Author: Joe Besogn
Date: 00:22:19 11/09/00
Go up one level in this thread
On November 08, 2000 at 19:38:59, Bruce Moreland wrote: > >On November 08, 2000 at 17:47:43, Joe Besogn wrote: > >>I try you on Kuhn's requirements, and will answer a response tomorrow, since it >>is late here: >> >>normal-science before null move you describe as what? > >I will take you at face value and attempt to answer these questions. Please >bear in mind that I haven't read the book so my answers might not conform to the >book's religious perspective. If you wave the book at me I'm not going to be >able to respond properly. > >Before the null-move, there was no null move. If you read anything written >about computer chess before the Donninger article, you won't find an >enthusiastic article about it. Campbell wrote one where he left out a crucial >point -- that it made your program a whole lot better to do it -- while talking >at some length about how it could mess up. > >Bob supposedly did it, but that's not published, I think. Fritz did it, and >that's not published of course. Some of the other commercials do it or god >knows what, but it's not like they talk about it. The amateurs were in a >slightly different world. > >The reason the Donninger article worked is that he include a little table that >showed search results for one position, and he found a crucial move radically >faster. > >>what exemplars from normal science time became anomolies? Were they important? > >I don't know what this means, since you are using terms-of-religion. If you >mean, who didn't use it, an obvious answer is Deep Thought, which to my >knowledge did not use it. They mistrusted it, and I recall hearing that it >didn't work alongside singular extension. Their downplay of null-move is one >reason people wonder if Deep Thought and Deep Blue were really any good. > >Talk about being "safe" is the kind of talk you hear from someone who has >written a micro program that does sixr plies in the middlegame. > >If you told me I would get to use my quad Xeon box against someone running on a >286, I wouldn't remove null-move in order to search the same number of plies as >they do, but more "safely". No way. I'd leave it on full blast and drop bricks >on them from great height. > >>what was the crisis in normal-science? > >Another term of religion, I think. The crisis is that the amateurs got beaten >by the professionals. > >>what anomoly was then solved by the null-move technology? > >We kicked ass. > >>how were chess revolutionarily programs different afterwards? > >Richard Lang had to deal with being beaten by people he'd never heard of. Maybe >these people had invented this stuff or stuff like it previous to us, but once >that Donninger article came out, the genie was out of the bottle. > >>did anyone resist the null-move? politically? > >Some people mistrusted it. Ask Bob if he likes it and he'll say no. I love it, >it's free speed, and there seems to be little if any downside, regardless of >what kind of program you are trying to write. > >>can you say that null-move was more than just "puzzling" (kuhn's phrase) within >>the normal-science? > >What? Someone promoted an idea that got you 1-2 extra plies for free, with the >minor drawback that every once in a while you'd make a mistake. As far as I can >tell everyone jumped at it. Talk to people about it, there are a lot of people >who read that article. It was given some sort of prize as the best article by a >new author, and I completely agree. > >>Personally, I think it was great idea that required a real leap of thought, so >>I'm open to the idea .... >> >>I ask because there are of course developments evolutionarily. Kuhn claims these >>as puzzling in normal-science. The revolutionary changes, he says, take place as >>a result of crisis in normal-science. Presumably Kuhn sees paradigm shifts as >>being at the top end of grey scale of change, where the scale is gradual, but >>with a catastrophic over-the-edge-flip in some central region. > >I think it depends upon how you graduate your change scale. There can be a >change forced by new hardware, such that we have to throw away all our code and >write a new kind of code for something that isn't a Von Neumann machine. You >can have changes that force us to throw away alpha-beta search and eval >functions and all that. Or we can talk about one small improvement to >alpha-beta that everyone embraces and results in a quantum improvement. > >>The other quick thought, is that Kuhn talks of pre-science, where developers are >>working independently, without too much communication. The first paradigm arises >>from this. Some of the items you describe sound like this pre-science period. It is the time at the start of a science when diverse researchers are separately working without a paradigm yet established (paradigms become established by peer acceptance of techniques, processes, measurement systems). In the beginning these systems still need to be 'agreed' and established. Hence pre-science = no dominant paradigm yet in place = lots of separate paradigms. There's then a Darwinistic struggle in which the dominant paradigm wins. > >I don't know what this means. > There are some posts under I made which endeavour to explain it. In the thread "for those interested in new paradigm", or whatever it was called. I found your answers difficult because of mindset and word-meaning discrepancies, but I think what you're saying is that null-move was a revolutionary change because later "we kicked ass". Meaning Lang and presumably Schroeder programs. In Kuhn terms, the crisis state you describe was that 'amateur' programs were weak. Hyatt's response is that null-move was a typical field evolutionary development over 15 years, from Kaissa, Berliner, through Beal and Donninger. Null-move is forward-pruning, right? Lang and Schroeder did this forward pruning at that time by working on some chess knowledge at each node. Null-move does the forward pruning by a depth-reduced search at each node. Schroeder, I have read, doesn't use null-move, even today. So, if you claim that 'amateur programs' were in one paradigm before nullmove, and then another after, you are going to be implying that Schroeder/Lang programs are in yet another one, now, and then. Unless you argue that the two forward pruning techniques (knowledge vs search) are sufficiently conceptually identical that the 'amateur' paradigm merged with the Lang/Schroeder. Too messy, I think. Hyatt's model of nullmove as evolution seems stronger. imo. >bruce > >>>Other programs have had success with techniques that weren't thought to be >>>useful, for instance self-teaching. This hasn't started a wave of self-learning >>>programs yet, but there have been some interesting articles and some interesting >>>attempts. >>> >>>We will probably see more interest in speculative evaluation since Christophe's >>>speculative program has been a success. >>> >>>All the programs that I know of now are built on a brute-force framework, with >>>selective extension and selective pruning. If anyone can make a strong program >>>that doesn't use these mechanisms, that will cause the most major shift we've >>>seen so far. >>> >>>bruce
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.