Author: Bob Durrett
Date: 12:19:41 11/13/00
Go up one level in this thread
On November 13, 2000 at 12:05:37, Howard Exner wrote: >On November 13, 2000 at 09:04:37, Bob Durrett wrote: > >>On November 12, 2000 at 21:30:07, Howard Exner wrote: >><snip> >> >>>These attacking engines are lots of fun and may prove advantageous over human >>>opposition. If programmers spend alot of time on exploring such possibilities >>>they might feel it a waste just to toss the engine out just because it does not >>>perform as well versus other computers. >> >><snip> >> >>This makes me wonder whether or not a program can contain two distinct blocks of >>code, one block playing chess the usual way and the other block being "an >>attacking engine" as suggested above. >> >>The idea is NOT to have both blocks of code working at the same time, each >>stealing computer resources from the other. >> >>Instead, this idea is to have some low-overhead evaluation function operating >>continuously [or just every now and then] whose purpose is to determine which of >>the two blocks is most likely to produce the best results. In positions where >>it seems likely that an attack would be appropriate, the "attacking engine" >>block of code would be turned on, but in positions where it seems unlikely that >>an attack would be appropriate the other block of code would be turned on. The >>block of code not in use may not need to be in RAM but, instead, would be copied >>from the hard disk when needed and erased from RAM when not. >> >>The real issue, it seems to me, is whether or not it is possible to create and >>use code which would perform a "low overhead evaluation function" without >>slowing down the program significantly. >> >>The envisioned benefit of this approach is that the programmer could "have his >>cake and eat it too," i.e. play strong both against humans and against >>computers. > >I've always thought programmers did this sort of thing in their eval. Have >knowledge that extends certain positions like king side attack patterns or >passed pawn scenarios. So in a sense they are changing mindsets but maybe not to >the extent that you are suggesting. The possibility of the "aggresive >personality" taking over may backfire too in positions that are not ripe for >that mode - the attack is ill conceived and dwindles away into a negative score. Sort of like a "dual-personality" chess engine? We may need psychiatrists [spelling=?]as part of the programming team? That brings up another concern: It is necessary that neither block of code be able to "take charge" and then ignore the potential need for changing back to the other block of code [i.e. "personality"]. The "low overhead evaluation function" would have to be part of some sort of executive code which maintained control of the overall configuration. Incidentally, this is just a special case of "adaptive software" in some general sense. [For the A.I. people, I apologize in advance for using the term "adaptive." For them, that word has very specific highly specialized meaning, but my use of the word "adaptive" is to be taken in a more general or intuitive sense.]
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.