I meant the GOFAI approach of symbol manipulation along the lines of Newell
and Simon, more sophisticated today, but basically using high-level symbols
like natural-language words and relationships between them to model the
world or the domain that an AI system is dealing with. This seems to still
be the belief of some researchers in AI, and some of them receive good
funding for it.
In opposition to that, is the concept of knowledge representation as a
fluid, non-discrete model (implemented in a computer obviously using
discrete bits, bytes, and operations) of the behavior of phenomena in the
world including an AI system's interaction with it.
I agree that if the universe could be considered a formal system (which may
or may not be the case, and which the common interpretation of quantum
theory says is not the case), and that is what is meant by symbol system
then "YES."
The primary intelligence I refer to IS supposed to be something that "can be
reduced to brain activity and physical law," and CAN "be expressed as a
nonlinear dynamical system or the effects thereof."
So, in the way that you've described this, I totally agree with you. I
guess I was attacking a paper tiger that any real thinking person involved
in AI doesn't bother with anymore.
Ben wrote:
Subject: Re: [agi] A question on the symbol-system hypothesis
My question is: am I wrong that there are still people out there
that
buy the symbol-system hypothesis? including the idea that a system based
on
the mechanical manipulation of statements in logic, without a foundation
of
primary intelligence to support it, can produce thought?
[John Scanlon]
The problem with answering your question is that I don't really know
what you mean, exactly, by "symbol manipulation"
Do you just mean the application of the rules of some formal axiom
system, like a formal grammar?
In that case, the answer becomes probably YES, because arguably the
whole physical universe can be effectively modeled as a "symbol
system" (e.g. if you accept it can be modeled as a quantum computer).
And, setting aside physical reductionism, "Symbolic dynamics" in
dynamical systems theory shows that any complex dynamical system can
be effectively modeled as a symbol system. This yields a YES answer
as well -- the mind can effectively be modeled as a symbol system.
(You refer to "primary intelligence" -- but is this primary
intelligence you refer to supposed to be something that can't be
reduced to brain activity nor physical law, and can't be expressed as
a nonlinear dynamical system or the effects thereof? Because physics,
brain and complex dynamics can be expressed in terms of symbol
systems!!!)
IMO, the real question then comes down NOT to whether the mind can be
effectively modeled as a symbol system (in its structures and
dynamics), but rather as to what kind of a symbol system the mind
is...
Clearly, it is NOT a just a symbol system involving symbols denoting
concepts on the level of abstraction of words in language....
So, if by "symbol system" you mean "symbol system involving symbols
denoting concepts on the level of abstraction of words in language" --
then no, the mind cannot effectively be modeled as a symbol system.
To pervert Fodor's terms, there **is** a formal mathematical language
of human thought, but its "words" and "grammar rules" are not remotely
restricted to being similar to the words and rules of human
language.... It is a big, ugly, complex language whose terms and
rules have as much to do with perception, action and
ineffable-to-consciousness intuition as with explicit linguistic
stuff..
-- Ben G
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
