Author: martin fierz
Date: 03:22:53 03/15/04
Go up one level in this thread
On March 15, 2004 at 05:06:09, Steven Edwards wrote: [snip] >So it is not surprising that some authors have an almost obsessive attitude >towards having the fastest move generator, the lowest branching factor, the >quickest evaluation function, and a never satisfied yearning for ever swifter >and more numerous processors. All of these can help improve the playing >strength of a traditional iterative A/B program. > >But little, if any, of the above has anything to do with artificial intelligence >or with making contributions to other areas of research. Neither is there much >here that can be seriously considered novel; the number of significantly new >ideas in the field over the past decade is embarrassingly small. The big >mystery to me is why intelligent people continue to work on the same old model >that is now thirty years old; indeed, a model older than some of today's >authors! > >I suppose that one of the attractions of the traditional iterative A/B approach >is that it is a low risk track. With relatively inexpensive high speed CPUs and >plenty of sample source on the Internet, a somewhat competent coder is almost >guaranteed produce a reasonable chess player in a matter of weeks. The problem >is the coder then spends the rest of his/her efforts on speeding up the program >(that's the source of the short term gratification); the alternative of slowing >down the program by tackling the harder task of adding chess knowledge is rarely >considered. Again, there's no surprise that nearly all of today's programs are >more or less variations of the same theme. [snip] your don't seem to like the traditional approach - "embarassingly small number of new ideas", "a mystery why intelligent people continue to work on...", "nothing to do with AI" and so on. i couldn't disagree more with you. -> the human brain is optimized for pattern matching, while a CPU is great at number crunching. is it really surprising to you that the *totally* different capabilities of the underlying hardware lead to different approaches for the software? it isn't to me... -> a quote by hans berliner i found on michael buro's webpage goes: "I consider the most important trend was that computers got considerably faster in these last 50 years. In this process, we found that many things for which we had at best anthropomorphic solutions, which in many cases failed to capture the real gist of a human's method, could be done by more brute-forcish methods that merely enumerated until a satisfactory solution was found. If this is heresy, so be it." -> you claim that the coder spends the rest of his time speeding up his program. this is most definitely wrong. all of us are concerned with speed, but all of us also know the rule "2x faster = ~50 elo". obviously, you can be 4 times slower than the fastest program and "only" get a 100 elo penalty for this. but most programs are much further than 100 rating points behind the best. the implication is obvious: the best programs are not the best because they are the fastest, but because they have better evaluations and better search algorithms (extensions and pruning). -> the reason why people use traditional algorithms? they work. it's survival of the fittest in algorithm world. if anybody had ever been successful at doing anything like you plan to do, people here might think differently. i don't know exactly what botvinnik did, but he also had ideas of making a great program which would be human-like in it's thinking. i certainly know that it never worked. i also know that all program based on a shannon B strategy died out. they existed, but they just didn't work as well as the more brute-force type programs. -> the bottom line: perhaps you are onto something great, we will see. until then, i (and i believe many others here) prefer to think you are somewhere between slightly and stark raving mad with your project goals. AFAIK you have absolutely nothing to show up to now (lots of code, but i'd like to see some chess moves...). AFAIK your engine has yet to make it's first move. it is a mystery to me how you can first implement opening book and tablebase access before addressing the real problems in your approach. it is also a mystery to me how you can spend lots of time posting here about what your program will be able to do in the end, when it can't do anything right now. good luck - you will need it... martin
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.