Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Dead Wrong!

Author: Robert Hyatt

Date: 13:56:43 07/21/00

Go up one level in this thread


On July 21, 2000 at 15:14:50, Ed Schröder wrote:

>On July 21, 2000 at 13:33:59, Robert Hyatt wrote:
>
>>On July 21, 2000 at 11:32:45, Ed Schröder wrote:
>>
>>>On July 21, 2000 at 11:03:13, Robert Hyatt wrote:
>>>
>>>>On July 21, 2000 at 07:42:00, Chris Carson wrote:
>>>>
>>>>>On July 21, 2000 at 07:06:08, Alvaro Polo wrote:
>>>>>
>>>>>>On July 21, 2000 at 01:11:57, Ed Schröder wrote:
>>>>>>
>>>>>>>On July 20, 2000 at 19:57:16, Jeremiah Penery wrote:
>>>>>>>
>>>>>>>>On July 20, 2000 at 19:11:03, Ed Schröder wrote:
>>>>>>>>
>>>>>>>>>[Event "DB-GK the rematch"]
>>>>>>>>>[Site "am Rd1; bm Rf5+"]
>>>>>>>>>[Date "2000.05.31"]
>>>>>>>>>[Round "?"]
>>>>>>>>>[White "DEEP BLUE"]
>>>>>>>>>[Black "Garry Kasparov"]
>>>>>>>>>[Result "0-1"]
>>>>>>>>>[WhiteElo "?"]
>>>>>>>>>[BlackElo "?"]
>>>>>>>>>[FEN "4r3/8/2p2PPk/1p1r4/pP2p1R1/P1B5/2P2K2/8 b - - 0 1"]
>>>>>>>>>
>>>>>>>>>DB played Rd1?? which caused a giant immediate material loss where Rf5+
>>>>>>>>>is the obvious defence. The most convincing argument is the score DB gave
>>>>>>>>>for Rd1?? way too positive (perhaps Uri remembers) and the very negative
>>>>>>>>>(and correct) scores some of the micro's gave when they did an analysis.
>>>>>>>>
>>>>>>>>Didn't someone say this one was caused by a C macro being improperly expanded or
>>>>>>>>some such?
>>>>>>>
>>>>>>>I don't know and I really don't care. All we have are a few games and
>>>>>>>hiding behind bugs is not a very convincing argument.
>>>>>>>
>>>>>>>Ed
>>>>>>
>>>>>>If what you are seeking is the truth you should care.
>>>>>>
>>>>>>Alvaro
>>>>>
>>>>>The truth is that Ed, Uri, and Amir are right.  DB had bugs
>>>>>and a simple eval (so that HSU could put it into ASICS, HSU
>>>>>was a HW guy Murry was the SW guy, trade offs were made to
>>>>>create ASICS).
>>>>>
>>>>>Best Regards,
>>>>>Chris Carson
>>>>
>>>>
>>>>You are obviously an ASIC expert?  And their claim of 8,000 adjustable eval
>>>>terms is therefore bogus?  And it was our imagination that it beat Kasparov
>>>>3 years ago?
>>>
>>>From the IBM site:
>>>
>>>    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.
>>>
>>
>>
>>What is the point?  "not using AI"?  Who is?  Who isn't using a simpler
>>search and evaluation compared to what was thought to be necessary 20
>>years ago?  Don't you use full-width?  Aren't you faster than Crafty?
>
>The point is the IBM site claims: DB (I quote)
>
>    "Deep Blue relies more on computational power and a simpler
>    search and evaluation function."
>
>A simple evaluation function. This is in contradiction with another
>claim (I now quote you)
>
>    "And their claim of 8,000 adjustable eval terms is therefore
>    bogus?"
>
>See above. This implies something else. Who made that "8000" statement
>BTW?
>
>Ed
>

Hsu and Campbell.  It was reported here by more than one person that
attended different lectures they gave.

BTW, you are overlooking a very important character in the above
quote...  the letter "r"

"simple" and "simpler" do _not_ mean the same thing.

maybe that is just a language problem.  But I can have a very complex
algorithm that is simpler than a more complex algorithm.  While it would
not be considered "simple" at all.

Seemed obvious to me, anyway...



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.