Re: An AI program that doesn't learn

["P.J.Leonard" <P.J.Leonard@xxxxxxxxxx>]

I wrote:
> >  Hmmm . . . humans use plans. A plan persists into the future. A plan
> > requires
> > a context. The context comprises a description of the current objects
> > and their
> > state. You can judge the success of a plan by monitoring the change of
> > state of
> > the objects.  If your plan was to kill a group and it fails then you
> > should really
> > do some detailed analysis to identify the flaw in your plan. e.g.
> > oponent does a move
> > that you did not consider.

Peter.Smith@xxxxxxxxxxxxxxxxx wrote:

> If the program is being tuned by hand then analysing flaws, and re-tuning,
> is also being done by hand rather than by the program.

 For a program to decide stratergies it must search.
To search you must focus. To focus you need to identify objects that
are the subject and watch for changes of state.  In my opinion a GO
program that does not try to understand the evolution of the GO game
is at a serious disadvantage.

> If the program is tuning itself against a set of problems then again it's
> not interesting for the program to keep history information for a board
> position (it won't have any when it is being trained.)

 Why not ?  It is very interesting for a program to know how it got 
into a particular state. If a program losses then it should perform a
backward
analysis to find the point at which it lost. Then it can present the
problem
to the programmer together with it's own (incorrect) reasoning.   

 cheers Paul.