Re: computer-go: minimax and go

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

It was strongly argued many years  ago that a  brute force approach in
chess  was totally ridiculous.   Now I'm  hearing arguments that sound
completely identical to those arguments concerning computer go.

Be careful though,  a lot of this  is just semantics and gets confused
with peoples notions of   what "brute force" and  "evaluation"  means.
Your question is full of ambiguity because you say "won't work", but I
don't know what that  means.  In some  peoples  opinion nothing we  do
"works" because all go  programs  are relatively  weak.  On  the other
hand,  if  you are willing   to  settle for less,   you  could say any
approach "works" if it plays legal moves.

Todays Chess programs are  called "brute force"  but the same programs
10 years ago would have been  called highly selective.   So it's all a
matter of definition.

If you are  asking   whether you can    just construct a  full   width
searching  program that  does almost  nothing at end  nodes  in a very
simplistic way, then, just like in chess, you will have a weak program
compared to other programs.

The thing that made Chess programmers emphasize speed, was that no one
was able to produce a very strong "evaluation function", but they knew
how to make it   search farily deep, so  that's  what they did.    The
knowledge  problem has never been  solved or  even improved very much,
todays chess programs are still very stupid about chess.

I  have yet to see how  this is any different in  GO.  Knowledge in Go
seems  even harder than knowledge in  Chess, so it doesn't seem likely
that a  program that does  virtually no searching will  make up for it
with knowledge.  If  this is possible, it should  be possible in Chess
too which  is much  simpler from  the knowledge  engineering  point of
view.  Try to  write a Grandmaster chess program  that only searches a
move or two ahead and you will see what I mean.

I don't  know any arguments that a  brute force approach will or won't
"work" in go.  All I know is that either approach, at least in current
hardware doesn't make a very strong program.

Someday, a "good" Go program  may be written, that  combines a lot  of
powerful   knowledge, an excellent search  (almost  certainly a highly
selective  search) and    wonderful engineering.  This  program   will
respond to advances in power, and that is how  I define "BRUTE FORCE",
which I admit doesn't  follow the classic  definition.  But a program,
to be best,  must take advantage of  all the hardware available to it.
It's   possible that "taking  advantage"  means  being smarter and not
searching that much more.

Just my opinon, don't blast me.


Don



   Date: Tue, 7 Nov 2000 12:57:54 -0500 (EST)
   X-Authentication-Warning: osf1.gmu.edu: mabramso set sender to mabramso@xxxxxxxxxxxxxxxxx using -f
   From: myriam <mabramso@xxxxxxxxxxxxxxxxx>
   Sender: owner-computer-go@xxxxxxxxxxxxxxxxx
   Precedence: bulk
   Reply-To: computer-go@xxxxxxxxxxxxxxxxx
   Content-Type: text
   Content-Length: 346


   Hi!

   I need to know (succintly) what are the arguments that a brute-force
   minimax approach to Go won't work assuming increased computer power 
   just like it worked for Chess? 

   I know it has probably been discussed here but the search capability
   on the archives is still not working. 

   Thanks,

					       myriam