Author: Don Dailey
Date: 10:15:16 05/11/98
Go up one level in this thread
On May 11, 1998 at 03:38:24, Danniel Corbit wrote: >On May 11, 1998 at 01:19:15, odell hall wrote: >>I was wondering If anybody thinks that perhaps karpov would be a better >>opponent for deepblue. Since his style is positional and it's hard to >>get anything tactical on him. If such a match were to occur what are >>your match perdictions >This is very much a "How many angels can dance on the head of a pin?" >question. I don't this question is in the same category! But I get your point that it still is just speculation and I agree. > Despite that, I feel compelled to answer anyway. Deeper Blue >beat the best player in the world. So matching a lesser player would be >silly. Computers are not afraid of styles. Computers ARE afraid of styles, at least in their own heuristic way! Matching a lesser player would NOT be silly, we are not talking about much lesser, but we are talking stylistically much different. Any way, I'm not picking on you, your point that other players are weaker should definitely be a consideration. > Computers never blunder, >unless programmed incorrectly and never get tired or psyched out. I think they do blunder but I'm nitpicking here a little. They blunder in a positional sense, everyone has seen them play ridiculous horizon moves or completely misunderstand a beginner position. However this is getting rarer and rarer. >Deeper Blue has been dismantled since the publicity stunt had its >impact. They have nothing to gain by playing again. Hence, we will >never see this match nor any like it. Despite this, Deeper Blue would >probably win anyway. With any player, Karpov, Anand or whoever, they >can load in every game that they have ever played to tune the computer >just for that opponent. You would prepare the same way for a special >foe, but probably would not remember every facet of every game like a >computer database can. A computer can tirelessly analyze day and night >the favorite openings and tendencies of a given player and store that >data for future use. Again, not to be disagreeable, but I disagree with you here! Tuning for a particular opponent has a minor impact at best. Probably 90% of this is done with special opening preparation that may or may not be particularly helpful. I spoke to Murray Campbell about a month ago and he said there was a LOT of opening preparation for this match and it was all pretty much wasted. I pointed out that I didn't think it was because it may have had the effect of limiting Kasparovs opening options. But we'll never really know. But in principle it should be possible. I just think the technology to super tune against a particular opponent does not exist for computers yet. > Wow, with a garden variety computer program, it >will still get clobbered. But Deeper Blue was not a garden variety >computer program. Computer programs for PC's can already trounce most >club players. We don't really know if Deeper Blue is the equivalent of >a GM, but we do know that it can be tuned to beat the best player in the >world. It does not make sense to believe that it cannot be tuned to >beat the second best or fourth best or tenth best or any other player. Same arguments. >This is not the same thing as entering a tournament and playing a large >collection of players. And it is also possible that Deeper Blue has a >fundamental weakness of some kind that repeated games would uncover. I definitely agree on this one. Players will get a good sense of their opponents strengths and weaknesses with time. >But such a weakness can be corrected in the software too. >But all of this is blowing smoke anyway. Deeper Blue has had all of >it's jim dandy chess processors jerked out and is now doing database >queries for someone. IBM has _nothing_ to gain and _everything_ to lose >by playing another match. In short, it will never happen. What a shame! Your response was well thought out, in my opinion you overestimate the tuning aspect, although the DB team itself does claim some work went into the specific tunning. By the way, I think reasonable tunning for matches against humans in general is possible and I have done a little of it myself. I have no real sense of whether it helped or not. There is some anecdotal evidence that certain programs are better against people than others programs (relative to each other.) I have my doubts the effect is large but I believe it probably exists. - Don
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.