Author: Don Dailey
Date: 10:19:54 05/13/98
Go up one level in this thread
On May 13, 1998 at 02:43:40, Hristo wrote: >>There's nothing fundamentally "flawed" about the way we're doing it now. >>It's led to extremely strong programs. It also seems to be more in >>harmony with what computers do well and what they do not do well at. >> >>The fact that we haven't solved chess is a good thing. There is still >>a challenge. But it may well be that another much better approach to >>chess exists which we have yet to discover. An interesting thought >>experiment is to ask the question, "how strong is the strongest chess >>program?" What do I mean by "strongest chess program?" Every possible >>program that can be written by a person, already exists in principle >>but is simply waiting to be discovered (and most never will.) So >>there exists a "very strongest one" or if you prefer a huge class of >>them all among the very best possible. Let's add the constraint that >>this program must fit on a 32 MEG modern Pentium machine. I have >>already written this program unless you also add the constraint, "very >>best program at 3 minutes a move." >> >>Would this hypothetical program play perfect chess? I think it either >>would, or it would come very close. At the very least it would blow >>away anything we have now. If it was search based (which seems >>plausible but let's not limit ourselves) it would have powerful pruning >>algorithms to >>avoid ridiculous lines of play with little or no risk. It would be >>capable of doing more than just search and evaluate (just like we do) >>and ... WAIT! I think I've figured it out!!! No just kidding. >> >>So I agree that we should keep looking for something better and not >>be afraid to try new ideas. >> >> >>- Don > >Don, Fernando, >I don't know if the current approach to solving the Chess game is >fundamentally wrong. >"We are also in alchemy times." Yes Fernando! I couldn't agree more. We >read those postings about *tweaking* this and *tweaking* that. Well >after years of *tweaking* it seems like the objective(of the game) has >been lost. So can we start with redefining the OBJECTIVE. > >1. What is the objective of the chess game? >2. What determines the path to achieving this objective?(the rules?!) >3. How are the chess-pieces different from each other?(do not apply >common knowledge, since it will lead us to predetermined and possibly >incorrect assumptions) >4. What is the relative *weight*(mass) for each chess-piece in relation >to one another?(Consider a clear board with only one chess-piece on the >board at a time) >5. How do the chess-pieces of the same color relate(interact) to each >other on the board?(Consider chess board with only white chess-pieces) >6. How do the chess-pieces of opposite colors relate to each other? >7. How do we evaluate a chess-position based on the answers to the above >questions? It's pretty fashionable to call for a completely new approach and this has happened continuously since computer chess began. But I don't see how your questions are new in any way. Everyone has been trying to answer these and each program reflects it's own way of answering them. The assumption is that we haven't made any progress and MUST be doing something wrong. And it sounds great in print. It makes the proclaimer seem like he has some special insights too and yet no solution is offered nor is any proof offered that the current approach is all wrong. But what has happened with our current "boring" approach is amazing. We've made incremental (but substantial) progress over the years in software design and major progress in hardware. Didn't a computer just beat the world champion in a match? I don't think this calls for a totally fresh approach, although I certainly would encourage great new ideas and breakthroughs. Now I can already hear some of you saying that hardware does not count. This is a predjudice point of view based on our view that brute force is not elegant. But if this is true, then the human brain is not elegant either since it contains massive (brute force) storage capabilities and untold billions of parts. Faster hardware IS elegant and totally legitimate. It just doesn't tickly your fancy. So why don't we put things into perspective a little and recognize the success of our approach. This still leaves plenty of room to speculate philisophically about breakthroughs and eventually scrapping our current approach for something far better. It's my hope that this someday will happen. If it does though, chess may very well be too limit a domain to relize the full benefit of this new approach. Perhaps GO or other games and endeavers will be. >There is a huge amount of other questions that one can ask. But how >about if we just answer these questions first! Remember to write down >all other questions .... > >Don, you take too many things as a given ... "to avoid ridiculous lines >of play" ... "perfect chess" ... Huh? Reread my post, this was all speculation, that's what we are doing now right? >Imagine we take all chess programs >currently available and take them out of the *human* context of things, >meaning we will not evaluate them against humans. Let those programs >play ... you will see that every program will *lose* at least one game. >Program that plays *perfect chess* will never lose unless it plays >against a program that plays *perfect chess*( we don't even know if >chess is a theoretical DRAW or win for WHITE(black)). This is elementary game theory and is nothing new. We don't have to imagine taking all the chess programs and playing them against each other because this is already being done. Actually I'm losing you here, I don't understand your point in this paragraph. > >Cheers. >Hristo But anyway, this stuff is all well and good and I fully support it. I am doing speculative reasearch on chess too and have in the past. I think most chess programmers do to one degree or other. I hope someone does write a great "right brained" chess program that can write poetry and philosophize while simultaneously playing chess. It would be a wonderful thing. - 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.