Re: Implementation

From: <hal.domain.name.hidden>
Date: Sun, 25 Jul 1999 16:30:25 -0700

Let me offer an alternative approach to the Maudlin/Marchal implementation
question which is consistent with computationalism but avoids some of
the difficulties.

Computationalism says that implementation of a certain class of algorithms
is necessary and sufficient for consciousness. Maudlin presents an
experiment where a seemingly minor and (he argues) irrelevant change makes
the difference between whether a conscious computation is implemented
or not.

However to conclude that Maudlin's change (adding an inert block) makes
consciousness go away (even assuming computationalism) is a fallacy.
We agree that the initial computation (sans block) is conscious.
We agree that adding the block changes the computation (by changing
counterfactual behavior). But we cannot conclude from this that the
resulting computation is not conscious.

The new configuration, with the block, implements a different computation
than without (if you take counterfactuals into consideration). But have
we proven that this new computation is not conscious? No! All we have
proven is that this is different from the computation we started with.

But more than one computation can be conscious, obviously. It is
conceivable that the new computation, although different, is conscious
as well. This is a possible escape from Maudlin's argument.

Let me further offer a rationale for why this might be plausible.

Consider a human brain which is listening to music with its eyes closed.
It will not be having very much activity in the visual cortex. Now we,
acting as a demon, come along and shut off much of the visual cortext.
We disturb the neurons so that they will not behave properly. In fact,
we set it up so that the whole brain will shut down if a certain level
of activity is reached in the visual cortex.

Now as long as the brain keeps its eyes closed and keeps listening to
music, it will not notice these changes. But if it opens its eyes, it
will experience a drastic change, including loss of consciousness.

The question is whether, as long as the brain keeps its eyes closed, the
brain will continue to be conscious even though we have made this change.
I think the answer is clearly, or at least very plausibly, yes, it will
remain conscious. All the neural activity will stay the same as it
would have if we had not intervened, and there is no reason to expect
consciousness to evaporate just because we made a change which *could*
terminate conciousness if certain events occur.

If you agree with this, I would argue that getting to Maudlin's
situation (with the block) is now just a matter of degree. In the
listening-to-music case, the change we made was relatively minor, and we
left the brain with a considerable amount of freedom of action before
it would stumble onto a counterfactual state which would terminate it.
In Maudlin's case, the brain is much more restricted. It has only a
minutely narrow path it must follow before it falls into a counterfactual
state which will make it stop. As long as it stays on that path, doing
exactly what we predicted, it will behave normally and will not be "aware"
that danger lurks at every step.

We can imagine a range of situations where we provide greater or
lesser amounts of restrictions in terms of counterfactuals. If minor
restrictions leave the brain conscious, it seems plausible that major
restrictions would do so as well.

If we agree with this argument, we can have both supervenience and
computationalism, it seems to me. We agree that Maudlin's machine changes
the program which is instantiated, but we claim that the new program
is also conscious.

Hal
Received on Sun Jul 25 1999 - 16:45:20 PDT