2017-05-08 8:26 GMT+02:00 Bruce Kellett
<[email protected] <mailto:[email protected]>>:
On 8/05/2017 3:59 pm, David Nyman wrote:
On 8 May 2017 4:53 a.m., "Bruce Kellett"
<[email protected]
<mailto:[email protected]>> wrote:
On 8/05/2017 3:14 am, David Nyman wrote:
I've been thinking a bit more about this and I'd like
to set out some further tentative remarks about the
above. Your professional expertise in these matters is
orders of magnitude greater than mine and consequently
any comments you might make would be very helpful. By
the way, it would also be helpful if you would read
beyond the next paragraph before commenting because I
hope I will come by myself to the fly in the ointment.
Firstly, and "assuming computationalism" on the basis
of CT + YD, we are led to the view that UD* must
include all possible "physical" computational
continuations (actually infinitely reiterated). This of
course is also to assume that all such continuations
are finitely computable (i.e. halting). Now, again on
the same assumptions, it might seem reasonable that our
observing such a physics in concrete substantial form
is evidence of its emergence (i.e. epistemologically)
as the predominant computational mechanism underlying
those very perceptions. Hence it might seem equally
reasonable to conclude that this is the reason that
these latter correspondingly appear to supervene on
concrete physical manifestations in their effective
environment.
Now wait a minute. We cannot escape the question of
measure. Why would it be reasonable to assume that a
physics of this sort should predominate in the manner
outlined above? Well, firstly, it would seem that the
generator of the set of possible physical computations
is infinitely reiterative​ and hence very robust (both
in the sense of computational inclusiveness a la step
7, and that of internal self-consistency). But who is
to say that the generators of "magical" or simply
inconsistent continuations aren't equally or even more
prevalent? After all we're dealing with a Library of
Babel here and the Vast majority of any such library is
bound to be gibberish. Well, I'm wondering​ about an
analogy with Feynman's path integral idea (comments
particularly appreciated here). Might a kind of least
action principle be applicable here, such that
internally consistent computations self-reinforce,
whereas inconsistent ones in effect self-cancel?
Also, absence of evidence isn't evidence of absence.
I'm thinking here about the evaluation of what we
typically remember having experienced. I can't help
invoking Hoyle here again (sorry). