I find that discussions around the comp thesis keep coming back to the
'Movie Graph Argument' (MGA). Each time I read one of the accounts in
Bruno's SANE04 or COMP(2013) papers, or Russell's 'MGA Revisited', I get
the feeling that something crucial to the argument is missing.
The account in COMP(2013) is probably the most detailed, so I will look
at this in some detail. The MGA (FGA) is introduced as "a direct
Mechanistic argument showing that consciousness ... cannot possibly
supervene on physical activity of the brain." This is supposed to be
shown by deriving a contradiction from the assumption of physical
supervenience.
We can use an original biological brain, or an equivalent digital
replacement -- it does not make any significant difference to the
argument. The first point is that in some conscious experience, be it a
dream or anything else, there might be a portion of the 'brain' (in
quotes because it can be biological or digital) that is not activated,
so this can be removed without affecting the conscious experience. More
generally, we suppose that there is some part of the 'brain' that is
required, but is defective for some reason. However, serendipitously,
when that part is required, some cosmic event luckily stimulates the
required activity, so that the physical activity of the 'brain' is
maintained, and it corresponds to the actual physical activity relevant
for that computation. This breaks counterfactual correctness, but if the
'dream' and the processing is properly determined, the physical activity
corresponding to the computation, and relevant to it, is maintained.
Counterfactual correctness is thus shown not to be relevant to *that*
conscious experience. If the lack of counterfactual correctness were
able to change the personal experience, then the brain could recognize
where its internal inputs came from, which is considered to be absurd
since the assumption does not provide cognition of the elementary parts
of the computing machinery.
So far, so good. The MGA seeks to extend this line of reasoning to show
that the physical activity is not relevant at all. We image that we are
able to make a film of all the internal activity of the 'brain' during
some conscious experience. (A dream if we wish to reduce dependence on
physical inputs and outputs.) We now run the computation again after
breaking some or all of the original physical connections in the
'brain'. But at this time we also project the film directly on to the
machinery. Now, when the broken connections are needed, despite the fact
that they cannot give the relevant outputs, the machine will still
perform the original physical activity relevant for that computation --
the movie, which comes from a film of the correct activity, will
supplement any lacking information. The movie plays the role of the
lucky cosmic event in the previous example.
Now the first person experience will be absolutely identical to the one
which would have been obtained by the unbroken machinery. We can
eliminate more of the machinery, indeed we can eliminate /all/ of it,
without changing an active consciousness into a fading consciousness,
because this would be experienced by the subject, contrary to the
assumption that the person is never conscious of any part of the
internal machinery.
Counterfactual correctness might be restored at any stage by adding
additional counterfactual machinery, but, as before, this does not
really make any difference. Counterfactual correctness is relevant only
to the general requirement that different inputs can give different
experiences, but we are always using the same inputs here as we
repeatedly run the same program, although with less and less of the
original machinery intact -- always providing the missing bits from the
film of the original conscious computation.
Again, this all seems reasonably clear, but then the argument becomes
very clouded, and it is not at all clear what conclusions are being
drawn directly from this thought experiment. Bruno talks about the
possibility of lowering the level of substitution for the digital
'brain' replacement. He also mentions the MWI of quantum mechanics, as
suggested by Russell as a way to overcome the "non-robust universe"
objection to the dovetailer. The relevance of these comments is quite
opaque to me. Because then Bruno simply says: "Then, as an applied
science in the fundamental realm, we can use Occam's razor to eliminate
the 'material principle'."
The conclusion is "the FGA (MGA) shows that any universal machine is
unable to distinguish a real physical realm from an arithmetical one, or
a combinatorial one, or whatever initial notion of Turing universality
is chosen as initial basic ontology." Hence, consciousness is not a
physical phenomenon, nor can it be a phenomenon relating to observed
matter at all. Consciousness can no longer be related to any physical
phenomenon whatsoever, nor can any subjective appearance of matter be
based on a notion of primitive matter.
I hope I have summarized the argument as in COMP(2013) sufficiently
accurately for the present purposes. Most of the above is direct
quotation from Bruno's text, with paraphrases in some less significant
areas in order to shorten the presentation.
Now, having read this many times, and looked at the other summaries of
the MGA, I still feel that something crucial is missing. We go from the
situation where we remove more and more of the original 'brain',
replacing the removed functionality by the projections from the movie,
which, it is agreed, does not alter the conscious experience of the
first person involved, to the conclusion that the physical brain is
entirely unnecessary; indeed, irrelevant.
I am sorry, but this just does not follow. The original physical
functionality is admitted to be still intact -- provide, admittedly, by
the projected movie, but that is still a physical device, operating with
a physical film in a physical projector, and projecting on to the
original (albeit damaged) physical machinery. How has the physical
element in all of this been rendered redundant? The original
functionality of the 'brain' has been preserved by the movie; the
conscious experience is still intact even though much of the original
functionality has been provided by another external physical device. How
does this differ from the original "Yes Doctor" scenario in which the
subject agrees to have his brain replaced by a physical device that
simulates (emulates) his original brain functionality? I submit that it
does not.
The only difference between the movie replacing the functionality of the
original experience and having that functionality replaced by a computer
would seem to be that the computer can emulate a wider range of
conscious experiences -- it is 'counterfactually correct' in that it can
respond appropriately to different external inputs. The film, being a
static record of one conscious experience, cannot do this. But it has
been admitted that the film can reproduce the original conscious
experience with perfect fidelity. And the film is every bit as physical
as the original 'brain'. So the physical has not been shown to be
redundant. It cannot be cut away with Occam's razor after all. If it
were, there would be no conscious experience remaining.
I conclude that the MGA fails to establish the conclusions that it
purports to establish.
Bruce
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.