From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!psych.toronto.edu!michael Tue Feb 11 15:25:20 EST 1992
Article 3544 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: Strong AI and panpsychism
Message-ID: <1992Feb6.191559.12739@psych.toronto.edu>
Organization: Department of Psychology, University of Toronto
References: <1992Feb5.020733.21580@bronze.ucs.indiana.edu> <1992Feb5.185805.15433@psych.toronto.edu> <1992Feb6.053810.22191@bronze.ucs.indiana.edu>
Distribution: world,local
Date: Thu, 6 Feb 1992 19:15:59 GMT

In article <1992Feb6.053810.22191@bronze.ucs.indiana.edu> chalmers@bronze.ucs.indiana.edu (David Chalmers) writes:
>In article <1992Feb5.185805.15433@psych.toronto.edu> michael@psych.toronto.edu (Michael Gemar) writes:
>
>>Is there any independent principled reason for not allowing time-dependent
>>definitions, or are they ruled out merely so that the above situation is
>>not a problem for functionalism?

>Firstly, there's nothing wrong with modifying one's theses in order to
>take account of potential counterexamples.  It happens all the time.

No doubt.  My question, however, is *why* one would want to change
the thesis.  If the counterexample you want to rule out is rocks
implementing FSAa, I want to know *why* you want to rule that out, and
if the reason is not simply ad hoc.

>Secondly, I very much doubt that any functionalist ever intended
>time-varying states to count as state-realizations.

But *can* they count?  If not, *why* not?  Again, what is needed is 
a principled reason (or perhaps a definition of "state").

>Still, Putnam doesn't actually need time-varying states for the result
>to go through, on the assumption that any object has an (implicit
>or explicit) "clock" built into it, that ensures that it is in different
>states at different times.  But the states that appear are nevertheless
>highly gerrymandered.

"Gerrymandered" by what objective criteria?  

>I haven't examined Putnam's proof as closely as I should have, but it's
>not at all clear to me that it establishes that any object implements
>any FSA in the strong sense of not just getting actual but counterfactual
>state-transitions right; and in particular getting the transitions upon
>counterfactual inputs right.  If someone has a demonstration of this
>I'd like to see it.  Putnam's proof doesn't provide this, and I suspect
>that it may not hold up.

This may be - I don't have the sophistication necessary to determine
this myself.

>Finally, I think it's reasonable to require that states are what might
>be called "causally efficacious" states, i.e. such that the property of
>being in the state is capable of doing causal work, and not just within
>an incestuous circle of related gerrymandered states; so they can be
>hooked up to potentially cause simple actions in the environment for
>instance (obviously, this is closely related to the last objection).
>Pure time-varying states may not satisfy this, although Putnam's
>gerrymandered clocked states may, under certain assumptions.

This objection assumes that you can distinguish between functions
"in the enviroment" and functions "in the entity."  I have yet to
see a good way of telling these apart.   Also remember that 
FSAs can (like SHRDLU, my favorite example), *include* a "virtual
environment".                 

>I don't have a definitive line on this.  I'm not certain that
>restrictions on allowable states are needed, but if they are, I
>suspect that they will be motivated by reflecting on our reasons for
>thinking that a clock doesn't *really* implement an arbitrary FSA,
>and fixing the definition of "implementation" accordingly.

You're not sure if "restrictions...are needed" to do *what*?  To
accomplish *what*?  If it's just to rule out the possibility that
rocks can implement an arbitrary FSA, then this seems suspiciously
ad hoc to me...

- michael




