Re: facts of reality, context, possible worlds

>I've done a lot of snipping, and added a few comments prefixed with #####.
>============
>Dick McCullough
><http://rhm.cdepot.net/>knowledge := man do identify od existent done
>knowledge haspart proposition list
>
>----- Original Message -----
>From: <mailto:phayes@ai.uwf.edu>pat hayes
>To: <mailto:rhm@cdepot.net>Richard H. McCullough
>Cc: <mailto:www-rdf-interest@w3.org>RDF-Interest ; 
><mailto:latimer1@att.net>Richard S. Latimer
>Sent: Friday, December 06, 2002 8:27 AM
>Subject: re: facts of reality, context, possible worlds
>
>>1. RDF triples and facts of reality
>c. My response
>The purpose of the Semantic Web is to provide man and machine with 
>easy access to the facts of reality.
>
>I would characterize it somewhat differently. The purpose of the SW 
>is to make propositional content available to machine processing on 
>the Web. Indeed, most of the time, that content will *correspond to* 
>the facts of reality; but those facts themselves cannot be 
>accessible to machines. One cannot send a fact over an optic fiber; 
>one can only send symbols; a syntax of some kind. Now, exactly how a 
>formal syntax can specify real facts is a good question: as I said 
>above, getting that correspondence clear is what model-theoretic 
>semantics is about.
>##### Personally, I will be very disappointed if I cannot directly 
>access RDFS/OWL Classes & Properties.

Then I am afraid that you are going to be very disappointed. You will 
be able to directly access *descriptions* of classes and properties 
in OWL and RDFS, but that's the best you (or anyone else) will be 
able to do. There isn't anything very surprising about this. After 
all, it also applies in everyday English conversations. You and I 
converse by exchanging *words*, not thoughts or concepts of pieces of 
reality. Those things don't fit into email; it has to rendered into 
ASCII strings. I can't directly access the things that you talk 
about, and you can't directly access mine.

>
>>2. context
>         definition of rain
>         definition of fall
>
>Most natural-kind terms like this do not have definitions.
>##### My dictionary defines "rain" and "fall".

I doubt if it does, in fact. Most dictionary definitions are not 
definitions in the required sense, a much stronger sense than the 
everyday one. I have never seen a dictionary which defines any 
natural kind term except by referring to some other natural kind 
term. What it probably does is give a phrase which is sufficient to 
identify the corresponding concept in the mind of a competent human 
reader, but all such readers bring to bear an incredible amount of 
information which is not stated explicitly in the dictionary or even 
in an encyclopedia.

>
>>       3. possible worlds
>c. the KR alternative
>It is not necessary to pin down the actual world in full 
>detail.  You can select the appropriate level of detail, depending 
>on your purpose, and express it in KR propositions.

Then KR is exactly like all other languages. Getting this idea of 
'level of detail' pinned down precisely will involve a semantic 
theory, and if you try to construct one with a mathematical level of 
rigor you will probably find yourself rediscovering Tarski's ideas.

>>
>Sure, so KR is a syntax, as I suspected. In which case it needs a 
>semantics. Just saying that it 'corresponds to the facts' doesn't 
>cut it.
>##### "a syntax" and "corresponds to the facts" are your words, not mine.
>##### KR is an English-like language with its own syntax and semantics.
>##### The meaning of any KR proposition is the facts of reality 
>which it denotes.

That sentence doesn't make sense, I'm afraid, as it stands. For 
example, if KR has the power of English, it can presumably express 
negation. If so, then it can express a sentence of the form 'P and 
not P' for some proposition P. Now, there cannot be any facts of 
reality denoted by such a sentence: it has to be false. So on your 
criteria, it would presumably be meaningless. But it is not 
meaningless: it says that P is both true and false, which is clearly 
meaningful if P is; that is how we know it must be false.

>##### Consider the KR proposition
>#####    at view=tabula rasa {
>#####        Dick isa person;
>#####        Dick has sex=male
>#####    }
>##### I think that the meaning of this proposition is immediately obvious
>#####  to any English-speaking adult.

Well, first, it isn't to me, for example (I have no idea what you 
mean by 'tabula rasa'); but in any case, if KR is on a par with 
English, then why not just use English? KR *looks* like an attempt at 
a formal language, not just another notation for human natural 
language.

>##### Although Knowledge Explorer (KE) does not "understand" this proposition
>##### in the same sense as a human, KE can "reason" effectively using its
>##### internal entity-characteristic-proposition taxonomy.

Claiming to be able to reason effectively in a language with the 
expressive power of English is a very large claim, and one that I am 
afraid I cannot take seriously. If you could do this, you would have 
solved the central AI problem; and if you have done that, then the 
world is your oyster. You should be doing better things than talking 
to me ;-)

Pat




-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam

Received on Saturday, 7 December 2002 00:45:04 UTC