Author: Jesper Antonsson
Date: 09:21:08 11/18/01
Go up one level in this thread
On November 18, 2001 at 06:55:04, Otello Gnaramori wrote: >On November 17, 2001 at 20:22:37, Jesper Antonsson wrote: > >>On November 17, 2001 at 18:05:09, Otello Gnaramori wrote: >>>Hi Jesper, >>>The evidence that nowadays chess programs are better optimized than Deep Blue's >>>software is that when you are talking about D.B. you are actually talking about >>>a project stopped in '97. >> >>Excuse me, but what evidence? > >You probably don't know that have been some advancement in computer chess since >then or do you think that programmers are bound to old schemes with no new >ideas coming from the research world...? Yes, basically. Please show me one novelty since 1997 that give significant extra strength. I think there is none. Remember that you have a 200-fold speed advantage to offset. >>>Since then many things changed and evolved both in hardware and in software >>>research to make better computer chess machines: Hardware yes, in software only a little better eval and perhaps some extension tuning. Not even close to be enough. >>Sorry, but DB ran clearly superior hardware compared to micros of today. And >>software, please show me some new (since 1997) software algorithms that gives >>decisive advantages. > >This is taken from IBM D.B. FAQ : >http://www.research.ibm.com/deepblue/meet/html/d.3.3a.html#ai > >Does Deep Blue use artificial intelligence? >The short answer is "no." Earlier computer designs that tried to mimic human >thinking weren't very good at it. No formula exists for intuition. So Deep >Blue's designers have gone "back to the future." Deep Blue relies more on >computational power and a simpler search and evaluation function. So? They relate to AI in a text for lay-men and just state that they do approximately what everyone else does, basically a simple tree search with some, compared to humans, simple eval. >BTW If you want to be more informed about advancements I think you need to >search through I.C.C.A. and/or A.C.M. archives. If you want to convince anyone, perhaps *you* should search. I feel I'm quite up-to-date, and I would have heard if something revolutionary came by, since I have lurked here a long time, and before that in RGCC. >>>we have a clear evidence of >>>this fact looking at recent man-machine challenges with the outstanding results >>>of Chess Tiger in Argentina as an example, and with the incoming challenge >>>Kramnik vs. Fritz7 , unbelievable few years ago in the personal computer class. >> >>Why do you call a not yet played match "evidence"? > >Sorry , but the only fact to have organized it is significative IMHO. I think not. It signifies only that the Fritz team found money to do it. > And why is the "man-machine" >>challenges "evidence"? Has any computer but DB succeded in beating any of the >>top ten grandmasters at standard time controls in a six game serious match? > >Simply because it wasn't organized...in other words you need many funds to have >a match against any top ten players. Yes, but if any micro programmer thought they had a good chance to win, they would probably find a way to cough up money. But I think the Fritz team know that they are going to lose big-time, however. >>I think not. That top commercial micro programs fare well against GMs at fast >>time controls is nothing really new, and DB is even more superior in that >>domain, as faster hardware is worth more at fast time controls (due to >>"diminishing returns"). > >In Argentina the time controls weren't *fast*... I'm sorry to disagree on that >point. In Argentina, if we're talking about the same event, only one opponent were above 2500 ELO and the time controls were, if not *fast*, at least not as slow as they could be. (75 min/40 moves followed with 15 min finish with each player getting 30 sec extra per move. That is quite quick and favours the computer.) That tournament is quite different from a slower multi-game match against one well prepared super GM. >If I remember well there was a match organized between Rebel and DB junior... I >let you guess who win ... The rebel team claimed to have played one such game, but as I remember it, that was more or less bogus, it wasn't at all like playing a real DB junior. It was some version they had one the web for all to play and that can't be compared with meeting it one-on-one in a match. Regards, Jesper
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.