A quibble: Go is already solved, but not when the board is empty!
It may sound stupid and obvious but I think its a good starting point.
Even between two 20kyu players, when the board is solved and there
is only a 1-cell gap between the walls and the border, the last 4 moves
threaten, block, connect, defend *are* perfect moves. When, as we
say here "all the fish has been sold" there are only the last 2 points
left.
After that, there is nothing left. I am not good enough to know how
many perfect moves can a pro player play at the end of a game, but
is would not surprise me if he/she can play 20 moves or more.
Assuming a pro player can play 25 perfect moves at the end of a game
(or does not even play them, because he expects the same from
his opponent) that is still an overwhelming complexity for a computer.
Where is the difference?
IMO: Because humans intelligently "connect" known optimal local
answers in the correct sequence. And computers have two options:
a. Global search. The easiest way, but when search space is too
big, any search technique fails or returns something too weak
(with time restrictions). This gets even worse because much
computing is required to determine if a game is finished.
b. Humanlike thinking. Trying to "link" perfect local answers. Which
today alternates the most "intelligent" with the most stupid behavior
one can imagine. And games penalize stupid moves too much.
Most of the objections I have read in this thread apply to "a" but
it is not clear why they should apply to "b".
If "b" is ready in Terry McIntyre's lifetime, computers should "solve"
the last 40 moves of a game, if pro humans "solve" 25.
Jacques.
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
