Author: Miguel A. Ballicora
Date: 20:51:33 12/28/01
Go up one level in this thread
On December 28, 2001 at 21:57:03, Christophe Theron wrote:
>On December 28, 2001 at 02:30:07, Tom Kerrigan wrote:
>
>>On December 27, 2001 at 23:11:24, Christophe Theron wrote:
>>
>>>On December 27, 2001 at 20:12:02, Tom Kerrigan wrote:
>>>
>>>>On December 27, 2001 at 13:10:08, Christophe Theron wrote:
>>>>
>>>>>On December 27, 2001 at 02:13:34, Tom Kerrigan wrote:
>>>>>
>>>>>>On December 26, 2001 at 17:47:12, Christophe Theron wrote:
>>>>>>
>>>>>>>I don't think so, but I think at some point the only way to improve will be to
>>>>>>>incorporate a way for the program to learn without the programmer, to remember
>>>>>>>its experience and improve on it, and to adapt its play to its opponent.
>>>>>>
>>>>>>I don't see this as likely because of the numbers involved.
>>>>>>
>>>>>>For a computer to recognize features, it has to loop over them. And there might
>>>>>>as well be an infinite number of features possible on a chess board, so that
>>>>>>loop is going to take a while.
>>>>>>
>>>>>>The human brain is sloppy and bad at this task, so maybe there's some way to do
>>>>>>sloppy and bad learning that does better than what we have now, but I wouldn't
>>>>>>know how to go about that.
>>>>>
>>>>>
>>>>>
>>>>>That's the problem, I do not know either.
>>>>>
>>>>>But the reason we know that there is another way of playing chess is because
>>>>>human chess players have a "NPS" that is a very small fraction of todays
>>>>>computers NPS. The best human players only look at 1/100000 of the nodes a
>>>>>computer looks at in order to play a move of comparable quality.
>>>>
>>>>I have always believed that this "human NPS" stuff is a load of crap.
>>>>
>>>>I do not play chess well and I'm not close with anybody who does play chess
>>>>particularly well, but the idea that a human would "visit" distinct "nodes" to
>>>>"search" for a good move strikes me as absurd. Maybe this is what strong humans
>>>>actually do, but I have a hard time believing that.
>>>>
>>>>-Tom
>>>
>>>
>>>
>>>I think that's what they do, but the problem is that they visit probably more
>>>nodes "inconsciently".
>>>
>>>So the order of magnitude of the human NPS is not very well established. It can
>>>be in the order of 10 if you count the nodes the player knows he has visited,
>>>but it could be in the order of 100 to 1000 with the ones he has visited without
>>>actually knowing it.
>>>
>>>But anyway, this number is still 100 to 1000 times smaller than the NPS of
>>>todays PCs.
>>
>>I still believe this is apples and oranges.
>>
>>I can stare at a chess position for a long time, think about it, and pick a move
>>without consciously visiting any other positions. Does this mean I'm visiting <<
>>1 NPS?
>>
>>Various experiments (most involving image/sound recognition) have proven that
>>our subconscious is much less sophisticated than people believed a few decades
>>ago; if I'm not willfully visiting different positions, I doubt I'm doing it
>>subconsciously.
>>
>>I've read that Kasparov searches 3 NPS. Does he think real hard about a position
>>for a third of a second, willfully change the position in his imagination, think
>>about the new position for a third of a second in pretty much the same way he
>>thought about the previous position, ad nauseum?
>>
>>If that's not what he's doing, then I disagree with the 3 NPS figure.
>>
>>It's my firm belief that the work that the human brain does is so vastly
>>different from the work that a computer does that any attempt to compare the two
>>is an exercise in silliness.
>>
>>What really annoys me (and I'm not accusing you of doing this) is when people
>>compare the processing power of the human brain to the processing power of a
>>computer and predict when computers will become sentient and take over the world
>>(presumably) based on some crappy best-fit Excel charts. I've seen a few web
>>sites that do this, and I believe Bill Joy and Stephen Hawking have done
>>something similar recently. Makes me sick.
>>
>>Hmm, who said that studying artificial intelligence will make you believe in
>>God?
>>
>>-Tom
>
>
>
>I think it is rather well established by now that human players are, like
>computers, studying a chess tree, trying to find the best possible continuation.
Actually, not. That is the vicious influence from Kotov's teaching that
made everybody think that they should think like a tree.
There are recently two books that finally made in writing what everybody
suspected. "Improve your Chess Now" by J. Tisdall and "Secrets of Practical
Chess". Not even amateurs are taught NOT to think like a tree nowadays, and the
best book about it is "How to reasess your Chess" by J. Silman.
First quote from Tisdall's book, first Chapter ("The Fabled Tree of Analysis"):
"I do not think like a tree - do you think like a tree?" GM Anatoly Lein.
There are certain situations where a strong player think like a tree, but
their thinking should certainly not be characterized by that.
>Their way of searching this tree is probably very different of the way current
>alpha-beta algorithms do it, but still they are studying a tree.
>
>The NPS idea is based just on this fact: in a tree you have nodes, and after a
>while you have visited a number of nodes, so you can compute a "NPS".
It is possible that in an endgame, you can be staring at the position for 10
minutes and make a very strong move without calculating like a tree a single
move, based only on general considerations and _retrograde_ analysis or a goal
seeking approach. Once you find the plan, everything falls into place.
That is the reason why computers play way much worse than GMs in that stage
of the game. Computers use the wrong algorithm.
What is most powerful from humans, is that they (we) know when to calculate
briefly like a tree and when not. In other words, when we have to change the
algorithm.
Regards,
Miguel
>
>I don't think we can give an accurate measure of this NPS. We can only guess a
>vague order of magnitude.
>
>It is also possible that for every position the player visits consciously there
>are many successor positions the brain sees inconsciously "in parallel". This
>might add another order of magnitude of inaccuracy (linear, not exponential).
>
>But still, even if you allow a big margin of error for the human NPS, you are
>still 100 to 1000 times under the computers NPS.
>
>This is exactly the kind of observation that tells me that the predictions you
>are talking about are indeed stupid (and I'm sorry for Mr Hawking :).
>
>Processing power is so badly used at this time in some areas that it seems that
>no amount of it can replace a little bit of human intelligence.
>
>And human intelligence is still spoiled in such a way that we might be far away
>from seeing it successfully simulated by a computer. :)
>
>
>
> 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.