On 9 May 2017 6:16 a.m., "Bruce Kellett" <bhkell...@optusnet.com.au
<mailto:bhkell...@optusnet.com.au>> wrote:
On 9/05/2017 1:57 am, David Nyman wrote:
On 8 May 2017 8:21 a.m., "Bruce Kellett"
<bhkell...@optusnet.com.au <mailto:bhkell...@optusnet.com.au>> wrote:
On 8/05/2017 4:53 pm, David Nyman wrote:
Both Hoyle's pigeon holes and Barbour's time capsules
assume that there is a coherent underlying physics with
regular exceptionless laws. Until you have something like
that, you cannot define consistent continuations.
But I'm afraid that's implied by assumption unless one
takes the view that the evolution of physical states is
fundamentally incomputable,
But I thought that that was what Bruno claimed. If one
assumes physics in one's derivation, then the circularity is
vicious.
Oh dear no, that's not right at all. What is uncomputable is any
extrinsically conceived *extraction* of physics from the
computational Babel consequent on the theory. There is no
possible search function for this. That extraction then is
necessarily a complex consequence of observer selection. Post
such extraction, the evolution of physical states is then by
assumption finitely computable, modulo the FPI, else
computationalism must fail as a theory of mind or of physics. At
this point the objective situation, mutatis mutandis, is
essentially equivalent to Everett's relative state assumptions.
The other point on which I must take you to task is again the
question of circularity. It's not the job of computationalism's
toy model to explicate every detail of the extraction of physics,
although it's already the case that it *predicts* the multiple
continuations implicit in the wavefunction, which is more than
can be said for QM itself which merely retrodicts them (again
modulo the FPI). Given the conjunction of the assumption of
computationalism and our observation of the physical environment
described by QM, all the theory has to show at this stage is that
it is not incompatible with these data (as it would be if, say,
the evolution of the wavefunction itself were shown to be
uncomputable). It should further explicate some reasonably
convincing justification for why just such a physics might be
expected to underpin the effective environment we observe. But
the *facts* of our observation of such a physics are not at
issue. There is no relevant question of circularity to deal with
here.
As to the so called Occam catastrophe, as exemplified in your
Boltzmann brain scenario, potential resolution necessarily can be
understood under computationalism only from a first person
perspective, as I previously suggested to you. We need to
justify, in terms of a subjective measure, why we should indeed
expect the physics we observe to emerge as the predominating
computational mechanism underlying our normally intelligible
perceptions. To do this we only need to show that "last Tuesday"
computational snippets can only reinforce, and magical or
unintelligible ones cannot interfere, with "normally
intelligible" and complexly connected continuations. A way to
grasp this intuitively is in terms of something like Hoyle's
"amnesic multiple personality" heuristic which, though as you
say it was originally based on the assumption of physics, IMO
illustrates the relevant considerations equally intuitively on
computational assumptions. In any case, the analogy of a
multitasking OS that I also mentioned suffices equally well in
this regard.
From this perspective, no amount of white noise in continuations
of "Boltzmann" computations could make a substantive subjective
difference. The reason being that the consequence is
overwhelmingly likely to be a total subjective unintelligibility
which will plausibly tend to be utterly swamped, in the struggle
of forgetting and remembering, by "normally intelligible"
continuations. The FPI is, obviously, the relevant consideration
in this regard. This is what I meant when I said that an absence
of evidence for this sort of pathology or unintelligibility is
not evidence of its absence. It suffices that these out of phase
components of experience be swamped in the battle for what one
might term personal subjective emergence. They just typically get
forgotten far more frequently than they get remembered by Hoyle's
multiply solipsistic agent. Hence what we may think of as
pathological scenarios would be expected to be very poor and
haphazard candidates in the ongoing struggle for apparently
persistent, pervasive and lawful subjective emergence. What would
emerge with these characteristics would then be consistently
remembered histories underpinned by a robust and reiterative
physical mechanism whose highly selective observation by us would
then be the final evidence of its predomination in this epic
personal struggle.
I gave you an illustration a few days ago (on which you didn't
comment) of what one might term the "psycho-theological" aspect
of computationalism. I said that consciousness or first person
subjectivity was really a pointless cherry on the cake of physics
whose mechanism must be assumed to proceed without any a priori
need of such a baroque supernumerary assumption. Indeed it can
only be an a posteriori datum tacked on to the physical scheme of
things. Computationalism, by contrast, can only be understood in
the final analysis as a synthesis of all possible subjective
personal histories. "Point of view" is then just what prevents
them from all happening at once. Thus physics, under the same
assumptions, can in turn be understood finally as the successful
computational generator underlying the "dreams of the machines".
David
I find most of what you say here very much a matter of wishful
thinking, and not entirely consistent at that. Let me come at it
in a different way.
I find Barbour's idea of time capsules quite helpful here. Each
time capsule is a self-contained conscious moment. There is no
progression necessarily involved, so the computation that gives
one conscious moment is complete in itself, and independent of
other such conscious moments. (In Barbour's picture, these moments
are points in configuration space that are related physically, but
we do not use that aspect here.) In the moment, you are
self-aware, and aware of memories that give you a concept of self.
But in that moment there is no way that you can know whether these
memories are veridicial or not -- they could well all be
completely false, in which case there is no "you" that continues
through time as a related series of experiences. Each experienced
moment is complete in itself, and there is no continuation. If all
you have is the moment of consciousness, you can go no further
than this. It is all an illusion, and there is no physics to extract.
