From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!bronze!chalmers Tue Apr  7 23:22:53 EDT 1992
Article 4784 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca sci.philosophy.tech:2454 comp.ai.philosophy:4784
Newsgroups: sci.philosophy.tech,comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!bronze!chalmers
>From: chalmers@bronze.ucs.indiana.edu (David Chalmers)
Subject: Re: A rock implements every FSA
Message-ID: <1992Mar29.003736.25807@bronze.ucs.indiana.edu>
Organization: Indiana University
References: <1992Mar25.094354.10243@husc3.harvard.edu> <1992Mar26.034816.29572@bronze.ucs.indiana.edu> <1992Mar28.100350.10367@husc3.harvard.edu>
Date: Sun, 29 Mar 92 00:37:36 GMT

In article <1992Mar28.100350.10367@husc3.harvard.edu> zeleny@zariski.harvard.edu (Mikhail Zeleny) writes:

>Irrelevant.  Although ideally we would like to get a trace that is both
>Eulerian and Hamiltonian, this is only desirable for aesthetic reasons.
>Your automaton needn't have a connected graph, for all I care: just pick a
>selection of initial states and input strings exhaustive of all of its
>combinatorial possibilities, and string their traces together.

This won't work.  Between consecutive traces, there will be a
discontinuity at which your "implementation" makes a state-transition
that is disallowed by the table.

>>See above.  I should note that even if such a sequence were possible,
>>it still wouldn't provide an adequate correspondence of logical
>>states to physical states.  e.g. take a 2-state, 2-input machine, such
>>that we want the state/input pair (S1,I1) to lead to S1, and (S1,I2) to
>>lead to S2.  Following the above means of construction, logical state
>>S1 will correspond to the disjunctive physical state P1-or-P2.  The
>>construction here will ensure that the combination (P1,I1) will lead to
>>a state realizing S1, and that (P2,I2) will lead to a state realizing
>>S2.  However, it ensures nothing about the behaviour upon the
>>combinations (P1,I2) and (P2, I1) (which never come up in the actual
>>sequence), and so cannot guarantee the appropriate transitions, i.e. 
>>that (P1-or-P2,I1) leads to a state realizing S1, and that (P1-or-P2)
>>leads to a state realizing S2.  So counterfactual sensitivity is still
>>lacking.
>
>Once I actualize all state transitions in the machine's table, your
>"counterfactual sensitivity" will be accounted for.

I note that this doesn't address the substance of my objection.  It
probably doesn't matter much, as that objection was made under the
assumption that you, following Putnam, were mapping inputs to inputs
and states to states ("the relevant remarks on p. 124" can only be
read that way).  Your exchange with Joseph O'Rourke has made it
clear that in fact you're mapping both inputs and states of the FSA
simultaneously onto internal states of the rock, which may as well
be entirely causally disconnected from the outside world.

This is an audacious strategy; if one allowed such mappings, then
anything might follow.  But of course one cannot map inputs this
way.  I quote Putnam himself, on p. 124:

  The inputs and outputs have certain constrained realizations,
  or at least their realizations must be of certain constrained
  kinds depending on our purposes; usually we are not allowed
  to simply *pick* physical states to serve as their realizations,
  as we are allowed to do with the so-called "logical states" of
  the automaton.

As your argument violates one of Putnam's basic assumptions, there's
no use trying to pretend that it's a version of his argument.

As I said in a previous message, the required constraints on
inputs and outputs are themselves a matter for debate, but at
the very least they have to be *inputs* -- i.e. they have to
supervene on a distinct region from that on which the internal
states supervene.  Which your "inputs" manifestly fail to do.

(In fact, your construction has the strange property that given
the state and input at any given time, not only is the next
state determined, but the next input is too.  Most unusual.)

>Again, I wish you had
>addressed my modal objections instead of jumping to conclusions.

If you wish to argue that strong conditionals can never have
determinate truth values, then good luck to you.  If you succeed,
it won't just be functionalism that falls.  (What was it that
Einstein said about "*if* a body is traveling at half the speed
of light..."?)  The conditionals invoked here are as innocuous
as they come, being truth-functionally determined by simpler
conditionals concerning the behaviour of objects whose maximal
states and environmental conditions are specified; physical laws
are enough to determine the truth-values here.  And the problems
with haecettism are no worse here than they are with conditionals
like "if I flip the switch on the lamp, then it will turn on".
Pardon me for not being too worried about them.

>For the third time, I ask that you
>formally publish your answer to Putnam's argument, or cease claiming that
>it has no force against functionalism.

I'm not interested in publishing trivialities.  If I published a
reply to every bad argument that I came across in the literature,
it would yield nothing but a padded resume.  It's not impossible
that I'll publish a substantial piece on computation and
implementation one of these days, but if I do, it will need a
lot more than a refutation of Putnam to justify its existence.

-- 
Dave Chalmers                            (dave@cogsci.indiana.edu)      
Center for Research on Concepts and Cognition, Indiana University.
"It is not the least charm of a theory that it is refutable."


