Author: Uri Blass
Date: 01:03:11 12/11/05
Go up one level in this thread
On December 11, 2005 at 03:20:05, Joseph Ciarrochi wrote: >Hi folks, > > When engines go through successive revisions, do the programers tend to add >things on top of what the engine already does (evolution; e.g.: add knowledge >or tweak paremeters but don't change the fundemental nature of teh program)? Or >do they ever start from scratch and do a radical rewrite of the whole program >(intelligent design.)? > >If engine development is more like evolution, than it seems like each engine is >inherently limited in how far it can develop, depending on how well it was >designed to begin with. e.g., maybe fritz 9 can't go much farther, without them >throwing the whole engine out and starting again with what is learned from fruit >and rybka? Rebel 15 and crafty seem to no longer be able to keep pace (), which >suggests that its original design is inherently limited and perhaps can't make >use of all the new programing tricks. > >This would also mean that newer engines, like fruit and rybka, have much more >potential to improve, given they have only just been born (excuse all the >metaphor). So in three years, we would predict that fruit and rybka will improve >more than fritz 9, hiarchics, and shredder (which are old in computer years) > >Any thoughts? I think that there is no limit based on design even if you do not rewrite the program from scratch. You can get the same result if you rewrite the program from scratch and if you replace part after part. one change by itself may not be considered as changing the fundemental nature of the program but set of many changes may be considered as changing the fundemental nature of the program. It is possible that rewriting the program from scratch is a faster way to get big improvement but I do not believe that there is a limit that you need to throw the all engine in order to get progress. Uri
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.