Author: Don Dailey
Date: 15:30:18 06/12/98
Go up one level in this thread
On June 12, 1998 at 17:02:06, Thorsten Czub wrote:
>On June 12, 1998 at 15:10:17, Don Dailey wrote:
>
>>As humans, we talk about tactics and positional play and try to view
>>them as separate things. But I agree with Mark, that in reality
>>there is no difference.
>
>??? I am completely crushed by your statements.
>Although I now understand your , better - our problems here.
>We live in different worlds.
>"as human, we talk about tactics and positional play and try to view
>them as separate things" - this sentence circles in my mind. I find it
>very funny.
>If I would go to my chess teacher and tell Wolfgang:
>He - Wolfgang, as humans we talk ... he would laugh about me.
>He would say: Thorsten - you have been to long with all these
>programmers.
>And I guess he would be right IF I would say something you, and MArk and
>Uri says.
>You are all materialist ! And your statements prove this crystal clear.
Hey, just relax a little. I am simply being pragmatic and accurate.
Everyone knows chess is ultimately tactical and I'm just stating the
obvious. Since I consider myself to be a scientist I will always
collect facts and definitions and work within them. I find it
extremely useful to step away from the natural human definitions
you prefer and try to look at things from "outside" this. Human
biases often mislead.
In my post I called for new terminology. Bob just defined tactics
as sequences of moves that win material, and this is the classical
definition of it. He is completely correct. But I want to move
away from this definition because the definition is not strict,
it is open to interpretation. My point has several aspects:
1) So what if it wins material, does it win the game?
2) What the hell is material anyway? Saying you win material
is a judgement call since no one knows in any given position
what any piece is really worth. Is the bishop pair material?
3) Why is there a distinction between "material" and other good
things (like rook occupation of the 7th rank?) I don't think
there should be.
4) At what depth does a positional winning move become tactical?
Why is a move that wins a knight (and eventually the game) in
7 ply a tactical move while a gradual king side attack (which
will eventually cause a knight to be sarcraficed to win the
king) is a positional move? The classic definition is flaky!
Do you see my point?
>> Really, the only thing that exists in chess
>>is won positions, drawn positions and lost positions.
>
>?? I would say, if you cannot calculate until the end (from the root)
>you will have to build superpositions of the 3 degrees too.
Of course. Who said any different?
>> A given move
>>can take you from one kind to another.
>>
>>Also, this means there is no such thing as a great move, unless you
>>consider it to be any move that preserves the win if the position is
>>a win or preserves the draw if a position is a draw. I suppose we
>>could loosely define a great move as the fastest win if it exists.
>
>But thats the main axiom, IF IT EXISTS. The main axiom is:
>WE DON'T KNOW if it exists. If we would be able to do so, chess would
>not be that interested anymore.
>
>>We humans muddle the distinction. If my program does an incredibly
>>deep search and finds a way to improve the position of the knight
>>as a result, it will be applauded as having great positional ability
>>when in fact all it has is technique.
>
>Sorry, i guess this way computers will never make any progress. You try
>to rebuild nature by recalculating all chances. I don't think this is
>ok.
>Or it would work. The main strength in life does not come from
>precalculating but from errors, forgetting about facts, and evolutionary
>recombination.
>You have to teach your program about chess, not about programming
>techniques.
>I know you will disagree. I know you will tell me rebuilding the human
>way is not the right way. I know. But you are wrong.
You do not know what I will tell you here, and if I told you what you
expect me to, you do not know if this is right or wrong. You are
guessing. But in fact I do not know if a more human approach is
better or not. My intuition is that humans are different from
computers and that raises the possibility of a better approach,
perhaps not human at all but also like nothing we are currently
doing. I prefer to stay open minded about this and do not pretend
to know the right answer.
>Computers will not jump to another level of intelligence. In the moment
>they are less intelligent than a polyph. Even this primitive life-form
>FORGETS. And this is its main strength.
>It is the main reason it lives.
There is a school of thought here that claims we are very arrogant
to define intelligence by our own standards. I have noticed over
the years that everytime a computer gets good at something, we
stop viewing the same behavior as intelligent. Super fast human
calculators used to be viewed as intelligent. A few years ago
it was widely held that playing a strong game of chess was viewed
as something not possible without intelligence.
We will never be satisfied. Big hash tables (something you seem
to oppose) was a big step forward in making programs more
intelligent. It made very wise use of memory, something we humans
take for granted. When you study a position, you become familiar
with it, and when you see the position again you only have to recall.
At one time the only "good" programs were brute force. People
tried to write selective ones but were not successful. Finally
though, people hit on the null move assumption and suddenly
programs were playing more human like (with selectivity.) But
it seems now that people are viewing these programs as brute
force. At some point programs will be just like us and then
we will think we are stupid too!
>>I think we need new terminology to express these things that makes
>>it more clear what is really going on. Saying positional play, and
>>tactical play is misleading because humans tend to think (incorrectly)
>>of tactics as material winning, or game winning.
>
>What do you want to differenciate when your point of view does not need
>or is unable to see anything into different things.
>You call anything tactics, since chess is finite.
>Point.
>Thats all you can do. Give him the biggest computer in universe and
>thats all.
>What do you want to differenciate ?
>Your point of view has no range to differenciate ! You believe junior is
>strong. And Virtual, and Genius. And than you have to say that Genius or
>junior or virtual were mistaken. And than you say: if they would have
>thought/considered more time on the right (in your case : WRONG
>calculated) moves, your program would defend against this trap.
>This is true. But it would run into the next trap.
>And you don't have a computer that will defend any trap.
>So why do you always try to argue this way ?
>Give me enough time and my program sees it too.
>Brilliant. This way we don't need to talk about chess anyway.
>
>> And they think
>>of positional play as something not involving wonderful tactics
>>which is also completely ridiculous in my opinion. This terminology
>>make people categorize programs as fast and dumb, or slow and smart,
>>which is very simplistic in my view. In my opinion, being able
>>to win a queen in 20 moves is great positional play too.
>
>In my opinion YOUR categoriation of chess as TACTICS is very stupid.
>And ridiculous. Your terminology and you method of asking for more time,
>than your program will see the problem, is ridiculous. You don't have
>MORE TIME.
>This is chess. You have to move in 40/120 or 60/60 or even faster and
>more limited when it comes to 5' blitz. So what ? You don't have
>infinite time.
>You have limited time. Now you will say: I need a faster computer than.
>Aha !
>But this will only help you in this position.
>
>
>>It turns out that most of us have programs that can play perfect
>>chess as a result of the search mechanism (given enough time of
>>course.)
>
>Nonsense. No program I know plays perfect chess. I know no program
>playing perfect chess, despite the fact it comes to a position where a
>forced win or a forced mate is the problem. But these situations are
>very rare. The main problem comes BEFORE the position is lost. Don't you
>see this ?
>Ah - I know your answer : for these positions I need a faster computer -
>or more time.
>
>Ok - I will wait until it is 2010 !
>
>
>>The positional heuristics do not help much here, and that
>>is why I reluctantly consider the search the heart of each chess
>>program and the biggest source of their strength. Run any modern
>>day program on a 8086 without hash tables (which are mainly a search
>>speed advance) and you will see exactly what I mean.
>>
>>- Don
>
> don't need this. I have given you 3 examples of very easy and primitive
>plans against top programs. 2 from a human beeing, 1 by CSTal.
A lot of these plans work against most humans.
- Don
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.