Clarification

[Matt Gokey <mgokey@xxxxxxxxxx>]

David Mechner writes:

>I don't think enough effort was put into that line of research to say
>that it failed; it was less successful (in terms of playing strength)
>than work on search, and was abandoned.

Ok, "failed" is too strong a word, and of course given that there
existed a better approach its natural that the path of least resistance
was followed.   Point taken.  However, my point was a hypothetical one,
and I still believe that a chess program of this nature would very often
make errors in judgment like Go programs do today.  Modeling knowledge
does not give a program intelligence or understanding...only the
appearance of it, until it blunders.

> How can one  predict future positional characteristics accurately
using only the
> current position?  I don't think it's possible.?  Anyone disagree?

>>I disagree strongly, and there's a simple proof: the existence of
>>human experts. We predict future positional characteristics accurately

>>using only the current position. Of course human experts do read
>>things out, but not in any sort of exhaustive way that can be compared

>>to the search done by chess programs. Human players' ability to play
>>speed-go demonstrates, I think, that a great deal of our skill likes
>>outside of our ability to search.

Exactly, only human experts exhibit this kind of ability.  Of course its
not perfect either.  Chess masters do make mistakes and so do go
players.  Blitz chess games contain many more errors, and I assume
speed-go does as well.  Anyway, my point was that only human
intelligence can analyze positions this way, and that our AI, GA, or
other machine learning techniques are so inadequate compared to the
brain, that I don't believe its the answer.  My comment about starting
from simple and building complexity step by step was referring to the
artificial life field of research that is starting simple, trying to get
emergent behavior from complexity formed by interaction of simple rules,
etc... A whole different field, not related to go research at all and
many many years away from exhibiting intelligence.

I'm no expert in cognitive science, but If I were to characterize human
play I would have to question whether we really know how the brain is
dealing with a problem domain like go.
Its surely a combination of vast knowledge from study and experience,
pattern recognition, intuition, analysis of risk and dependencies
between local sub games, reading out lines, and probably a host of other
things we are not aware of and that the practitioner may not even be
aware of either.  Vastly complex - very fuzzy.  I still tend to believe
that its far too complex to model explicitly and too complex to develop
using today's AI techniques.   And if I'm wrong, well I know its too
complex for me, so my research will take me elsewhere.  However, I do
believe that many current techniques being employed are quite useful for
certain sub problems needed to play go, like life & death, eye space,
territory assessment, etc.

> Our concentration has been on life and death analysis and we expect,
> by the end of the summer, to have results showing that programs can do

> quite good general life and death analysis by explicitly representing
> the knowledge and *logic* that human experts use when reading out life

> and death problems.

Sounds interesting and would be a great help to current programs I
expect.  Will you have any published papers from this research
available?

>I'd advise restricting the scope of your research - pick a limited
>task to start; perhaps move ordering, low-liberty tactics, territory
>assessment, eye assessment, group strength assessment, pattern
>learning, whatever; and try to use whatever your favorite technique is
>to accomplish that.

I'm definitely going to limit my scope; I'm thinking along the lines of
advanced
searching techniques that may be able to exploit certain properties of
go.  I'll post more about it later, because I'd really like to get the
opinions of this list's regulars before committing a lot of time. :-)

Matt