[ Ray Allis 930128.1530 ]
[From Bill Powers (930128.1230)]
I don't see how there can be any rational account of language
that doesn't include an account of the properties of the
experienced world.
Exactly. I was trying to understand why AI is such a bust ...
···
================================================================
CHAPTER 19 Language (Ray Allis, unpublished)
19.1 Definition of Language
The languages here under discussion are human "natural languages".
These include all human languages which have developed "naturally",
regardless of whether they use sounds or gestures.
Certainly there is a "deep structure" in linguistic ability; it is our
physical implementation (embodiment). The "meaning" of language is
that state which it evokes in us.
19.2 Words evoke experience.
Words, spoken, written or gestured, are symbols: they symbolize
experiential knowledge in humans. Meaning is not in the word; words
have an effect on a human mind - through our eyes and ears, and meaning
is the effects which are associated with the effect of the word. This,
so far, can only occur in a human mind; machines are not yet complex
enough to have effects to be associated. It is not entirely a matter
of complexity; it is experience that machines lack. Words are
associated with sensory experience first; hot, cold, soft, red, doggy.
Later words are associated with more abstract experience; love, fear,
justice.
19.3 Words have no intrinsic meaning.
A poet selects words for their connotations. Our reaction to a poet's
work clearly depends on the richness of experience we bring to the
interpretation, and not on any intrinsic meanings of the words. In
fact, as I shall assert later, words have no intrinsic meanings. Our
"common sense" provides us with an interpretation of the symbols based
on our understanding of the world. An example: "The fog comes in on
little cat feet" [Sandberg]. A very small child might at first try to
understand that statement literally, but people with more experience
realize that poems have a different purpose or intent than other types
of writing, such as instruction manuals. After a person has felt the
softness of a kitten's paw, and seen its delicacy and heard the silence
of its motion, those experiences and impressions can be evoked by
associated symbols; by words. We know that fog doesn't have feet. But
we can see the similarity, the analogy, between silent, soft kittens
and silent, soft fog, and that is what the poet wanted.
[ . . . ]
The major frustration in the present communication is caused by the
fact that the words do not have exactly identical meanings for each of
us. Our communication depends on our personal experiences in childhood
and school having given us sufficiently similar concepts associated
with the symbols we are using. And in fact, much time and effort (and
frustration) is necessary to build our beliefs that we are indeed in
agreement as to definition of the words we use. Only then can we
consider the concepts we are trying to communicate.
The way this works is that you send out an assemblage of words you
think might convey your concept and you assess my reply in terms of
whether you believe I "understood". I, in turn, choose words and
structure them in a way which, I hope, conveys my intent. You and I
judge the effectiveness of our communication by observation of each
other's behavior. We negotiate, we cut and fit, we restate.
If symbols (words) had meaning outside our minds, we would not need to
do any of this. There might be problems with our perception or
acquisition of such meaning, but we could believe given the identical
perception, we must have the same meaning.
[ . . . ]
19.4 Not a symbol system.
Because human languages consist of symbols, it is easy to confuse
natural languages with artificial symbol systems. But they are two
very different things. Language is not useable as a set of symbols
through manipulation of the relationships among words in anything like
the same way that arithmetic is. Human natural languages do consist of
symbols, but the "operators" and "rules" for use make no pretense to
logical validity or completeness. Human natural language is not a
symbol system; there are no useful interactions among the symbols
themselves. Words do not interact with each other. Their
connotations, which are evoked responses in human minds, interact with
each other.
19.5 The Utility of Language: Communication With Other Minds
Of course, if our concepts and those of other minds do not overlap, we
cannot communicate. If they "see" in different frequencies we cannot
talk of "red" meaning danger, or appreciate a field of flowers, or
wonder at the beauty and mystery of our lady's blush. They won't
notice that we feel pale and ill, representative painting will be
wasted on them. We, on the other hand, cannot appreciate the night
sky at radio frequencies, or the hunting cries of bats or the shades
of intensity a flowing electrical current causes in a high voltage
transmission line.
19.6 Communicating Experience
Actually, experience is not communicated. Experience is _evoked_
through transmission of symbols which evoke internal (non-symbolic)
state to much the same internal effect as if it were evoked directly by
environmental sensors. Language understanding is exactly this evocation
of state. In other words, if the language evokes in you experiences
the speaker intended to evoke, then you understood the language.
The ability to construct internal imaginary events and situations is
fundamental to symbolic communication; where symbols are derived from
and in turn evoke internal state. Computer systems cannot understand
language until they possess the analog, sensory, experiential states
symbolized by linguistic elements.
=====================================================================
I'm still working on comments about Holland's et. al. book,
"Induction", but I see no reason to change the opinion I had at the
time it was published; they are talking about deduction.
From the Fringe,
Ray Allis
ray@atc.boeing.com