Author: Dan Newman
Date: 12:27:16 04/28/00
Go up one level in this thread
On April 28, 2000 at 10:48:14, stuart taylor wrote: >On April 28, 2000 at 08:47:05, Aaron Tay wrote: > >>On April 28, 2000 at 06:44:16, stuart taylor wrote: >> >>>On April 28, 2000 at 04:39:53, Aaron Tay wrote: >>> >>>>On April 28, 2000 at 00:35:30, Michael Neish wrote: >>>> >>>>>On April 27, 2000 at 21:58:09, stuart taylor wrote: >>>>> >>>>>>Is it possible to program a computer to set many possible future scenarios of >>>>>>favourable positions possible to arrive at with same material ballance (from >>>>>>that current position-some time later, even 40-60 ply's or more), then have >>the program find ways how to get there? >>>>> >>>>>I'm not sure I understand what you're saying. Do you mean having the >>>>>computer shuffle the pieces about on the board without removing any, >>>>>until it finds a position which is positionally superior for itself, and >>>>>then try to find a way to get there move by move? >>>> >>>> >>>>> I'm no expert, but this sounds terribly difficult to me, especially since its >opponent isn't exactly going to allow the computer to reach that position >without a fight, >>>> >>>>It seems to me that many chess books try to discourage the "I move there, you >>>>move there, then I will move there" kind of thought..Not that ,tactics arean't >>>>important. >>>> >>>>But it seems that many times,intermediate players like me are advised to do >>>>this.. >>>> >>>>1)Visualise the best squares where your pieces should go, >>>>2)form a plan based on the pawn structure, etc etc then >>>>3)try concrete moves to achieve that plan.. >>>> >>>>I think he is sugguesting that..But I doubt that such a plan if succuessfully >>>>implemented would lead to "Top class human planning skills". >> >> >> >>>I mean-with or without exchanges. If the computer could know which type of >>>"possible" position "would" be good, then it can know what to try and >calculate. >> >> >>>So the machine will be doing three things instead of the normal >>>two. Normally it calculates, and it does so based on knowledge and its >>>positional assesments. But a third thing would be to visualize plans and >counter plans by way of something which a computer could do-to estimate >possible resulting possitions, and to decide which of those positions should >be striven for. >> >>Isn;'t your third aspect already covered in a sense by the first two things? >>"Visualize plans and resulting positions?" sounds extremly general, something >>that computers are not good at yet... >> >>I suspect though, most engines already have special codes that trigger off, when >>caluculating positions that will lead to a big change in character or position >>i.e after large number of exchanges to a endgame, >> >> >>> That would be the easy part. >> >>It;s sounds hard to me.. > >The computer has to imagine future scenarios which "could" be reached. >Then it has to decide which of them would be good to reach. >Then it has to see what it can reach even against good play. >And that if the plan isn't refuted, so it will be carried out, or better. >And if it is refuted, the program must not have become worse as a result. > >I'm speaking of conjuring up positions independantly of ply count, and before >knowing exactly how it will be done. > They do that already? I don't think they do that exactly. >It's a deep concept I was thinking about it for a long time. And I want to know >if any top programmers have any opinion, or could actually do it. >The machine would be quite creative! >S.Taylor I've thought about doing something very much like this--it's very like the process I go through when I play chess... Unfortunately it's a lot easier to do what's already been done before (because it's inherently easier and because it's more or less fully described) and simply do the usual alpha/beta search. There is a short section in the book "Problem Solving and Artificial Intelligence", by Jean-Louis Laurière, that describes a program called Robin, developed by Jacques Pitrat in 1972, that did something very like this. What the program did was to come up with several plans (move sequences by one side that achieve some goal). Then it would proceed in a backwards direction from the goal identifying (and attempting to eliminate) things that the opponent might do to stop the plan. A plan might be to take some piece with your queen. But that move is blocked by another opponent piece, so the program would look for some way to drive that piece off its square, and so on. It would then add these elements to the plan until it got back to the root. (So, as you can see, every idea that you can come up with has likely already been thought of by someone else back in the dark ages :). I do think (given the improvement in programming tools and so forth) that it's an idea that ought to be explored though...) -Dan.
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.