Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Khalifman and Gelfand on computer

Author: Dave Gomboc

Date: 07:52:29 05/22/00

Go up one level in this thread


On May 22, 2000 at 01:35:38, blass uri wrote:

>On May 22, 2000 at 00:05:20, Dave Gomboc wrote:
>
>>On May 21, 2000 at 16:51:47, blass uri wrote:
>>
>>>On May 21, 2000 at 13:52:58, Dave Gomboc wrote:
>>>
>>>>On May 21, 2000 at 00:01:24, Christophe Theron wrote:
>>>>
>>>>>On May 20, 2000 at 17:03:26, pete wrote:
>>>>>
>>>>>><snip>
>>>>>>
>>>>>>>>Another problems with the idea is that there is no way to check that there is no
>>>>>>>>hidden opening book.
>>>>>>>
>>>>>>>Of course there is. If the tournament organisers supply the hardware and check
>>>>>>>the files before tournament start. I think it's practically possible to agree on
>>>>>>>a certain file standard that would make cheating difficult.
>>>>>>>
>>>>>>
>>>>>>I really wonder how you want to do this :-) .
>>>>>>
>>>>>>I come with a single file "chessengine.exe " . How do you want to know if it
>>>>>>includes an opening book or not ? What should the programmer stop to let the
>>>>>>engine think some random time before playing out the book move ? Or should the
>>>>>>programmers provide their source code before the tournament starts ? And if you
>>>>>>say yes , think one step further and imagine the next logical step :-)
>>>>>>
>>>>>>Let me explain with another example : some year ago one of the popular topics
>>>>>>was how programs like Fritz or Goliath were said to be tuned for BS2630 or the
>>>>>>Nunn positions . How could one prove that ?
>>>>>>
>>>>>>I think if it is about limitting the programs I mostly like limitting the
>>>>>>hardware and think this is in fact really similar to the limits of Formula 1 .
>>>>>>
>>>>>>You could also limit the programs size and say for example all of its components
>>>>>>have to fit on a CD .
>>>>>>
>>>>>>But limitting how the programs should achieve their goals under given conditions
>>>>>>is too much I think . It is a battle of minds , that is what affects the
>>>>>>interest and attraction ; at this period of time an interesting one .
>>>>>>
>>>>>>The programmers with their ideas against the GMs with their ideas .
>>>>>>
>>>>>>That the automates achieve similar goals by completely different means is
>>>>>>another part of the thrill.
>>>>>>
>>>>>>If some kind of FIDE rules for computer-human events were agreed on it would be
>>>>>>a fair battle for all as the programmers could adapt.
>>>>>>
>>>>>>But randomly disabling certain program features just how they come in mind ( and
>>>>>>this is my feeling about the TB decision in NL2000 although it obviously had
>>>>>>zero influence ) just makes no sense to me .
>>>>>>
>>>>>>All this might or might not come in the future.
>>>>>>
>>>>>>But what really attracts the public most IMHO is that at the moment the GMs
>>>>>>still can compete with the programs when they run on whatever hardware they want
>>>>>>using every trick they can invent . And this still seems to be the case .
>>>>>>
>>>>>>So why not wait with the limits until the humans really need them ?
>>>>>>
>>>>>>This will come one day , agreed ; for example I personally like to play Shredder
>>>>>>with a rook in advance . Maybe some future day this will be the only way to be
>>>>>>competitive for the best human players too :-)
>>>>>
>>>>>
>>>>>This must absolutely be avoided.
>>>>>
>>>>>No strong human player will ever accept to play with an advantage. It's just a
>>>>>question of respect. It's like saying: "OK, I know you are not good enough for
>>>>>my program, so I give you a rook in advance!".
>>>>>
>>>>>Like a slap in the face of the human player!
>>>>>
>>>>>The only acceptable way to solve the problem is to have restrictions on the
>>>>>hardware.
>>>>>
>>>>>In particular, restrictions on the total amount of memory (including hard disk)
>>>>>the program is allowed to use. By making it small enough, it will not be
>>>>>possible to have big opening books and tablebases. The programmer will have to
>>>>>make a choice on what he wants to load on the computer.
>>>>>
>>>>>Maybe even choices on the opening lines he will load. That is similar to what
>>>>>the human player does: before an important game, he prepares for his opponent by
>>>>>doing his opening revision.
>>>>>
>>>>>
>>>>>    Christophe
>>>>
>>>>I don't see you offering any comparable restrictions on the humans' brains.
>>>>These have orders of magnitudes more processing power and memory than a
>>>>gigahertz PC!
>>>>
>>>>Dave
>>>
>>>I disagree.
>>>No human can compete with a 386 in simple actions like multiplying numbers.
>>>
>>>I believe that a 386 should be enough to win kasparov if people write the right
>>>program.
>>>
>>>Uri
>>
>>Who needs to multiply numbers when you can check 50000 piece patterns in
>>parallel?
>>
>>Dave
>
>No human can check 50000 piece pattern in parallel.
>
>I believe that a 386 has a huge advantage in speed relative to humans and when
>humans will think about the right ideas kasparov will have no chance against a
>386.
>
>Uri

How do you know that?  What if I said that a human with 50000 piece patterns
(e.g. a master chess player) checks them in parallel, in about the first second
or so that they look at a position?  I didn't say that the action had to be a
conscious one.  I am, however, reasonably sure that humans do it regularly.
(Well, in my case, maybe it's only 5000 patterns. ;-)

Source: experiments beginning with deGroot 1965, which may not jive exactly with
what I've suggested, but my recollection is that it was along those lines.

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.