From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!psych.toronto.edu!michael Mon May 25 14:06:02 EDT 1992
Article 5722 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!psych.toronto.edu!michael
>From: michael@psych.toronto.edu (Michael Gemar)
Subject: Re: Comments on Searle - What could causal powers be?
Organization: Department of Psychology, University of Toronto
References: <1992May17.225236.3588@oracorp.com>
Message-ID: <1992May18.185935.13789@psych.toronto.edu>
Date: Mon, 18 May 1992 18:59:35 GMT

In article <1992May17.225236.3588@oracorp.com> daryl@oracorp.com (Daryl McCullough) writes:
>In article <1992May5.204157.23037@psych.toronto.edu> michael@psych.toronto.edu 
>(Michael Gemar) writes:
>
>> While I agree that a computer program is, in itself, not a physical
>> entity, it is hard for me to see how implementation changes things.
>> Remember that,in order to counter Searle, *all* implementations of the
>> same program must generate semantics or have the appropriate "causal
>> powers" [...]  However, what do all possible implementations of a
>> program have in common *except* the abstract structure?
>
>A specification of a real-world system should really involve a
>program, together with a specification of the interface, which is the
>real-world interpretation of the inputs and outputs. The program
>itself is a purely mathematical object, but the interface
>specification grounds the system in the real world. With this notion
>of a specification of a system, different implementations of the same
>specification share causal properties, in addition to abstract
>structure. Two implementations are affected by the world in the same
>way (or in analogous ways), and in turn affect the world in the same
>way.

I take it then that you agree with Harnad that, without such real-world
connections, there can be no semantics.  Thus, Searle is at least right
in saying that symbol manipulation in and of itself is insufficient for
semantics.

This position still seems to leave SHRDLU's Dilemma, the problem of a
program whose "world" is "a purely mathematical object".   Could SHRDLU
ever have semantics?  As I interpret your claim above, it could not,
simply because it has no interface with the "real" world.  However, we can
imagine that SHRDLU's internal "World" can mirror ours to an arbitrary
level of accuracy.  This seems to lead to a problem - the "real" world
generates meaning, where a reproduction of that world to an arbitrary 
level of detail in a computer does not.  Something seems wrong somewhere...


>> Remember that the same program can be implemented on a computer,
>> with beer cans and string powered by windmills, by the Bolivian
>> economy, by a school offish directed appropriately, by the interaction
>> of galaxies, etc.
>
>I don't think that an electronic computer and the Bolivian can both be
>implementations of the same specification, because they don't provide
>the same interface. However, if there is a translation device that allows
>me to communicate with the Bolivian economy in terms of alpha-numeric
>characters, then I will admit the possibility that the combined system
>might be a perfectly good implementation.

And thus might be conscious?

- michael



