Author: Mark Young
Date: 13:59:18 06/15/98
Go up one level in this thread
On June 15, 1998 at 14:52:32, Robert Hyatt wrote: >On June 15, 1998 at 13:52:33, Fernando Villegas wrote: > >>On June 15, 1998 at 00:14:19, Robert Hyatt wrote: >> >>>On June 14, 1998 at 20:11:08, Fernando Villegas wrote: >>> >>>>Bob: >>>>With all the respect due to your intelligence and merits as a programmer >>>>not only in chess field, but in computer sciences as an all, even so >>>>your words about this issue remember me that paper that was writen by an >>>>european scientist to show how flight by machines heavier than air the >>>>week before of brothers Wright would flight. Have you read it? Is a >>>>wonderful example of reasoning, you can bet. Heavier than air machine? A >>>>thousand reasons were against that. >>>>Fernando >>> >>>There is a big difference between "the bumblebee can't fly" or "an iron >>>bird can't fly" and my position on search. You only have to check the >>>results of any computer chess event you care to choose, and notice >>>whether >>>the tournament was won by a "fast searcher" or a "knowledgable >>>human-like >>>selective algorithm." >>> >>>My comments aren't "theory".. they are *reality*... >> >> >> >>Dear Bob: >>Why you must limit the discusion to what actually slow programs do? Why you must >>assume that a different approach involves necesarily slowness? Why you identify >>my arguments in favour of different approach with a defense of CSTAL or whatever >>current program supposed to be knowledgeable? > >this is a pretty easy question to answer. First, machines can not play chess >like a human, because we have *absolutely* *no* *idea* about how a human plays >chess. So that's "out". > >That leaves search and non-search approaches. I claim that any non-search >based approach is going to have to do just as much work as my search-based >approach, in order to come to the same conclusions. I don't say that knowledge- >based approaches won't work, I simply believe that the won't work *any better* >and, in general, will be worse because they will be far larger (lines of code) >with the attending larger number of special cases/broken cases/bugs... > > > >>I believe that a superior kind of approach will make a program faster, not >>slower. Sure it is as much as human brains can compete with computers and >>nevertheless the hardware is thousand times slower. So, that involves that the >>heuristic the humans use, not based in analysing move by move, are far superior. > > >But that's a pointless argument. For example, from my AI class that I teach >every year: Name a flower that rhymes with "nose". Everyone comes up with >"rose" instantly. I then ask how they did that, thinking in terms of their >computer science education. And after the usual answers, I quickly point out >that "surely you don't intend to tell me that I maintain a mental linked-list >of the phoenetic pronunciations of every word I know, linking all words that >sound alike, look alike, or have similar meanings together. Because I would >(a) have more pointers than data by several orders of magnitude, and (b) it >would horriffically slow to search them. So the brain does this (somehow) in >an associative manner, which machines can *not* do at present (and no, I refuse >to enter into the discussion about associative hardware memory which is *not* >the same thing at all). > >So I can't do it like a human, until (a) I know how a human plays chess, which >might not be understood for hundreds of years yet and (b) even then, after we >know how humans "think" it will take another hundred years or so to produce a >machine that can emulate that. So, maybe, in 200-300 years, your idea might >be workable, but at present it is not. > We don’t even know if every great chess player thinks in the same way. I find it interesting that some of the world champions and other strong grandmasters have played correspondence chess. In the correspondence game with much weaker players, the weaker players are able to play even if not beat the much more knowledgeable grandmasters. I think this shows that many people can play at a grandmaster level if given time to work out the tactics. So I’m not sure if grandmasters have some kind of high level of knowledge or they just know how to calculate tactics faster. > >>I canm bet that you yoursefl will be one of the guys to create a superior, >>different kind of approach. > > >There are many of us that might find new ideas, but my money says that for my >future days, any enhancements will be either in stop-gap approaches like the >knowledge trinkets I include in Crafty as needed or in new search ideas that let >it understand tactics even better... > > > >>Fernando
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.