Author: Robert Hyatt
Date: 20:32:40 05/08/00
Go up one level in this thread
On May 08, 2000 at 21:53:31, Hans Gerber wrote: >On May 08, 2000 at 18:20:51, Robert Hyatt wrote: > >>On May 08, 2000 at 16:09:41, Hans Gerber wrote: >> >>>On May 08, 2000 at 16:02:51, Robert Hyatt wrote: >>> >>>>On May 08, 2000 at 15:27:15, Hans Gerber wrote: >>>> >>>>>Are you _really_ convinced that cheating could _not_ be excluded? In the past? >>>>>In general? Forever? >>>> >>>> >>>>Absolutely, yes. >>> >>>Well, I have to believe you. Now, the whole problem get's into a new light IMO. >>> >>>Thank you very much. I'll have to digest that one. >> >> >>Here's a way to think about this. >> >>You and I are going to play a match (human vs human). I am going to cheat, and >>use an SMP crafty to play my moves. I won't tell you how, yet. So let's see if >>you can come up with a plan to prevent this from happening... a plan that is >>realistic (ie we can't play the match on mars). >> >>you already know I am going to be using a box that is not going to fit in my >>pocket (a quad xeon). How are you going to detect or stop my cheating? Let's >>go from there... >> >>then we have the remaining issue of instead of me playing, it really is crafty >>sitting there at the table. How do you make certain that Crafty plays every >>move in the game, with no outside influence (same problem as stopping my >>ability to cheat in our match). And having it replay every game won't reproduce >>the same moves each time, so that a 'validation run' won't work. >> >>Together, those two issues say "can't be done" to me... > > >Perhaps I'm not the expert to find a solution. Why not invite the scientists to >invent a possibility. You argue as if such a research would be a waste of time. > >Let me ask another question. If a solution for control existed would you accept >my point that normally the scientists had to examin tsuch questions since a long >time? Would you then accept my opinion about Hsu and his interpretation of >science? No... because the solution doesn't exist, which means that the logs are just pieces of paper that won't prove cheating, nor will they disprove cheating. As such, their importance is really only in giving us some insight into what DB could do, things that many didn't know (depth, etc). As far as Hsu, you are on the wrong person. Hsu didn't have _any_ control at the match. He designed and assembled the hardware. He (and others) wrote the software. But legal and marketing folks took control because they realized how valuable the P/R was going to be, particularly if DB won, but even if it lost. > >As a little note let me state that your stressing the impossibility of a >continual repetition of the moves in 'parallel' systems is not necessary because >humans also do vary in their judgement. What I'm trying to say is that control >should still be a realistic goal. But if the computer is non-deterministic in its behavior, _how_ will you ever prove whether it played some particular move or not? And if you can't, you just lost any chance of using the logs (which Kasparov wanted) to prove that it either did, or did not, cheat. > >A second note to your example one where you played connected with a box. IMO >human chessplayers of the top ranks would never "cheat" with the help of such >boxes. It would lead too far to discuss that in a computerchess group. On the >other side we have scientists and technicians who don't have a high education in >chess. They should make some reflections on how to find controlling mechanisms. >This should be done facing the future reality that the machines get more and >more stronger. The machine should be able to "play" completely on its own >through a complete match. If you want to have it with different personalities, >then do it in advance. At that moment chessplayers would accept a machine as >genuine 'chessplayer'. Then the problem of cheating will be solved. You should look at a tournament played last year. In a well-known scandal, someone used a computer program to whack GM players like flies. He was a 2300 player himself I believe. He had a TPR over 2600. So yes, humans will cheat, given the chance. As far as "on its own" how would you confirm that? How to be sure that there is no 'access'? IE no rf link, no magnetic link, no laser link, no sonic link, no optical link, etc...
This page took 0.01 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.