Please see notes and comments interspersed below,
Sincerely,
Rich
Cooper,
Rich Cooper,
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F Sowa
Sent: Sunday, May 24, 2015 10:09 AM
To: ontolog-forum@xxxxxxxxxxxxxxxx
Subject: Re: [ontolog-forum] Semantics of Natural Languages
Tom and Rich,
I certainly agree with that point:
TJ
> the internalist/externalist argument among cognitive
scientists (the
> "brain in a vat" discussions) are
orthogonal to the
> connectionist/representationalist argument.
But I don't believe that induction is a riddle:
TJ
> a level at which we can begin to understand how we
could solve such
> representationalist issues as the riddle of
induction.
RC:> Here is another relevant quote from PG:
However, it became apparent that
the methodology of the positivists led to serious problems in relation to the
problem of induction. The most famous ones are Hempel’s16 “paradox
of confirmation” and Goodman’s17 “riddle of induction.”
What I see as the root of the troublesome cases is that if we use logical
relations alone to determine which inductions are valid, the fact that all
predicates are treated on a par induces symmetries which are not preserved by
our understanding of the inductions: “Raven” is treated on a par
with “nonraven,” “green” with “grue” etc.
What we need is a non-logical way of distinguishing those predicates that may
be used in inductive inferences from those that may not.
So he sees something which he believes induction misses,
in his view.
But just taking the background of "Raven" as
the complement of Raven is meaningless, IMHO. It would be better to use the
class of symbols that are known to distinguish an object (not a background)
from other objects (not other backgrounds) called ~Raven, which could include
all other objects in the domain which are, literally, not Raven objects.
JFS:> The people who think
that induction is a riddle are so enamored with deduction that they think that
all reasoning should be algorithmic.
But deduction cannot generate
anything new. It can only work out the details of what we had already
discovered by induction, assumed by abduction, tested by trial and error -- and
repeat, and repeat...
For more about that cycle, skip
to slides 41 to 46 of
http://www.jfsowa.com/talks/micai.pdf
Why has AI failed? And how
can it succeed?
I posted micai.pdf earlier, but
I recently added more slides at the end. That cycle, which is based on
Peirce's "logic of pragmatism", captures the essence of all reasoning
-- from the so called "commonsense" to the most advanced science.
For the more advanced issues,
see the talk I presented in April:
http://www.jfsowa.com/talks/ppe.pdf
Peirce, Polya, and Euclid:
Integrating Logic,
Heuristics, and Geometry
These slides show the
relationship of induction, abduction, deduction, and analogy to perception.
TJ
> [Gärdenfors's] work (in
Chapter 7 of Conceptual Spaces) makes it
> something more than a
metaphor. But until its connection (pardon the
> pun) to the ANN paradigm is
better articulated, it remains a
> mediational layer between
representationalism and connectionism whose "connection"
> to connectionism remains
unestablished.
My major complaint about that
work is the implication that there are only three major paradigms. There are
*infinitely many* paradigms.
I like to quote the slogan of
the psychologists: "Beware of anybody who has a one-factor theory."
An article "The Amazing
Teen Brain" in the June issue of _Scientific American_ summarizes the
issues. It's only 5 pages long, but it shows the complexity. The three models
that PG discusses each have a small grain of truth, a larger amount of falsity,
and many more unknowns.
RC
> I took [PG's definition of
the symbolic approach] to mean
> mathematically modeled by
Turing machines, but NOBODY except
> mathematicians and
philosophers with lots of time on their hands
> actually programs TMs.
PG was using TMs as a generic
representative of anything running on any digital computer. In his examples,
he cited Fodor and other advocates of symbolic methods in AI. But that ignores
the huge differences ranging from Roger Schank to Richard Montague.
In any case, there are many
related issues in the fruit-fly thread.
I'll reply to them later.
John