Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: EIGHT new poll questions now up....

Author: G.Mueller

Date: 06:26:56 01/29/99

Go up one level in this thread


On January 29, 1999 at 00:31:49, Robert Hyatt wrote:

>On January 28, 1999 at 23:01:31, Steven Schwartz wrote:
>
>>On January 28, 1999 at 20:32:43, Robert Hyatt wrote:
>>
>>>On January 28, 1999 at 17:27:21, Steven Schwartz wrote:
>>>
>>>>In our attempt to make up for lost time,
>>>>we have just put up EIGHT new poll questions!
>>>>
>>>>Please click on the link to the "Computer Chess
>>>>Resource Center" at the top or bottom of this
>>>>page and place your votes.
>>>>
>>>>The latest SSDF results are also on the WCCR
>>>>in the Resource Center.
>>>>
>>>>- Steve (ICD/Your Move)
>>>
>>>
>>>The 'deep blue' question is really a bad one.  The idea just doesn't make
>>>a lot of sense.  IE if I took a word processor, made in work on my blender,
>>>would I have a better automobile would make just as much sense.
>>>
>>>DB is _not_ "just a fast machine that a program can run on."It _is_ a silicon
>>>embodiment of what the DB guys think a chess computer ought to be.  _if_ you
>>>could take a commercial engine and modify it to work on their hardware, do you
>>>know what you would have?  Easy answer:  "Deep Blue".  Because you would have to
>>>do things exactly as they do things.  Evaluate the same things, use the same
>>>kind of search, the same data structures to fit the hardware...
>>>
>>>ie it is a silly question...
>>
>>We'll add it to our list of silly questions:-))
>>Would it have been phrased better if we said,"If you took the best
>>commercial program and run it at the relative speed of Deep Blue, would
>>it be stronger or weaker than Deep Blue is now?"?
>>- Steve (ICD/Your Move)
>
>
>even that won't quite do...  NPS is not NPS here.  IE the best answer no
>matter what is that you would just get 'deep blue'.  Their eval is all hardware.
>to run at their speed would need their hardware, or else some truly remarkable
>cpu speeds.  IE they do 200M nodes per sec minimum.  Commercial programs are
>roughly 200K nodes per sec today.  That is a factor of 1,000.  When you factor
>in what they do in their eval in parallel, we need another factor of 10 to make
>up that difference.
>
>_then_ we could compare.  But a factor of 10,000 isn't coming for a _long_
>time, unfortunately... :(
Hello Bob!

For me this experiment you planed with Ed to compare, what a factor of 100 in
time brings was interessting, so sad that it does not work...
Best wishes
G.Mueller



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.