Author: Christophe Theron
Date: 08:34:09 12/12/01
Go up one level in this thread
On December 12, 2001 at 08:50:48, Gordon Rattray wrote:
>On December 12, 2001 at 08:22:44, Steve Maughan wrote:
>
>>Gordon,
>>
>>>>The least important kind of experts from the list above are the chess experts.
>>>>
>>>>Yes, sorry.
>>>
>>>
>>>I'm not entirely sure what you mean by this, i.e. it's possible for something
>>>to be the "least" while still being vitaly important, it's just that other
>>>things are even more vitaly important. ;-)
>>>
>>>I agree that chess knowledge isn't essential in order to write a decent chess
>>>program, but that doesn't mean that having chess knowledge wouldn't make it
>>>even better or easier to write. For example, computers are generally weak in
>>>the area of long term strategic planning. Maybe if more grandmasters wrote
>>>chess programs (assuming they were strong in the other areas of course!) this
>>>aspect could be tackled more. i.e. how does a programmer program "planning"
>>>functionality if they're not very good at it themselves?
>>>
>>>Also, chess knowledge must help while testing a program. e.g. if it loses a
>>>game, where did it go wrong?
>>
>>I think there is a subtler point that Christophe is making - that is *too*
>>*much* chess knowledge can sometimes hamper program development. A program that
>>is written by a chess master may be too ambitious in the knowledge that
>>contains, especially early on in the development. The programmer may find it
>>inconceivable that a concept such as 'tempo', is left out of the engine. This
>>results in a slow knowledge rich engine that is outsearched by the dumber
>>competitors.
>>
>>It is interesting to note that most, if not all, of the top programs are written
>>by sub-experts i.e. with a rating of 1400 - 1800 ELO. It would seem that this
>>range is 'optimal'. Authors with strong chess ability never seem to quite make
>>it to the top e.g. Chris Whittington, Vincent Diepveen. Maybe Vincent has the
>>best chance as he is also a professional programmer.
>>
>>Regards,
>>
>>Steve
>
>
>What you say is true, but it assumes that the developer is weaker in some of the
>other areas (e.g. game theory, program design). I said "assuming they were
>strong in the other areas of course", so the grandmaster/programmer would know
>about things such as knowledge versus speed.
>
>I'm not suggesting that chess knowledge is the most important factor, but I
>think it could play an important part in conjuction with other aspects.
>Consider, for example, if a grandmaster offered to assist a chess programmer as
>and when the programmer saw fit. Surely this would be a significant benefit?
>
>I just don't think that chess knowledge can account for next to nothing. But
>maybe nobody was suggesting this anyway, as my interpretation of "least" (see
>above) highlights.
>
>Gordon
I'm not saying that chess knowledge accounts for nothing. I just dispute the
fact that chess experts have chess knowledge:
1) that would be useful for a chess program
2) that a weak player like me cannot either find by itself or find around him
(by asking to better players or finding it in books)
I'm doing chess programming since a long time and I don't think I would have
benefited much from a grandmaster by my side, if I had been given the
opportunity to have one.
And as I have said many times, I even believe that a gransmaster involved in a
chess programming team would be counter productive.
Human players have to rely on heuristics often designed to compensate for some
weaknesses of the human brain. Computers do not have these weaknesses, and so
should not use these heuristics. An obvious example is the ability for a
computer to compute thousands of complex variations without fearing to miss one.
The human brain cannot do the same. I believe the concept of tempo is also an
example.
On the other hand there are heuristics the human players use that are of too
high level at this time for computers. I guess the concept of planning is one of
them. Human players are able to build on their past experience to come up in a
given position with a set of plans that are most likely applicable. That's
something I have no idea how to do with a computer. I guess other teams are
working on this idea, but it seems that nobody actually succeeded.
Christophe
This page took 0.01 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.