On 26 Mar 2015, at 08:05, Bruce Kellett wrote:
Bruno Marchal wrote:
On 25 Mar 2015, at 12:25, Quentin Anciaux wrote:
Multiple realisation does not undermine physical supervenience...
what undermine it, is that you're forced to accept (with the movie
graph argument) that the consciousness is supervening on the movie
+ broken gate... which is absurd, and the conclusion is either
that physical supervenience is false or computationalism is...
Good summary. If you accept physical supervenience, you need to
accept that non active part of the brain have active part in the
brain, basically. It makes clear that it is not the material brain
or material computer which does the thinking, but the abstract
person run by any sufficiently robust programs, with a robustness
defined to its most plausible computations above its substitution
level above and below the substitution below.
I think that all the MGA establishes is that if the film taken of
the physical states of the brain is a good copy, then consciousness
can supervene an that copy as well as the original.
Let me try to summarize the argument as I see it. We are conscious
and we have brains that seem to be connected with the conscious
state, such that a reasonable first model is that consciousness
supervenes on the physical brain -- we alter the brain, we affect
the conscious state, and the conscious state, being deterministic,
reciprocally affects the brain. (Changed thoughts are correlated
with changed brain states.)
The observation is then made that we could, quite probably, simulate
the brain state to any desired level in a computer (universal Turing
machine). The question is: does consciousness supervene on the
physical state, or on the abstract calculational state represented
by the computer?
Given that the computer simulation has the same conscious state as
the original brain, it follows that copies of the conscious state
can be made. In so far as these are accurate copies of the original
physical state, they are all the same conscious moments -- we only
create different consciousnesses when the inputs differ between
copies -- and then the states are no longer identical.
None of this argues against conscious supervening on the physical
rather than on an abstraction in Platonia. The MGA, as I understand
it, was designed to undermine this conclusion. The movie image
projected on the original neural plate recreates the original
conscious state. But we can degrade the neural plate. As long as we
project the same movie copy, the conscious state is unchanged. It is
argued that this is absurd. As far as I can tell, such an argument
hinges on the notion of conterfactual equivalence: the original
movie and the degraded plate are not counterfactually equivalent.
I simply say, so what! Counterfactual equivalence does not have any
independent justification, and it is highly unlike to be sensible,
even in the context of computationalism.
You are quick here.
I might explain the stroboscope machinery which might help me to ask
you what you mean by consciousness supervening to a recording, given
that no computation at all is involved in the recording.
Basically, because the simulation of any given conscious state can
be carried out an any computer -- whatever the architecture,
physical construction, or programming language. As long as the
original state is accurately simulated, the conscious state will be
the same.
It is not the state which must be correctly simulated, it is the
relation between those state/ You need the truth of the proposition IF
the input change, I do some this or that". That truth is part of what
define my person, and it is not applied in the movie. The movie is a
sequence of description of states, not a sequence of states related by
some universal number.
But these different instances of the calculation are generally not
counterfactually equivalent, nor need they be -- they only have to
simulate the original state to the required degree of accuracy
But there is no simulation here, in the technical sense of simulation.
I say that x simulates y on z if for all y z, the sentence phi_x(y, z)
= phi_y(z) is provable in RA.
-- they may differ to any degree whatsoever for their calculated
states before and after the target conscious moment.
If one x simulated y, an infinity of different x will simulated y as
well. We agree that there is a measure problem.
This comes back to my original question: since all possible programs
are run by the dovetailer, how do we ensure that conscious beings
see an ordered and predictable world.
Indeed that is the question. To answer it I have asked a Löbian
number. The answer, to make simple, is that if you abstract on
falsities (for knowledge), or illusion (for physics) the logic of self-
reference constraints the problem and show that in those direction the
reality kicks back ans is structured, so we can extend the formal
"measure one" into a physics (calculus of uncertainty).
Only a set of measure zero among all possible programs would give
that result.
I suspect you don't take the first person views (the modal variant of
relational justification) into account.
I don't pretend it is simple. You need to understand some theorems in
computer science, which has been exploited a lot to get those machine
logics.
Bruno
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.
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 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.
