Author: Bob Durrett
Date: 17:45:49 11/02/02
Go up one level in this thread
On November 02, 2002 at 20:15:52, Peter Berger wrote: >On November 02, 2002 at 20:04:07, Bob Durrett wrote: > >>On November 02, 2002 at 19:47:10, Peter Berger wrote: >> >>>On November 02, 2002 at 18:12:26, Bob Durrett wrote: >>>>>>: ) : ) : ) : ) : ) : ) : ) : ) : ) : ) : ) : ) >>>>>> >>>>>>Bob D. >>>>>> >>>>> >>>>>That's a Turing Test and there is nothing funny about it IMHO. I don't know how >>>>>good and professional the "Cheater Cops" at ICC do their job but if you really >>>>>managed to let everybody believe that your chessengine is a human player of >>>>>course this would mean that your engine really plays human-like IMHO. >>>>> >>>>>Peter >>>> >>>>I know the Turing Test is Sacred among AI people. It is absolutely taboo to say >>>>anything bad about it. However, it is fatally flawed! It falsely assumes that >>>>the human mind is the ultimate thinking machine. Typical of the human ego to >>>>think that! >>> >>>How has that anything to do with what I posted? >> >>Perhaps I misunderstood you. I thought you said [paraphrased]: "There is >>nothing funny about the Turing Test IMHO." >> >>[Are we communicating???] > >No, we are not. You gave a very good, although unethical experiment to test your >own hypothesis, namely to emulate a human player with a chessprogram on ICC - if >ICC and other players payed attention to this account, this would be a good and >valid experiment to test if computers are able to play human-like IMHO. > >> >>> >>>I don't think there is any chessprogram that can really emulate a strong human >>>chessplayer so that it will fool a reasonable amount of humans at expert level, >>>that's it. That's all I posted , that's all I wanted to discuss. >> >>OK. But there's not much to discuss about that with me, since I agree with you >>100%. Today's chess engines are not yet stronger than the top human GM. We >>agree. Maybe others would have different ideas about this than ours. >> >>Bob D. > >No Rolf, I didn't say that at all. I have posted zilch about playing strength >and in fact I don't even have a clear opinion about this topic at all myself. OK. I see what you're saying. You say that no current engine can be used to emulate a human player enough to fool anybody at or above the expert level. I believe that's true. It would take a design modification, starting with a much stronger engine than currently available. I see human chessplayers as being very flawed creatures. If we were, somehow, able to obtain a chess engine [possible in the future?] which was MUCH stronger than the strongest human of that time, then maybe it could be reprogrammed to make human-like mistakes occasionally. If the statistical occurrence of those mistakes matched the statistics for humans, then it might be extremely difficult to distinguish such a modified chess engine from a human. Of course, there must be an absence of computer-like moves because such moves would reveal the true non-human nature of the engine, as someone noted here recently. But the use of an extremely strong engine [of the future] might take care of this. I see such an "extremely strong engine," MUCH stronger than the strongest human, as being a real possibility sometime during my remaining lifetime. Hence my interest in this topic. Please recall that this discussion began with my expressed "USER Desirement" that a human-like sparring partner be created by the chess engine programmers. I would like to have such a silicon sparring partner, and the sooner the better. Bob D. > >IMHO the much more interesting experiment would be to emulate a 1800 human >player in fact - here playing strength is a no-issue but the principal problems >remain the same. Since my rating was USCF 1856 at the time I quit playing chess, I would benefit greatly from having a "silicon human-like sparring partner" performing at about the 1800 level. I don't know if I am representative of all USCF 1800-1900 level chessplayers. But in my case, my opening repertoire had major holes in it and so did my endgame knowledge. I believe an ELO 2500 chess engine would be plenty strong enough for use in a "silicon sparring partner" for an 1800 player. The idea would be to make the engine err almost every move. The errors must be typical of 1800 players and occur in the games consistent with the statistics for an 1800 player. The other characteristics of human 1800 players, such as fatigue, should also be emulated. Of course, the engine programmer would have to get rid of those "computer-like" moves, such as Bf8 in the Kramnik vs DF match. Bob D. >> >>> >>>> >>>>As an aside: Would you say that a lightning bolt did not exist if no one saw or >>>>heard it? >>> >>>As I don't even know what a lightning bolt is, that is beyond my abilities. >> >>It's an electrical discharge involving the motion of an extremely large number >>of electrons between two clouds or between a cloud and the Earth's surface. >> >>[You knew that.] > >I didn't - my English is not that good. But I understand it now. > >Peter
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.