Re: [computer-go] Pattern matching - example play

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 20:43 1-12-2004 +0100, Frank de Groot wrote:
From: "Chris Fant" <chrisfant@xxxxxxxxxxxxxxxxx>
>Subject: Re: [computer-go] Pattern matching - example play
>
>
>> > I think human Go pro's play Go very sub-optimally so learning from pro
>games
>> > will never produce anything "better" than a Go pro, *but it will keep
>pace*
>> > with pro patterns, for example.
>>
>> I disagree with this.  The limiting factor in Go skill if the bad
>> moves that one makes.  As many of the pros who are generating your
>> learning games are not making the same mistakes, their mistakes will
>> disappear from the system's memory due to the law of large numbers.
>
>
>
>Sure, I realize that.
>But there are so many factors that make a human play better than a compuyer
>program, that I thought this would outweigh it, an result in something that
>at best would be slightly better than a pro.

The only advantage human has is that he can play a move which in the far
future is good too.

In all other respects a human has a disadvantage.

Let's start with the first one. 

A humans brain works, to say it very rudely (as it's not really like that)
at around 10-12Hz.

That means that a human can *see* the best move, but he can't see more than
10 moves a second ever in his life. Add complex considerations and the
average pro will basically recognize patterns and just *feel* that the
tactical outcome of the dogfight is a win. He doesn't need to calculate it
like the computer.

Yet the computer *can* calculate where human cannot.

Secondly go, it's just a game. It has a limitation of 19x19 and the rules
are clear. In short a computer *can* play it.

Now the big problem. Humans make mistakes. Of course amateurs more than pro's.

But also pro's make horrible mistake, seen from their level.

As we all know the weakest moves you make lose the game.

Additionally when a pro of the future will play the computer, the pro
probably will not be motivated to win from the machine. Either becuase he
slept well, because he gets paid the money anyway and next week is an
important GM tournament, because he is not used to such positional horrible
opponents, and sometimes because he has put millions of dollars at his own
via via at some betting sites. The last few reasons have been a major
problem in computerchess and will come in go too. 

And *those* problems will let the first few pro's lose from go programs
long before anyone here guesses the go programs even remotely have a chance
against mankind. Most likely that wil happen at a time when i still win
from all go-programs.

Human players, myself included, tend to overestimate the strength of
mankind when looking from worst case viewpoint.

The worst case of a human really is way more horrible than you can imagine.

And boardgames like go just are about the worst case you show in a game.
Nothing else.


>_______________________________________________
>computer-go mailing list
>computer-go@xxxxxxxxxxxxxxxxx
>http://www.computer-go.org/mailman/listinfo/computer-go/
>
>
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/