Author: Robert Hyatt
Date: 09:05:25 05/29/99
Go up one level in this thread
On May 29, 1999 at 10:41:35, Dave Gomboc wrote: >On May 29, 1999 at 10:11:19, Robert Hyatt wrote: > >>On May 29, 1999 at 09:48:05, Dave Gomboc wrote: >> >>>On May 29, 1999 at 00:21:16, Robert Hyatt wrote: >>> >>>>On May 28, 1999 at 19:25:15, blass uri wrote: >>>> >>>>> >>>>>On May 28, 1999 at 18:27:20, Dann Corbit wrote: >>>>> >>>>>>On May 28, 1999 at 18:15:30, blass uri wrote: >>>>>> >>>>>>> >>>>>>>On May 28, 1999 at 15:08:15, Dann Corbit wrote: >>>>>>> >>>>>>> >>>>>>>>If Hsu enters next year, they will have to have three categories. >>>>>>>>0. Hsu (they can start polishing the brass now, I think). >>>>>>>>1. Supercomputers >>>>>>>>2. Microcomputers >>>>>>> >>>>>>>I think that Hsu should be in the same category as Supercomputers. >>>>>>Why will they even bother to show up then? Unless all of the supercomputers >>>>>>borrow Hsu's chips as add-ons like IBM did, they would have no chance >>>>>>whatsoever. >>>>> >>>>>I do not agree that they would have no chance. >>>>> >>>>>Super Computers are becoming faster every year and even if Hsu has hardware >>>>>advantage(I am not sure about this) they have chances by doing a better program >>>>>or by luck. >>>>> >>>>>In the last WCCC microcomputer(Fritz3) won 2 Supercomputers(Deep thought and >>>>>Socrates) and won the championship. >>>>> >>>>>It is not clear to me that Hsu will win. >>>>> >>>>>I believe that programs can be significantly improved and that there is a chance >>>>>that in 2000-2002 another supercomputer or even a micro will be better than Deep >>>>>blue chip. >>>>> >>>>>Uri >>>> >>>> >>>> >>>>to make this short and succinct: >>>> >>>> "not a chance in hell..." >>> >>>Fortunately, chess players are generally not the sort to resign before a game, >>>even if their opponents happen to outrate them by a few hundred points. >>>Otherwise, there wouldn't be much of a tournament. >>> >>>Unlike some people, I have a high amount of respect for their achievement. This >>>does not mean that I think their winning is a foregone conclusion. Conclusions >>>are what are made after an event, and if the arrangements a team of people make >>>prohibit them from competing, those arrangements prevent them from winning too, >>>and of course, from people making conclusions about their play, because they >>>don't play. >>> >>>I expect that if they were playing, they would completely thrash everybody. My >>>expectation, however justified, isn't really worth squat, because neither is >>>this trememdously big "if". >>> >>>Hsu's company will have different needs than IBM, and I hope this will result in >>>a lot of games being played that are publicly verifiable by the chess card, with >>>the likely result that people will be able to infer from its play that DB was >>>very, very strong. Not being in a position to guarantee such a thing, I won't. >>>That is a job for Hsu's company. >>> >>>Dave >> >>My "not a chance in hell" was a response to the last sentence... "by 2000-2002 >>another supercomputer or even a micro will be _better_ than DB". DB can always >>lose a game. But in 2002 there won't be a program around that can win one of >>10 games vs the thing... Because they will still be 2-3 orders of magnitude >>slower, particularly if Hsu gets his 10-15 times faster chip redone... That >>will put them another 3-4 years (at least) ahead of the pack, beyond where they >>already are. > >Okay, but by 2002 someone else could decide to build custom hardware. Of >course, Hsu has a big head start. :-) I don't know how much the IEEE Micro >article would help a potential competitor or not, I ought to get to the library >so I can read it for myself. 2002 is probably too soon for anybody to catch >him, but maybe not if someone decided to throw obscenely huge amount of money at >it. Then again, the easiest way would be to fund him, so... that is possible, but not what was said. Custom hardware could be done, although it wouldn't catch up to deep blue, because it is non-trivial to design that hardware, and they have a 10+ year head-start on doing so. However, the original "a supercomputer or microcomputer" won't fly. There isn't a cray on the planet that is within a factor of 50 of the speed of their custom hardware solution. IE the fastest we have ever seen on a Cray is 5M nodes per second. They are almost doing that on one chip, and they have 479 more to use beyond that. :) > >Okay, if Hsu's PCI card appears on time and on spec, something good enough to >win the next WCCC will include one or (probably more) of them. I'll go that >far. > >Dave
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.