COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	1. As to the intemperate language, it was Searle who started it.
C00007 00003	The argument from cryptography and information theory
C00008 ENDMK
C⊗;
1. As to the intemperate language, it was Searle who started it.
I deny being abusive.

2. Re: Chinese room.  Searle is mistaken that the symbols can
be assigned arbitrary interpretations.  If you assign, for
examples, the Chinese characters in their standard
order by counting strokes in the radical, etc. alphabetically
 successive English words
as interpretations, the interpreted dialog will obey no compactly
statable syntactic rules.  Cryptanalysis depends on this fact.
So does a child learning a language or a person learning a
language without a text, bilingual teacher or dictionary.  In
the language learning case, it is the combination of language
and behavior that turns out to have a substantially unique
interpretation.  Simple substitution ciphers in alphabetic
languages have unique interpretations at around 20 letters.
No-one has, to my knowledge, attempted to decrypt Chinese
cold, but cryptanalsysts have deciphered straightforward code books.
The deciphering is quite hard and other clues are used when
available but there is no problem verifying when one has
a solution.  It decodes additional material.

3. Searle paraphrases me as saying that semantics can be assigned
to programs arbitrarily.  If they are to be executable, there
are structural restrictions.  It's the cryptography argument
all over again.

4. The common sense mental notions are ascriptions of structure inferred
from behavior.  They don't depend on knowing about neurons.  Children
develop the concepts of belief, desire and intention that they ascribe
to other people and animals, because it explains their behavior.
They learn analogies between their self-observations and what
they observe about others and what others say.  This structure
turns out not to be entirely autonomous in humans, because closer
observation  turns up interactions with internal chemistry, etc.
However, the common sense mental notions turn out to have rather
close correspondents that seem to be useful in designing
machines that behave effectively in a world requiring
intelligence.  We then take the step Searle objects to --- treating
intelligence abstractly and ignoring certain aspects of human
intelligence, e.g. the effects of cocaine.  We are betting that
abstract theories of  abstract intelligence will prove useful
in AI.  Not only is this confirmed, but psychologists often
find the abstractions useful in treating human intelligence,
e.g. they often use theoretical that have no place for
incorporating the effects of cocaine.  On other occasions
they use theories that allow for cocaine.
The argument from cryptography and information theory