Author: Peter Fendrich
Date: 16:50:45 01/29/99
Go up one level in this thread
On January 29, 1999 at 13:42:56, KarinsDad wrote: >On January 29, 1999 at 10:27:58, Peter Fendrich wrote: > >>On January 29, 1999 at 09:51:36, William H Rogers wrote: >> >>>Would todays top chess program be as good as or better if they were played on >>>the Deep Blue Machine. >>>The heart of the Deep Blue Machine is hard wired mover generator. It is capable >>>of making millions of moves per second, while todays top programs rely on chess >>>logic, so if today top programs were using millions of moves per second combined >>>with their own logic, I believe they would destroy Deep Blue or any other Chess >>>Master. >>>Bill >>I can't agree... >> > >Although I cannot agree with Bill, I cannot agree with everything you said >either. You didn't seem to agree with anything I said :) >>1) It's absolutely impossible to transform the pc chess programs around to Deep >>Blue hardware. > >Why? The Deep Blue software was written in the first place, why can you not port >all of the ideas (if not the identical code) from the current PC software? If ideas are bad or good are tightly connected to what kind of hardware you have. It's not just ideas from nothing. Extemely fast parallell arhitecture with "wired" modules gives new options and possiblities and in the same time makes some of the techniques used in pc's obsolete. I wouldn't call it a transformation or a port. It would be a completely redesign. I would probably try to invent extensions that I would't dream of in my old pc! I wouldn't do nullmoves until down in the tree or maybe not at all. I would probably redesign my hash tables completely to meet the better memory bandwith. I would probably redesing my move generator of the same reason. And this, and that... If one didn't do it like this, DB's peformance would be outstanding and impossible to reach. >>2) If it would be possible, not one line of code would be optimized to use that >>kind of hardware so the performance would be a disaster compared to Deep B. > >Yes and if the Deep Blue hardware had no C compiler written for it, you would >have to either write one or re-write the code (similar to what Bionic did) into >a supported language. And, you would have to re-optimize the code. That does not >mean that the basic algorithms of the current PC software could not be used and >would not be better. As I said before it is not a port and not just an re-optimization it is a total redesign. > >In fact, the opposite postulate may be true. Since current PC software is so >limited on speed in comparison to Deep Blue, it is more probable that the >developers of that software have had to add elements to their algoritms to >improve their programs understanding of the game to make up for their more >limited speed. Take this example: If I only reached 3 plies with a limited evaluation function (just an example!) and had the choice of including 3 times more evalaution code or reaching 4 plies, I would certainly go for 4 plies. On the other hand, if I reached 20 plies with the limited evaluation function and had the option to reach 21 plies or include 3 times more evaluation code, I would certainly go for evaluation. Why should I assume that he DB team is more stupid than I'm? :) > >>3) If Deep B. didn't have a great deal of chess logic, despite its speed, they >>would have been crushed by Kasparov - they were not... > >I disagree here as well. Benjamin helped the team add a lot of chess logic to >the code for the second match. That is true. However, even before he did this >and when Deep Blue was running at half of the speed and with less knowledge, >Kasparov did not crush Deep Blue in the first match. He won, but he did not >crush (and most other GMs would probably not even have won). From what I have >read, the first incarnation of Deep Blue didn't have as much chess knowledge as >current PC software programs. The second incarnation had more knowledge, but it >also was 2x as fast. Twice as fast tends to correspond from 50 to 70 ELO, hence, >maybe it won more due to the speed increase and maybe not as much due to the >knowledge increase. Who can say? They had less chess knowledge in match 1 than in match 2. That doesn't mean that they had none in the first match! I don't know how much they have now or had then but they would be stupid to not take the opportunity given from high speed and stuff the monster with knowledge... > >>4) Todays pc chess programms, spend a great deal of effort to get faster and >>deeper - the cost is less chess logic and risky optimizations. Deep B don't have >>to do this. In fact they could concentrate mainly on chess logic between the two >>matches played against K. >> >>My conlusion is that Deep B must have a lot of chess logic inside the shell, >>compared to todays pc pogram, probably more than most of them. > >Probably less than most of them, at least for the first incarnation. When you >improve the speed by a factor of at least x1000 (over current PCs), you need >less logic, not more. The razoring and other pruning techniques are not needed >as much just as you said. However, just a standard search at great depth without >special knowledge code can produce a very strong chess program. x1000 will give you a very strong program even with just a piece/square evaluation, but that's not the point. It's all a question of what would give most, one more ply or more knowledge. > >> >>//Peter >>The speed gives you possibilities to concentrate on chess logic, not the other >>way around. > >Yes, it does give you that option. However, that does not mean that an increase >of chess logic beyond what some of the commercial programs have was done. > >Of course, this is all speculation on both of our parts. Speculation it is... //Peter > >KarinsDad :) > >>
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.