Author: Robert Hyatt
Date: 06:06:19 02/26/02
Go up one level in this thread
On February 26, 2002 at 01:14:12, Ricardo Gibert wrote: >On February 26, 2002 at 00:09:42, Robert Hyatt wrote: > >>On February 25, 2002 at 15:34:43, Uri Blass wrote: >> >>>On February 25, 2002 at 15:08:40, Robert Hyatt wrote: >>> >>>>On February 25, 2002 at 13:17:22, Uri Blass wrote: >>>> >>>>>On February 25, 2002 at 10:35:00, Slater Wold wrote: >>>>> >>>>>>I was recently in contact with Hsu, where I asked him if there was anyway that >>>>>>he would either a.) sell the technology in DB or b.) donate this information to >>>>>>someone (Hyatt comes to mind) who would put it to use and keep it safe from >>>>>>being commercial use. >>>>>> >>>>>>Basically he told me he only bought the rights to rematch Kaspy (who refused). >>>>>>And to keep IBM off his back, if he decided to make a Shogi engine. Period. No >>>>>>other reasons. He will never sell/commercialize/donate/share his information. >>>>>>Ever. >>>>>> >>>>>>What a terrible, terrible dissappointment. >>>>> >>>>>Not disappointment for me >>>>>I guess that the thing is simply not strong enough. >>>> >>>>That is absolutely the _worst_ reasoning I have ever seen. Do you also >>>>guess that when it rains when you have something planned, that the clouds >>>>have something against you? >>>> >>>>It was strong enough to smash computer programs for a long while. It was >>>>strong enough to beat kasparov in a 6 game match. I'll bet _other_ engine >>>>authors wish theirs was "not that strong"... >>> >>>I said *is* not strong enough and not *was* not strong enough >> >>Hsu has already written that his chess chip in .18 micron would search around >>30M nodes per second. I think everyone would find that plenty strong enough >>since it is 15 times faster than quad boxes... > >30M NPS and an EBF=4 will search a little less deeply as 1M NPS and EBF=3. At 2M >NPS... > Yes... and it will _also_ have fewer errors in the search. Given two searches, one with null-move R=3 to depth N, and one with no null-move, to depth N-1, I'll take N-1 every time... >> >> >> >> >>> >>>> >>>>IE they didn't win most every ACM event after 1986 because of of luck... >>>> >>>>> >>>>>It is not clear if the result of deeper blue against kasparov is better than the >>>>>result of Rebel against van wely if you remember that van wely trained a lot >>>>>against rebel before the match when kasparov could not train against something >>>>>similiar to deeper blue. >>>>> >>>>>Uri >>>> >>>> >>>>There's _still_ quite a jump from van Wely to Kasparov... And van Wely wasn't >>>>playing for a $1,000,000 prize either. >>> >>>I agree that kasparov is clearly better than van wely but van wely admitted that >>>he trained by playing 100 games against Rebel and the question in comparing the >>>results is how much elo you can get by preparing against a known computer and >>>not only against computers. >>> >>>The 1000000$ prize did not help kasparov to play better. >> >>I disagree. It was a _strong_ motivation. I would work _much_ harder to >>win 1M dollars than I would to win 1000. >> >> >> >> >>> >>>He played well in 4 of the games but in the games that he lost he did mistakes >>>that he usually does not do against humans. >>> >>>Kasparov never resigned in a drawn position against humans and he simply >>>believed that the machine is stonger than it's real strength(I guess that he >>>believed that Qe3 cannot be a draw because the machine could not blunder to let >>>him a tactical draw so he did not check it when it was clear that it had better >>>position) >>> >>>Uri >> >> >> >>I won't try to speculate on what he did or didn't think. But your idea doesn't >>make much sense. Why would he assume that some position was won, >>and then assume that the machine wouldn't make a mistake and allow a perp by >>Qe3? If that were true, wouldn't he have simply resigned at the start of the >>game rather than playing on?
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.