Author: Robert Hyatt
Date: 10:09:45 05/10/98
Go up one level in this thread
On May 10, 1998 at 11:29:33, Fernando Villegas wrote: >Dear Keith: >First, let me give you my warmest thanks for your extended and detailed >piece of information about your talk with Hsu. >Second, let me tell you that I feel very well to know that my enthusiasm >for CSTASL, showed here until the limit of fanatism, is shared by a guy >like Hsu and, If I can say it, by you. >Third, it also very flattering me to know that Hsu used a method of >evaluation baed in second degree levels that is not far the method I >proposed lot time ago to some programmers and in WCCR I think also, one >that I called "modular", a method based in the simple fact that >intelligent behaviour is not made out of an exhaustive calculation of >factors put together in an specific moment, but a feedback provcess >where in different moments different kind of progressive more detailed >factor are taken into account on the gvround of previous calculations. Note that this is *not* a new idea... IE we had a three-level evaluation in Cray Blitz for many years... some simple independent factors to be sure, but many that depended on interactions... IE king safety was initially defined as the pawn structure around the king, which was combined (non- linearly) with the pieces that were close to the king, which was then combined with the ability of other pieces to reach the king quickly. Ditto for weak pawns which depends not only on their being weak, but on their mobility and ability to advance and exchange themselves away... most good evaluations have such terms... It's just that DB can do them with no regard to computational consequences, while the rest of us have to decide what we can afford to do, and how we "make up" for what we can't afford. There is *nothing* they can't afford to do. They simply have to want to do it... >I >remember I putted the example of a tennis player that does not try to >calculate precisely the trayectory of the ball in the very moment the >adversary strike it with the racker, but take a serie of decisions in a >sucesions of moments; first, the general direction of the ball, right or >left so I begin to run right or left; then , the altitude, low or high, >etc, etc, until all this with corrections received from the very >trayectory the ball is really taking, etc... Of course, not being myself >a programmer, I never received about that the minimal commentary, not >even to reject it. Neither I am saying "This" what Hsu did. It just >enogh for me to suppoose than a vague intuition is not so far the real >thing made out for the the best chess computer of the world. >Finally, after all you have said I almost cannot wait the moment Mr Hsu >will launch his product to the market, if if he does it at all. It will *definitely* change commercial computer chess if he does, of course, which may or may not be a good thing. Time will tell. The only down-side is that the WMCCC event will likely go the same way as the WCCC and ACM events have apparently gone. When someone dominates to a level that makes it totally impossible to compete, interest dies, although some of us would still continue to see what can be done with commodity microprocessors. But we'd get killed by "the big dog" of course... :)
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.