RE: [ontolog-forum] Foundation Ontology [was Semantic Web shortcomings]
John,
These are good points -- one way to look at this is that the concept of
"grounded" is itself not well-grounded... or that the concept of grounded is
often used too absolutely... we should acknowledge that many symbols are
partially grounded / semi-grounded / semantically grounded... that often
refer to things we only partially understand, or cannot directly experience,
and often refer to things we understand somewhat abstractly...
So perhaps the acronym could be: SGFO - Semantically Grounded Foundation
Ontology ... which people might pronounce "sigfo"...
Phil Jackson
> -----Original Message-----
> From: standard-upper-ontology@LISTSERV.IEEE.ORG [mailto:standard-upper-
> ontology@LISTSERV.IEEE.ORG] On Behalf Of John F. Sowa
> Sent: Sunday, August 24, 2008 9:39 AM
> To: standard-upper-ontology@LISTSERV.IEEE.ORG
> Cc: [ontolog-forum]
> Subject: Re: [ontolog-forum] Foundation Ontology [was Semantic Web
> shortcomings]
>
> Folks,
>
> I suggested the term UFO as a joke, but we should recognize that the
> problem of symbol grounding is significant. And we should note that
> there are three methods of grounding the symbols we use:
>
> 1. Direct experience with the referents by perception and action.
>
> 2. Indirect connections to experience by associations created by
> patterns of words that are more directly grounded.
>
> 3. Communication by means of natural languages with other people
> whose grounding for the symbols is more direct than ours.
>
> The issue of grounding the symbols used in AI was a hot topic
> back in the 1980s with lots of spilled ink about Searle's
> Chinese room. The people who are working on robots are finally
> answering the symbol grounding questions by connecting symbols
> to the robots' systems of perception and action.
>
> In any case, I was never really worried about symbol grounding
> because people have been using ungrounded books for millennia.
> Some books link the words to pictures, but the main method for
> grounding words is to take advantage of the human methods of
> communication (point #3 above).
>
> For example, very few of us have any direct experience with the
> experiments that established the existence of atoms and bacteria,
> but we use those words all the time. We also have no direct
> knowledge of things inside our own bodies, such as bones, heart,
> liver, brain, etc. But we accept the information that people
> who have more direct experience with such things have told us.
>
> We can also talk about cities and countries we have never visited.
> When we actually go there, we discover new things we hadn't expected,
> but we also find that our most important expectations are fulfilled:
> there are people at those locations who can provide food and shelter
> and the means for us to get back home again.
>
> Therefore, I consider symbol grounding a solved, or at least solvable
> problem. But I suggest that we keep it in mind, because we must
> ensure that any new symbols are connected to ones that are grounded.
> Peirce noted that every proposition stated in any language, natural
> or artificial, must have at least one index to link it to its referents.
> That index could be outside the statement itself, such as a pointing
> finger. But more often, it's a name whose referent is determined by
> other propositions.
>
> And I would also note that the *most* frequent method of symbol
> grounding that people use is based on communication with other people
> by means of natural language. That is why comments are essential for
> telling us how to use any formal notation. The ideal comments are
> ones that can be stated in a controlled natural language that a
> computer can relate to the formalism automatically.
>
> John Sowa