Of course, this is a solipsistic conclusion, but there is nothing
in our experience of consciousness that shows solipsism to be
false. The "I" is the "I" of the moment, nothing more.
Now consider the UD in arithmetic. It dovetails all possible
programs -- does all possible computations -- but most
computations have nothing to do with consciousness. If we use
Boltzmann's thermodynamics as an illustration of the situation,
the computations of the dovetailer represent a state of thermal
equilibrium, a state of maximum entropy. The characteristic of
thermal equilibrium is that every microstate is equally likely --
a state of complete chaos. Similarly, in the dovetailer, every
computation is equally likely and there is no order whatsoever.
Occasionally, in Boltzmann's thermal equilibrium there are
fluctuations to states of lower entropy in which some order
emerges, but according to the second law of thermodynamics, these
always return to equilibrium. Similarly, in the computations of
the dovetailer, there are occasionally computations that make some
sort of internal sense. Some of these correspond to conscious
moments. But, as in the thermal case, these rapidly return to
meaningless noise. Small fluctuations to momentary order are
overwhelmingly more likely than larger fluctuations to order that
persists over time -- or computations that correspond to an
extended sequence of (consistent) conscious states. In fact,
within the dovetailer there are undoubtedly sequences of
computations that correspond to the entire history of the
observable universe, from the big bang through to the final heat
death. But such calculations are of measure zero in the overall
picture.
So, if one is to take the statistics of computations that pass
through one's instantaneous conscious state in order to extract
meaningful physics, one will find that the overwhelming majority
of these computations are of short-lived conscious moments that
rapidly return to meaningless chaos, nothing more. The dovetailer
would then say that no consistent physics can ever be extracted
from the statistics over conscious moments, because these
statistics are dominated by chaotic continuations.
That does not necessarily mean that no consistent physics exists
-- as I said, all of physics will be in the computations of the
dovetailer somewhere. All it means is that such physics cannot be
extracted by considering individual conscious moments as primary.
Physics has to have an independent existence, or it has no
existence at all, and solipsism is the only answer.
Thanks for your interesting comments Bruce. It's difficult to know how
to respond because I still feel we are somehow at cross purposes.
Perhaps I should start by commenting on the matter of wishful
thinking. As I said to Brent, I don't see myself as an apologist for
computationalism. Rather I'm trying to understand its possible
implications. My views on the matter have changed over the years,
particularly as a result of discussion on this forum, and I am perhaps
more persuaded than I was originally (which is to say not at all). But
I don't believe my wishes are really part of the picture. It's rather
that I deliberately allow myself a modicum of intuitive latitude in
these discussions - one might say going just a bit further than
confidence in my position would allow in less speculative topics -
mainly in order to stimulate the broadest possible critique from a
valuable community of varied expertise.
The thing that changed my view the most, in direct contrast to what
you say above, was considering the matter from a first personal
perspective rather than the traditional "view from nowhere". By this I
don't mean that I believe consciousness to be more "fundamental" than
physics, but rather that I began to see Bruno's point that physics
would have to be a subjectively driven filtration from the
computational Babel if computationalism as a whole were to make any
sense. The alphabetical Babel is a useless chaos because of the
impossibility of an extrinsic search function but the possibility of
intrinsic self-selection via subjectivity seems like a more promising
proposition. But of course this demands a fundamental theory of
subjectivity and it is indeed a fresh approach to this vexing question
that Bruno has brought to the party. I suspect the amount and
occasional ferocity of focus on his approach stems from something more
than the mere desire to prove him wrong.
All that said, the point where I feel we're talking past each other is
precisely on the issue of the extraction of physics. As I tried to
point out, the facts of the matter are not what is at issue here.
Rather the question hinges on whether it is reasonable to suppose,
against the assumed background of UD*, that the vast majority of
conscious moments would be a consequence of their supervening
computationally on the observed physical environment rather than on
random fluctuations. And the reason I make use of Hoyle's or Barbour's
analogies is that they seem to lead naturally to a form of solipsism
that I've called the amnesic multiple personality. Hoyle himself
points this out. And far from seeing this as terminal for the argument
(although frankly it amuses me that this despised philosophical
position might be thus rehabilitated) I find it genuinely
enlightening on the question of how to reconcile probability in a
quasi-frequentist framework in an overall context in which "everything
happens" with certainty.
This is where what I've called the struggle or battle between
remembering and forgetting comes to the fore. Admittedly it's an
unfamiliar idea and hardly easy to come to terms with, not least for
me, but in struggling with it I've found it genuinely enlightening
with respect to many otherwise intractable puzzles, not least to do
with personal identity and its putative history. In terms of our
discussion, the area of disagreement which most puzzles me is why you
take the view that the experience of a unique agent such as the one
analogised by Hoyle or Barbour would be dominated by random events
rather than the order imposed by the predominance of a robust
physical-computational mechanism. I don't see why a mechanism
"singularised" or selected in this manner would *at that point* fail
to possess the necessary "independence" in the sense you demand of it.
After all, this seems little different in effect from the so called
anthropic selection invoked in other frameworks. I'm probably being a
bit slow here, but could you explain again why you take this view,
preferably putting it in the context of the approach I describe above
if at all possible. As I said to Brent, a counter-argument is more
enlightening than one that is merely different.