[comp.ai] footnote to understanding

gls@odyssey.ATT.COM (g.l.sicherman) (03/11/89)

Here's a provoking quotation about understanding:

	In this argument ----- commits two blunders.  He interprets
	understanding as the limit of an evolutionary process of
	baconian observation, and he treats understanding, like
	intelligence, as a fixed property independent of its
	possessor.

	To understand is to assimilate a process foreign to oneself.  A
	machine does not "understand" how to make screw eyes, because
	that is part of its function. ... When we examine -----'s
	argument closely, it reduces to two familiar ideas: the
	logical idea that all understanding rests on knowledge of the
	principles of physics, and the psychological idea that
	understanding is necessary for the sake of controlling.
	... The ideal of -----'s theory would be a computer that
	"understands" natural language well enough to be able to make
	people do its bidding.
					--Maia I. Aimless (1979)

This position is extremist.  What I find relevant in it is the idea
of understanding as assimilating a foreign process.  In this Aimless
evidently borrows from the Gestaltists (growth by assimilation).  A
foreign language is obviously a "foreign" process!

If we accept this interpretation, then a being must have experience
to understand, because to understand is to experience an analogy to
one's experience.  On the other hand, a being need not have *human*
experience to understand something.  Ants may be said to understand
aphids, and so may we; but the two understandings are different.

By the same token, if an understanding is an analogy, it must be
limited.  It is a philosophical commonplace that we shall never
know what it is like to *be* an aphid.  We can compare an aphid's
emotions to our own, but the similarity goes just so far.

Example: I tell Snerd "your fly is open," and Snerd reaches for a fly
swatter.  Snerd did not understand what I said!  He made an analogy
between language and experience, but the analogy did not succeed in
utilizing my utterance.

Example: I tell Snerd "The earth is round." He says, "Like an orange?"
I say "Yes." Snerd understands the roundness of the earth by analogy.
Snerd asks, "Can I roll the earth down the sidewalk?" The analogy is
sound in its way, but Snerd has encountered one of its limitations.
Language has the same limitations.  If you've never heard of giraffes,
you will not understand the Chinese for "There's a giraffe in the
billiard room!" no matter how well you know Chinese.

One advantage of this model is that it requires communication but not
language as such.  To a naive maiden I might explain my intentions
by pointing out the activities of her livestock.  Even without words
she could understand the analogy well enough to respond with a show
of appetite or disgust.

-:-
		"Sitting quietly, doing nothing,
		  And the Net grows by itself."
-- 
Col. G. L. Sicherman
gls@odyssey.att.COM

bwk@mbunix.mitre.org (Barry W. Kort) (03/14/89)

I was intrigued by G. L. Sicherman's citation from Maia I. Aimless.
I wonder if that name is an anagram.  It is very close to "I am aimless."

Which brings up an interesting observation.  Any system which can
communicate its goal can induce others to "do its bidding" (provided
the goal is not too repugnant).

I am reminded of the Tin Man in the Wizard of Oz who managed to induce
Dorothy to oil him.  Even a squeaky wheel seems able to communicate its
needs, as we learned from Robert Pirsig's famous book, _Zen and the
Art of Motorcycle Maintenance_.

Like, Aimless, I am not convinced that an ability to successfully signal
one's needs counts as "understanding".

--Barry Kort