Re: computer-go: Pattern matching

[Tristan Cazenave <cazenave@xxxxxxxxxxxxxxxxx>]

> * Strongly doubt AI will work for Go programming. :)
>

Is AI a boring academic field with people publishing

a lot of papers about useless theories, or a domain

where people try to enable programs to behave as

'intelligent' as they can ?

I agree that when you look at what is published, AI

often looks like the first. But if a good computer Go

program is not called AI, then AI is sick of its theories.

I prefer to see AI as whatever you can put in a program

to make it do things in an intelligent way.

A better formalization of Go programs would enable to

make progress in Go, but a better understanding of

what is involved in learning Go would be even more

interesting. You cannot assume it will work, but it

is at least a promising approach. It may take time

to make it work, more time than traditional approaches

to Go programming, but once it works, the gains are

enormous. It is not even sure that it takes more time,

a program like NeuroGo for example relies on some

good ideas on Go and TD learning, it has not taken

much time to be developed and it is a good Go program

compared to many others. If you have some good ideas

on how to make a program learn Go, then it may be more

fruitful than if you try a standard approach.

--
Tristan Cazenave
Universite Paris 8, Departement Informatique, Labo IA
2, rue de la Liberte 93526 Saint-Denis Cedex - France
http://www.ai.univ-paris8.fr/~cazenave/