Author: Christopher R. Dorr
Date: 07:26:39 01/08/99
Go up one level in this thread
On January 07, 1999 at 21:51:12, Bruce Moreland wrote: > >On January 07, 1999 at 09:15:29, Christopher R. Dorr wrote: > >>On January 06, 1999 at 18:42:44, KarinsDad wrote: >> >>>On January 06, 1999 at 16:37:42, Christopher R. Dorr wrote: >>> >>>> >>>>At these salaries, would they need any NetWare Engineers/weak Masters? I know >>>>someone who might be willing to help out :) Heck...they'd only need to pay me >>>>$250,000 a year....I'm flexible like that. :) >>>> >>>>I'd really like to see the results of something like this. Obviously it won't >>>>happen, but it would be interesting, both from the perspective of chess >>>>programming and from that of software engineering as a discipline. >>>> >>>>While we could never really know what would happen until this was tried, my gut >>>>instinct still tells me that the program they would put out wouldn't be that >>>>much better than the best competing programs from Ed, ChessBase, etc. >>>> >>>>How much work (on the engine specifically; I know they could do great stuff with >>>>the interface and features) do you think could be parted out to the other >>>>programmers? >>>> >>>>When I wrote my pathetic little excuse for a chess program, it had many of the >>>>components of decent programs (piece square tables, various extensions, decent >>>>opening book [never really completed], relatively complex evaluation function), >>>>I couldn't see where I could have used the help of someone as or less >>>>experienced than I was with chess programming. Admittedly, I wasn't a very good >>>>programmer, but I had read the literature, and dissected some of the >>>>source-available programs out there. >>>> >>>>Perhaps at the more advanced levels, some programming assistance could be of >>>>help, but from a software engineering perspective, I have my doubts about >>>>involving a team in this; I think you'd reach the point of diminishing returns >>>>*very* quickly. >>>> >>>>Chris >>> >>>Chris, >>> >>>A team such as this would have several things going for them. >>> >>>1) The Microsoft talent could be of the caliber that they could search the web >>>(and the ICCA journals), run everything past the rest of the team in >>>brainstorming sessions, and within a month, have a reasonable understanding of >>>the concepts involved in the current technology (i.e. get up to speed). I >>>downloaded about 8.5 Meg of thesis papers and other information from the web and >>>digested most of it in a weekend. Does that make me as knowledgable as Bob or >>>Ed? Of course not. But then again, I don't have them sitting in the office next >>>to mine. >>> >>>2) The chess programmers could be an interface between the MS engineers and the >>>GMs. The GMs could relate deeper chess knowledge, the chess programmers could >>>come up with ideas on how to implement that knowledge into a program and the MS >>>engineers could do the prototyping and proof of concept. >>> >>>IBM introduced it's new 332 MHz microprocessor last year, the fastest chip >>>available at that point on the RS/6000 SP. This chip is 5 times faster than the >>>ones used in Deep Blue. Using this chip and a quad configuration, Bob could >>>create a program similar to Deep Blue (since MS purchased the source) that does >>>16 million nodes per second (200,000,000 nodes per second Deep Blue * 4 >>>processors in a quad / 256 processors in Deep Blue * 5 times faster). >>> >>>If you assume that Deep Blue was running at a 2775 level and that this new >>>program is running on the above system at 8% the speed of Deep Blue, shouldn't >>>this new program be able to run at least at a 2700 level? How much more could >>>Bob do with a specialized team, a lot of resources, and a case of light beer? >>>The diminishing returns comes in when you buy the second case of beer. >>> >>>KarinsDad >>> >>>PS. I think I'm going to bow out of this thread now. It has gone from mildly >>>amusing to just plain silly. >> >> >>Your choice, of course, but I do find this interesting in several levels. >> >>I still don't believe that that having a team of programmers (as you indicate) >>would make things any better. What could they do, that Bob couldn't by himself? >> >>Their 'brainstorming' after a mere month of immersion would be likely to amuse >>Bob more than anything else. Just like the Application programmers >>'brainstorming' about compiler optimization (after a month or reviewing the >>literature)would probably have the experienced compiler guys wasteing their day >>saying "And that won't work, because..." fifty times a day. "Yeah, I know that >>problem isn't in the recent literature, but 10 years ago..." >> >>How exactly are the programmers going to help? I don't believe that you can >>'part out' a program like a chess engine because 1. It is relatively small, and >>2. the components of it are so interdependent, you can't simply set up the >>message passing stuff from one module to another and expect that they will work >>together at all. Yeah, they can do the interface, and the database, but the >>engine itself (which is what I'm talking about here)? I can't see how they will >>do that. I've done team programming before too, and understand that it simply >>isn't the optimal solution for many kinds of problems. >> >>MS Could develop a world class program very easily. They could buy everything on >>the planet. But they wouldn't make things much better. Team concept wouldn't >>apply well to this situation. If MS wanted to get in the biz, then the only way >>the'd make a splash would be to buy Rebel or Fritz or Deep Blue, fit it to a >>great interface, give it a nice database, and release it on their label. >> >>Chris > >I think the quoted post is quite correct. > >They would put an experienced person on the engine and leave him alone, or >license an engine or whatever. They would spend a lot of time on the UI. They >would spend a lot of time on internet play. The non-engine aspects of it would >be the point of it as far as they would be concerned. > >I agree also that this thread is sillly, but I'd rather think of the "what might >we expect if we hear that they are working on a chess program" aspect rather >than the "what would happen if Bill got obsessed with chess" aspect. > >They might make a product, but if they do, it won't put the high-end guys out of >business. It would probably have one of the commercial engines in it, and it >would be a mass-market program. The target market would not be chess players. >The target market would be software buyers, who are sub-tournament players. > >bruce Of course nobody expects that MS would really do any of this. I see this as more of a theoretical discussion about the nature of chess programming. Is it amenable to the 'Team programming' concept that you and others have spoused? We've already established that MS could make a great interface, and a great database, but the question on the table is 'Could they improve the state of the art, regarding the engine?' And I think the answer is 'No'. You can't part out a program like a chess program, and expect to improve over an engine developed by a good, experienced programmer working alone. You nor anyone else has given a plan for doing so. Do you have an idea about how to do so? You can think the thread is silly all you want, but the fundamental question is still unresolved, and nobody has produced any ideas about how to have 'MS rock the chess world' other than to but the best program in the world. If they are going to do that, then they haven't made things better, except for the chess programmer who is now rich. If they are going to keep one programmer working on the engine by himself, the it begs the question 'Is the chess programmer making $500K a year going to make a better engine than the guy maing $50 K a year?' Just talking about the strength of the engine, how do you do this? Do I do the evaluation function, and you do the search stuff? I don't think that will work. Even a relatively large company like MindScape recognizes that developing the engine in house would not be very effective, although they certainly could if they wanted to. They simply buy a strong engine, and fit an interface around it. Some programs (like a chess engine) are simply best written by a single person. GNU has had many talented people work on it over the years. Crafty has had only one person working on the engine. Which is the better chess program? Chris
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.