RE: Unambiguous Language WAS: MEME: A.I.: Artificial Intelligence

From: Durant Schoon (durant@ilm.com)
Date: Fri Jun 22 2001 - 14:02:20 MDT


> From: "John Stick" <johnstick@worldnet.att.net>
>
> Ambiguity is more central to the general use of natural languages than
> anyone has noted yet in this thread. It is a valuable design feature, not a
> defect to be minimized at all costs. Moreover, this use of ambiguity is
> the reason why language use in open conversation requires general
> intelligence (or abductive reasoning, or a solution to the frame problem)
> and so is central to the task of building general AI.

I completely agree with you. In my post I was merely trying to answer the
Patrick's challenge of "Can an unambiguous language be created." I can now
sum up my answer as:

Yes (I think so). If ambiguity is uncertainty due to choosing between
multiple interpretations, then ambiguity can be eliminated by the author
choosing which interpretation is intended ahead of time (although
exhaustively stating all implications / assumptions will be resource
intensive and is computationally unattainable today).

Just because I currently think it's *possible*, I, in now way, advocate
that it would be desirable to speak unambiguously all the time even if
we could.

> Beyond imprecision in meaning of words and sentences, ambiguity arises
> because much of the meaning of a communication emerges from the interplay of
> words and context. This is built into natural language use at many levels.
> At the most simple are classes of words such as indexicals ( "this", "that",
> "here", "now") and pronouns ( can't tell who is "me" without context).
> Figurative speech, irony, sarcasm, humor, all rely on context (and the
> possibilty of multiple meanings that disambiguating would dissolve). For
> all words, while denotations might be disambiguated by Durant Schoon's
> Herculean Dictionary proposal, connotations will still vary relative to the
> experiences of speaker and listener. All social situations make available
> speech acts or language games where meaning is altered or supplemented:
> promising, flirting, calling out.

Yes! Reason #1 for ambiguity: Flirting!

[ ... many excellent examples snipped ... ]

> You could limit ambiguity by limiting use of figurative speech and
> references that are not fully specified. You could disclaim any intent to
> communicate connotations. But except in very specialized circumstances, why
> would you want to? The same features that yield ambiguity allow much more
> information to be expressed in limited space, and allow many subltle
> effects and meanings that cannot be communicated more simply (as in poetry,
> but not just poetry).

Yes. To vindicate myself, I did try to mention that jokes would not be as
funny. Poetry is another excellent example. And I know this is not friendly
(lowercase f), but insults should be added to our list.

> Even in scientific discourse, figurative language can be very powerful
> in suggesting new lines of thought, as with the meme meme, and the general
> use of evolutionary ideas throughout the social sciences.

Yes, thank you for your eloquent points!

--
Durant Schoon


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT