msellers@mentor.com (Mike Sellers) (01/09/90)
The following occurred to me while reading Searle's article in Sci Am.
Comments appreciated.
Assumption: humans do (at times) understand, think, and consciously
apprehend, whatever those words specifically mean; they have enough
common-sense content to let them stand on their own for now. The issue
at hand then seems to be whether computers in some possible configuration
can reproduce this same process or effect, especially within some practical
bounds of space and time. (I suppose the reason for asking this question
is, first of all, simply that we are continually fascinated by our own
magnificance. Secondly, if we can determine the answer to this question,
we may be better able to focus our efforts in trying to create other
intelligences like ourselves --the hard way, I mean. :-) )
If syntax can yield semantics, then it may be that the person in the
Chinese Room will, in fact, gain an understanding of the messages being
passed in and out after some manipulation: He will have learned the
meaning of the input from the process he goes through to create the output.
If this is the case, then the whole argument will have succeeded in showing
that a sufficiently complex symbol processor can attain understanding.
Or at least that a sufficiently complex symbol processor can appear to
have attained understanding, which in reality may be the same (how do I
know when my daughter has understood me? How about my dog? Or you?).
However, if syntax cannot yield semantics, then it would seem that the
person within the Room is destined to simply be a non-conscious symbol
pusher forever. In this case, let us suppose that the occupant is further
reduced to receiving the simplest of symbols and putting other simple ones
out after following some rules laid out in a rather thin rule-book. From
time to time the rule-book calls for a null response, and from time to
time it calls for the transcriptionist to add a new page to the book
containing new instructions constructed from an explanation in the book
itself. Usually though, he simply creates a new message based on a
manipulation of the old according to the book.
In addition, our symbol pusher is not dealing with anything so intricate
as Chinese, but is instead getting and producing a limited form of
Babylonian cuneiform. Unknown to our Babylonian transcriptionist,
there also exist some number of other Babylonian Booths, each receiving
and transmitting their own limited messages in a rather undirected fashion.
Together these comprise a Bablylonian Bureaucracy where each Booth receives
some simple input, processes it a bit, and supplies some output to some
other Booth. Each Booth receives symbols from some number of others, and
supplies input to still others, though the inhabitant of each need not be
aware of this fact.
This whole improbable set up may seem to be begging the question of the
original Chinese Room symbol manipulation scenario (though set up in the
large), but it contains an important difference: Each Booth is blind to
the functioning of the others, and knows only about its own activity.
Individually they are rather primitive, if flexible, cuneiform processors,
and possess none of the attributes we are interested in. However, taken
together, we must conclude that (through some process we do not as yet
understand (!)) the whole mass of them together will eventually --given
the necessary and sufficient environmental input-- take on awareness,
understanding, and cognizance. This must be the case, since the only
existing example of a thing possessing these qualities aquires and
maintains them by just such a system.
We each possess some number of relatively simple, though highly
communicative processors that, we assume, do not themselves possess any
of "our" faculties. The Babylonian Bureaucracy does not begin to explain
how such a system can generate consciousness, and yet such generation does
apparently occur. It does however extend the possibility that by starting
with an appropriate mass of processors --natural OR artificial-- arrayed
in some as yet undetermined but highly connected fashion, and supplied
with continual and coherent environmental input, we will eventually end
up with something that could in fact "be" a dog, a child, or a computer.
Again, comments would be appreciated.
--
Mike Sellers ...!tektronix!sequent!mntgfx!msellers
Mentor Graphics Corp. msellers@mntgfx.MENTOR.COM
Electronic Packaging and Analysis Division -- AutoSurface Project
"Amor est magis cognitivus quam cognitio"gilham@csl.sri.com (Fred Gilham) (01/10/90)
msellers@mentor.com (Mike Sellers) writes:
===============
This whole improbable set up may seem to be begging the question of the
original Chinese Room symbol manipulation scenario (though set up in the
large), but it contains an important difference: Each Booth is blind to
the functioning of the others, and knows only about its own activity.
Individually they are rather primitive, if flexible, cuneiform processors,
and possess none of the attributes we are interested in. However, taken
together, we must conclude that (through some process we do not as yet
understand (!)) the whole mass of them together will eventually --given
the necessary and sufficient environmental input-- take on awareness,
understanding, and cognizance. This must be the case, since the only
existing example of a thing possessing these qualities aquires and
maintains them by just such a system.
===============
This is true ONLY if you assume that the only processes going on in
the human mind are algorithmic. This seems to be the exact question
we are trying to resolve.
-Fred Gilham gilham@csl.sri.comandrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) (01/10/90)
How can Searle assert that syntax in and of itself cannot give rise to semantics? An existence proof for the contradiction is there for all to see in (at least) every mentally normal human child. Isn't it really this simple? -- ........................................................................... Andrew Palfreyman andrew@dtg.nsc.com Albania before April!
ele@cbnewsm.ATT.COM (eugene.l.edmon) (01/12/90)
In article <139@daedalus.nsc.com> andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) writes: > >How can Searle assert that syntax in and of itself cannot give rise to >semantics? An existence proof for the contradiction is there for all to see >in (at least) every mentally normal human child. Isn't it really this >simple? >-- No, it isn't this simple. As children learn semantics they are not picking it up from the syntax. If you still think this I would be interested in your further developing the argument. In fact, it seems to make more sense to argue that children learn semantics first as they learn words then how to string them together in sentences. -- gene edmon ele@cbnewsm.ATT.COM
andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) (01/12/90)
In article <8489@cbnewsm.ATT.COM>, ele@cbnewsm.ATT.COM (eugene.l.edmon) writes: > > andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head) writes: > >How can Searle assert that syntax in and of itself cannot give rise to > >semantics? An existence proof for the contradiction is there for all to see > >in (at least) every mentally normal human child. Isn't it really this simple? > > No, it isn't this simple. ok - so what's the scoop? > As children learn semantics they are not picking it up from the syntax. aha! - a contradiction; i'm all ears.... > If you still think this I would be interested in your further developing > the argument. i must have missed something somewhere.... > In fact, it seems to make more sense to argue that children > learn semantics first as they learn words then how to string > them together in sentences. at least it's obvious to you. that's nice, but hardly counts as a refutation, eugene. -- ........................................................................... Andrew Palfreyman andrew@dtg.nsc.com Albania before April!
ele@cbnewsm.ATT.COM (eugene.l.edmon) (01/14/90)
In article <492@berlioz.nsc.com> andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) writes: >In article <8489@cbnewsm.ATT.COM>, ele@cbnewsm.ATT.COM (eugene.l.edmon) writes: >> In fact, it seems to make more sense to argue that children >> learn semantics first as they learn words then how to string >> them together in sentences. > at least it's obvious to you. that's nice, but hardly counts > as a refutation, eugene. > Well let's see if this works. I make up a new word, say acadedementia, and tell you what it means. You then know all you need to use the word correctly. However, if I instead tell you not the meaning but the syntactic rules associated with use of the word, you will not be able to use it correctly. Now it seems to me (and by this I don't necessarily mean obvious) that an adult's ability to use a word correctly from learning its meaning depends on a long history with the syntax of the language. -- gene edmon ele@cbnewsm.ATT.COM
lee@uhccux.uhcc.hawaii.edu (Greg Lee) (01/14/90)
From article <8527@cbnewsm.ATT.COM>, by ele@cbnewsm.ATT.COM (eugene.l.edmon): >... >Well let's see if this works. I make up a new word, say acadedementia, >and tell you what it means. You then know all you need to use >the word correctly. However, if I instead tell you not the meaning >but the syntactic rules associated with use of the word, >you will not be able to use it correctly. It doesn't work. Whether syntactic rules can be sufficient to characterize correct usage is under discussion. In your scenario, you presuppose that the answer is that they cannot. Consequently, as an argument that syntactic rules are not sufficient, this begs the question. A way of phrasing the central difficulty is: Given an incorrect usage, what principles allow one to decide whether the violation is syntactic or semantic in nature? And since rules of usage are typically phrased in terms of co-occurrence of items of certain categories (tokens of certain types), this leads to the question: What is the difference between a syntactic category and a semantic category? Personally, I don't think there are any answers to these questions. Greg, lee@uhccux.uhcc.hawaii.edu