From: Stevan Harnad (harnad@ecs.soton.ac.uk)
Date: Fri Sep 20 2002 - 19:51:27 BST
On Fri, 20 Sep 2002, Srinandan Dasmahapatra wrote:
> On Thu, 19 Sep 2002, Stevan Harnad wrote:
>
> > We will be talking in circles until you twig on the difference between
> > (intrinsically) meaning
> > X (as when I think and mean X) and merely being systematically
> > interpretable as meaning X (by thinkers who can think and mean X):
> > extrinsic meaning (e.g., in blueprints, simulations, text, symbols).
>
> It's the interpretive filter that I cannot get rid of. I agree with you
> that ultimately we have to fall back on the brain and brain activity as
> result of complicated stuff and goings on in response to externally
> generated (snow) or internally generated stimuli (peekaboo unicorns). The
> only way of mediating the relationships between the two is via the dual
> aspect nature of natural language -- I guess your term hybrid for natural
> language syntax and grounded robots smacks of this dual aspectness --
I think most of the burden of grounding will be taken on by sensorimotor
category learning and manipulation, and language only secondarily. It is
the lexicon of language that will inherit the senorimotor categories as
grounding.
> which can even convince us that when we point to a white peekaboo unicorn
> in the middle of the Antarctic and refer to it via a mass noun ("that's a
> lot of peekaboo unicorn") we can use the syntactic (in the distributional
> sense) cues to guess that it is the white substance that has the name
> peekaboo unicorn.
I only mention such exoticisms so people stop worrying about them, and
focus on the world of horses, stripes and zebras. The peekabook unicorns
will fall out as a bonus. Ignore them for now.
> Your symbol grounding is a catchword that encapsulates
> this internal/external aspect of meaning. Fair enough. Once you are in
> the internal domain, there are no boundaries that set apart what goes on
> in the mind/brain just because that's how things are with the mind/brain.
It's not internal vs. external "domains." Input is input, and most of it
is sensorimotor projections of objects in the world. The internal stuff
is mostly sorting and processing of that, and then some spin-offs from
the transaction.
> We have strong internal "feelings" about "percepts" that are
> stimuli/responses/stimuli that are "real."
I think you are trying to force something there. We have feelings,
period. That some are strong and some weak, some "about" objects, some
simply felt is irrelevant, just a distraction. Feelings are most
certainly real, and what is meaningful is felt to be menaingful. But
that's all there is to it. The rest is just grounding and
interpretatability.
> Hence, by transitive extension
> of all these relations, meanings are feelings that the world and the way a
> community of speakers represent the world evoke in us.
Not at all. Meanings are the feelings going on in the heads of the
entities that are understanding or generating the meaning. Simply put,
they are the feelings accompanying the understanding and productioin of
words and sentences.
> It's a truism that
> in talking about this we are choosing specific representations. It is how
> these representations get mangled in usage that suggest traces of reality
> -- of the world as lived in and expressed via a language-using community.
I don't know why the community keeps getting dragged into it. There is
one. It's there. But it's hardly relevant.
> If that is your position, I do not have much quarrel with it.
I couldn't infer a position from what you described above. My positions,
to the extent I have them, are far, far simpler than any of that:
We can understand and generate meaningful words and sentences. The words
and sentences have extrinsic meaning (i.e., they can be interpreted, by
us, as meaning X), but in our heads, the thought that X cannot be the
same as when X is written down on a piece of paper. On a paper, an
reader reads the sentence, and its meaning comes to life in his head.
But no one else is reading the sentences in my head. So they must be
connected to the things they mean in some other way than via the
mediation of an external interpreter. An autonomous robot able to
manipulate objects and words the same way I do sounds like a good
candidate for the kind of system that makes that connection.
That's all.
> > Just as applying the right algorithm will compute whether P is prime. So
> > what? The menaing still isn't in the symbols but in the head. If "29" is
> > already "semantic arithmetic" then I rest my case (the notion of
> > semantic arithmetic is empty, and the semantic is just redundant). If it
> > only becomes "semantic arithmetic" if we add some "context" and some
> > algorithms, then I remind you that the result is every bit as
> > un-semantic as "29" alone (without the help of the thinker/user), just
> > more complex.
>
> There is no arithmetic without a thinker (for me, not Platonists). So, no
> worries there in the immediate sense.
But there are arithmetic symbols. Why don't we call them "semantic
arithmetic."
> > "Context" will not help, in any way. (It all reduces to I/O plus syntax,
> > as long as we are just talking about ungrounded computation.)
>
> Context is used in a way that roughly means "for every possible
> situation." Any realist (of any sort) will interpret that to mean
> "connected to a spatio-temporal universe." Metaphysical rumblings appear
> when the "situation" will not be rid of how we as humans have decided to
> influence it. And when counterfactuals seem to impinge on what is. But I
> don't think I ought to go into it. What sparked off the discussion was
> the Semantic Web, and as you say:
Why all this metaphysics? Let's just talk about symbols and people and
meaning, ungrounded and grounded, extrinsic and intrinsic.
> > I think we agree then: There is no difference between the "semantic web"
> > and the non-semantic web. It's all just syntax, meaningfully
> > interpretable by people. Same is true of "semantic books," etc. (i.e.,
> > all books).
>
> One of the objections I had was in the use of the term ontologies and also
> the notion of meaning as defined via truth which I choose to interpret as
> truth in a model. At this moment it is convenient for me to leave model
> suitably imprecisely characterised. Or else we start the same round of
> back-and-forthing.
I think we are still talking about different things -- you about various
truth theories, maybe even different ontologies, I simply about what it
is in the head that makes words meaningful to people (but not to books
or computers).
> > So much for the notion that this is all just a self-reinforcing symbolic
> > web...
>
> > At bottom, it's not just symbol systems that are systematically
> > resistant to misconstrual, nor even symbol-system users' shared
> > conventions that are resistant. At bottom, and most importantly,
> > it is the (mushroom) world that the symbols are about that resists --
> > but only in a grounded system.
>
> That does not make the detailed principles of grounding any more specific
> than a functional specification, though. That is surely a partly
> empirical issue that can't be settled by exchanging e-mails. I guess
> that's what you dedicate the rest of your life trying to do!
Of course successful grounding and TT-passing is an empical project!
> > > > I believe-I-know what "X" means when it feels as if I know what X means,
> > > > when it seems I can pick out X's and non-X's, and when neither the world
> > > > of objects nor other speakers error-corrects me.
> > >
> > > Despite your disinclination to engage with Tarski or Davidson, "formal
> > > semantics" does not say much more than that!
> >
> > I doubt T or D say what I say about feelings, otherwise they too would be
> > getting the flak I'm getting for it!
>
> There are formal constructions that hide their feelings -- they
> are more visible in intensional semantics, etc. in a multi-"agent" setting
> routinely invoke infintary arguments to establish common ground. That's
> usually a sign that formal representations have to engage with a
> meta-representation. All this is some formal intuition on my part, so
> please feel free to ignore this.
>
> For me, we do not have access to language except through a linguistic
> community. Stuff in the brain magically enables that to happen, creating
> languages out of nothing but interactions between speakers and the
> experience of the world. I think an expression often used is "brain in a
> vat" -- very obviously we are not. We carry around representations of our
> body and the way the body interfaces with the world as well. I don't
> really know whether all of this is necessary for meaning, but I don't know
> where to chop and separate.
Well, it was fun doing the copping and separation of this exchange,
anyway!
Cheers, Stevan
This archive was generated by hypermail 2.1.4 : Mon Nov 04 2002 - 22:08:43 GMT