Re: Fwd: COUNTERFACTUALS

From: <hal.domain.name.hidden>
Date: Fri, 9 Jul 1999 23:48:25 -0700

Christopher Maloney, <dude.domain.name.hidden>, writes:
> Marchal wrote:
> > what Maudlin and me have showed is that
> >
> > EITHER the appearance of hardware and consciousness
> > is explain(able) by the theory of possible softwares (computer science,
> > ...)
> >
> > OR the computationalist hypothesis is false.
>
> Please explain this, if you have time. I'm sorry if you've already
> given more detail in other posts, perhaps you could point me to the
> archives.
>
> What exactly is "the computationalist hypothesis"?
>
> I gather from the above that "the physical supervenience thesis"
> has something to do with explaining consciousness as a software
> program, but I'm confused. I would have *guessed* that that is
> the "computationalist hypothesis", but you say that those are
> incompatible, so it can't be.

I got hold of Maudlin's paper yesterday and I will hopefully post more
on it this weekend.

The computationalist hypothesis is that implementing a certain kind
of computation is necessary and sufficient for creating consciousness.
It is an important part of this hypothesis that the implementation must
include counterfactuals. That is, the implementation must mirror not
only the behavior of the program given the specific inputs encountered
on the run, but it must have been such that if other inputs had occured,
the implementation would still have faithfully followed the rules of
the program.

This last point is the crucial part for the paradox Maudlin constructs.

The supervenience thesis is that whether consciousness occurs depends
solely on the physical activity of the system. Two systems with the
same physical activity will both produce the same consciousness. In
particular, adding an inert object which does not interact with the
system should not change its degree of consciousness.

Although these two theses appear compatible, they are contradictory.
To oversimplify somewhat, the reason is because of the need to implement
counterfactuals in order to produce consciousness in the computationalist
hypothesis. Since counterfactuals don't actually occur, this contradicts
the supervenience hypothesis, which says that consciousness should only
depend on physical activity, which means it depends on what actually
happens.

Maudlin constructs an elaborate variant on a Turing machine. It is
used to replay a previously computed conscious experience, and we want
to ask whether the replay is conscious. Because it is a replay we
know exactly what states it will go through, but it is not enough,
by the computationalist hypothesis, to arrange that the machine
goes through such a series of states. In addition, the machine has
to handle counterfactuals - that is, it must be such that if the
input were different from what was used on the earlier runs, then it
will still do the right thing.

Maudlin proposes to add to the TM a mechanical clockwork computer which
handles counterfactuals. This portion of the computer would be activated
if the input were different from what is expected. It is only used in
that case, otherwise the main TM just plows through the states that
were pre-computed from an earlier run.

Now, because this is a replay and all the inputs are the same, the
clockwork computer will never actually be activated. Nevertheless
this extra clockwork mechanism MUST be present, according to the
computationalist hypothesis, otherwise the system does not handle
counterfactuals and so is not conscious.

Maudlin then proposes to insert a chunk of wood among the gears of
the clockwork computer such that if it ever tried to start running, it
would be blocked. This piece of wood does not need to touch anything,
it just needs to be positioned so that if the gears tried to turn they
would jam on the wood. (Keep in mind that the gears never actually do
start to turn.)

By doing so he has changed the machine so that it no longer handles
counterfactuals properly. Hence by the computationalist hypothesis,
the machine will no longer be conscious. However, his inert chunk of
wood makes no actual difference in the activity of the machine. Hence
by the supervenience hypothesis, the machine continues to be conscious.

The two theses give opposite predictions and hence are incompatible.

This short description does not really do Maudlin's 20-page argument
justice but hopefully explains the basic idea.

Hal
Received on Sat Jul 10 1999 - 00:00:44 PDT