten the period until the perfect solution is found. Seems premature. >Just look at certain strategies in other economy fields.) > > >>the playing strength of such a program would be nearly the same as of the >>program with perfect information. so your argument against tables would be >>circumvented, and the program would be nearly the same as before. > >Almost you mean. Yes, why don't you do it?? > >>but give me >>one single reason to do this... just to satisfy rolf tueschen? > >Of course not. Well, it's difficult to tell if Idon't know you better. So all I >can say that I have two arguments. Science. You know that checkers is a game. >Now, implementing perfect data is no longer interesting. Ok, checkers is a weak >example, because it's almost solved, but think of more complicated examples. It >is not the perfect automatism that interests, but the thought process. A >computer with a mosaic of well chosen little programs for strategies to find a >good move. The moment you can define the best move or a sufficiently good move >for all concrete positions, the gameis over, because it's solved. But the >technical problem to create the relations between many little programs and then >to optimize the cooperation that is the scientifically interesting work. > >Second argument. Fair play. I make it short. Realising that humans are >especially forbidden to use books or tables with the perfect solutions, it's >simply rather primitive fould play if you argue "Fantastic! Because I umpute all >the things into my program and my machine it doesn't matter, it wouldn't even >attract attention if I would hide it, if I implantated the complete solutions of >endgame theory. I have plenty of space left!" You know that is like doping in >sports. You want to have advantages because you know that without youwould lose >to your opponents. Because the endgame, at least in chess is a real mess for >computers. > > > >>i have tested my program playing with 4-piece / 6-piece / 8-piece endgame >>databases. the difference in playing strength is actually much smaller than you >>might expect. what really improves is the quality of the evaluation, i.e. if you >>get to close to the db, the program with the db will just say draw or win, while >>the one without it will say +a little or + a lot > >Did you test it against strong human opposition? Yes, perhaps checkers is too >far developped already. You must go for either chess or directly GO in future. >Hint for the chessplayers here. HEX is also a fantastic game to play. Although >also solved by computers yet. > > >Rolf Tueschen The value of the endgame tablebases in chess is also small and programs probably get no more than 20 elo from them. I also think that checkers may be interesting in the future only if you increase the board. Increasing the board may be also a good idea also for chess. It is posibble to increase it in a random way for the opening position so programs are going to have problems to construct tablebases. For example you can decide: with probability of 50% you add the 2 squares a0 and a9 to the board. With probability of 50% you add the 2 squares b0 and b9, ... with probability of 50% you add the 2 squares h0 and h9. You can continue: With probability of 50% you add the squares i1 and (a-1)1 with probability of 50% you add the squares i2 and (a-1)2 ... with probability of 50% you add the squares i8 and (a-1)8 2^16 options to add squares with probability of only 1/2^16 to play regular chess. In that case working on tablebases or opening book is going to be a waste of time. Uri
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.