Author: Uri Blass
Date: 01:53:18 04/10/02
Go up one level in this thread
On April 10, 2002 at 00:27:18, Robert Hyatt wrote: >On April 09, 2002 at 14:50:58, Uri Blass wrote: > >>On April 09, 2002 at 13:13:48, Robert Hyatt wrote: >> >>>On April 09, 2002 at 10:59:12, Mike Hood wrote: >>> >>>>On April 09, 2002 at 09:51:36, Robert Hyatt wrote: >>>> >>>>>On April 09, 2002 at 06:28:34, Uri Blass wrote: >>>>> >>>>>>On April 08, 2002 at 22:59:34, Robert Hyatt wrote: >>>>>> >>>>>>>On April 08, 2002 at 21:48:25, Michael Vox wrote: >>>>>>> >>>>>>>>On April 08, 2002 at 20:24:03, Dann Corbit wrote: >>>>>>>> >>>>>>>>>And if he were going to face Deeper Blue instead of Fritz, which would he claim is the stronger?< >>>>>>>> >>>>>>>>He already stated at the initial press conference that he feels Fritz on current >>>>>>>>hardware is stronger than DB. >>>>>>>> >>>>>>>>This commone knowledge. >>>>>>>> >>>>>>>>Regards >>>>>>> >>>>>>> >>>>>>>That is utter horse-hockey. AKA marketing hype and covering himself in case >>>>>>>the unexpected happens and he loses... >>>>>> >>>>>>Do you know names of GM's who say after looking at the games of deeper blue >>>>>>against kasparov that Deeper blue is better? >>>>>> >>>>>>Uri >>>>> >>>>> >>>>>Who won the match? That is the important data. When another program does >>>>>the same in a 40/2hr match of 6 games, then we can talk... >>>> >>>>I agree with you in part, Robert. Deep Blue vs Kasparov, the score sheet speaks >>>>for itself. My main contention concerning Deep Blue is that it was not allowed >>>>to play more games against other grandmasters and (this would be more relevant >>>>in 2002 than in 1997) against other computers. This has led to bloated estimates >>>>of Deep Blue's playing strength. I am sure that Anand, despite being "only" the >>>>world's second best player in 1997, is a better anti-computer player, and a >>>>series of Deep Blue vs Anand games would have relativised Deep Blue's abilities. >>>> >>>>I have no criticism of Deep Blue as an entity, but I am very critical of IBM's >>>>use of Deep Blue. As soon as the matches against Kasparov were won they couldn't >>>>dismantle the hardware fast enough. "Let's quit while we're ahead". What were >>>>they afraid of? Evidently IBM weren't as confident of Deep Blue's strength as >>>>they claimed to be. >>> >>> >>>Remember several important details: >>> >>>1. Deep Thought produced a 2650+ rating over 25 consecutive games against >>>GM players in 40 move/2hr games. These games were played in tournaments, not >>>matches. That was deep thought. >>> >>>2. Deep Blue 1 (and deep blue junior) played _lots_ of games vs GM players >>>at various exhibitions. I went to two consecutive SuperComputing conferences >>>(We had ACM events at these conferences several times) and DB Jr was playing >>>exhibitions against GM players at both. I watched it thrash Robert Byrne at >>>one, for example, in 4 consecutive games. >>> >>>3. Deep Blue 2 was barely completed prior to the 1997 Kasparov match. They >>>didn't play any games with it prior to the match, except for what they could >>>do to test the hardware... >>> >>>4. Deep Thought searched maybe 2M nodes per second. DB2 was 100x faster. >>>DB2 was far "smarter" also, according to documents published by Hsu after the >>>event. >> >>We do not know if deep blue was smarter because it is possible that it had bugs >>and wrong extensions that deep thought had not. >> >> > > >The extensions didn't change a lot. The eval did. But for _any_ program >you could make the point that it is possible that a newer version is worse. >But Kasparov played deep thought and had no trouble beating it. DB1 was >more difficult and DB2 beat him. There is room to draw a conclusion about >the three machines... I believe that DB2 was stronger than DB1 and that DB1 was stronger than Deep thought but being better does not mean having better program and it is possible that it was because of better hardware. It is possible that it was better inspite of inferior search rules or inferior evaluation. This was my point when I said that we do not know if deep blue was smarter and I did not mean to say that it was not better than deep thought. > > > > > >> >> >> If Deep Thought could produce a 2650 rating at 2M nodes per second, >>>how strong do you think DB2 might be? Deep Fritz on an 8-way box might well >>>be as strong as deep thought... maybe... but it isn't going to be much >>>stronger... and that is still a far distance from the speed/power of DB2... >> >>Deep thought could earn from the fact that it was not public and the players >>could not train at home to learn it's weaknesses. > >So... neither are other GMs "public" in this regard. Deep Thought played >games everywhere. At ACM events. On the internet (ICC or ICS as it was >known back then.) At public events. In chess tournaments all over the >country... I agree. The point is not to say that Deep thought did not play at GM strength at that time but that it is not fair to compare Deep blue's performance to the performance of the commercial programs against humans because of this reason. 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.