On 31 Mar 2015, at 07:17, Bruce Kellett wrote:
meekerdb wrote:
On 3/28/2015 11:36 PM, Bruce Kellett wrote:
Bruno has acknowledged that this is not what the MGA shows. MGA
simply shows that his version of computationalism is incompatible
with physical supervenience. This cannot be seen as surprising
since it is explicitly built into computationalism that
physicalism is false.
That's not my understanding. Bruno's argument starts with assuming
that a part, or all, of your brain could be replaced by a digital
AI with the same I/O and if done at a suitably low level of detail
(probably neuronal) you conscious inner life would be essentially
the same. That seems to me to be assuming physicalism as the basis
of consciousness.
This contradicts what you say below about Bruno assuming that only
certain special processes institute consciousness.
I think there is an ambiguity, or uncertainty, about just what the
program that is to replace part or all of your brain does. If the
program is just a simulation of the actual physical brain, neuron by
neuron, synapse by synapse, so that that physical laws that govern
the behaviour of these brain elements are instantiated by the
computer, and act on the initial data given by the state of the
brain when the program is started, then there will be no essential
difference between the program and the brain it replaces. In this
case you might say "Yes, doctor", with some confidence. The
necessary programming would presumably be well understood since the
brain is deterministic at the level with which we are concerned, and
the physical/chemical laws can be determined. If the initial state
can be ascertained with sufficient precision without killing you,
then the simulated computer brain substitute acts just like the
original, so should give no problems.
This understanding is based on the idea that consciousness
supervenes on the processes and states of the physical brain. These
have been replaced by equivalent physical processes, so
consciousness should remain intact. There is no appeal to
computationalism here. The simulating computer has to perform many
detailed calculations to carry through the operation of known
physical laws on the initial data, but I don't think anyone is
saying that consciousness supervenes on such calculations.
The other approach is to assume that the computer used to replace
your brain is running a true AI program. It is not simulating the
physical processes piece by piece, but running some black box
program that has been shown to reproduce known brain outputs for
some range of suitable inputs. The program is presumably supposed to
implement the universal TM computations upon which consciousness
supervenes independently of the underlying hardware/wetware. If this
is the model you have in mind, then the computationalist model
directly contradicts physical supervenience, right from the outset.
Now, I think the interesting question to ask is: "Given these two
different implementations of the brain replacing program, would you
have equal confidence in both possibilities?"
I think the answer would, in general, be "No!". The program that
assumes physical supervenience can be tested element by element, so
that once it has been shown to truly follow the known chemical and
physical laws, and accurately reproduces the structure of your
actual brain, it will be counterfactually correct, and could be
trusted into the future.
The alternative, computationalist model cannot be tested in this
way. Basically because it is necessarily holistic. Consciousness is
assumed to supervene on a particular type of computation, but is
your computationalist program the same as mine? How do we know? I do
not think the we could ever guarantee that such an AI device was
counterfactually correct for /your/ brain. Many artificial learning
programs, based on neural nets or the like, can be trained to
perform with great reproducibility on the training data set, but
fail miserably once one goes outside this data set. They are not
counterfactually correct, and I do not know how you could ever
ensure the necessary counterfactual correctness, even if you did
imagine that you knew precisely the sort of computation upon which
consciousness supervened.
So I would reject the computationalist program right at the start --
I would not say "Yes, doctor" to that sort of AI program.
Nor do I.
That is why I say that my definition of computationalism is weaker
than most in the literature.
Computationalism, as I defined it, assumes only the existence of a
level of substitution such that you survive with a digital (Turing
emulable) functional substitution made at that level.
In the six first step of the UD argument, I suppose the level high
(but still describing the biology of neurons and glials cells), to
make the reasoning more easy. But the conclusion hold up even for
someone who say that to get its relevant actual state, we need to
simulate the while universe, from the big bangs, at the level of
superstring theory, with (10^(10^100)) hexadecimals exact.
This is because that dumb little Robinson Arithmetic emulates that
"artificial brains", infinitely often, and with sometimes *much*
bigger number of decimals.
Note that I do not assume physical digitalism. But for someone making
the level so law that only some actual story of a universe is needed
for its/his/her consciousness to exist, + comp, in that case entails
physical digitalism. I doubt this makes sense, but if true, it has to
be deduced from inside arithmetic.
Whatever low the level you choose, even including oracles, once you
bet that it is Turing emulable, then you are emulate by all universal
numbers in arithmetic.
Are you OK with the step 2? Number like (10^(10^100)) might slow
down the machinery a lot, but if you agree with the step 2, the first
person cannot be aware of such times. Also, some tasks, like the
universal one, (or a still-weaker called subuniversal) have an
unboundable efficiency, by some speed-up theorems in proof theory and
computability. They might play some role.
Of course, I assume the Church-Turing thesis. This assumes some
realism on the possible digital machines and machineries, equivalent
with realism on a tiny fragment on which intuitionists and classical
mathematicians agree. Most physicists used stronger mathematical
theories. And Brent made me realize that RA is even a strct finitisme
in Van Bendeghem sense. RA is consistent with there is a biggest number.
May be comp is false, but that is why I make it precise and look for
the consequence. Without Everett QM I would still be sure it can't be
true, but perhaps still study it, for the beauty of mathematics.
The big discovery is the discovery of the universal machine, by
mathematician trying to clarify some paradoxes.
Bruno
Bruce
The MGA is, therefore, largely irrelevant, because it does not
prove anything that we didn't already know. It certainly does not
show that consciousness is an abstract process in Plationia,
independent of any physical process.
Bruno assumes that only some special processes instantiate
consciousness and these are characterized by being computations of
some kind, i.e. a sequence of states that could be realized by a
program running on a Universal Turing Machine (not necessarily
halting). Since the consciousness computation defined this way is
an abstract mathematical process in Platonia; it is equivalent to
assuming consciousness is instantiated by an abstract mathematical
process.
Brent
--
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 [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
http://iridia.ulb.ac.be/~marchal/
--
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 [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
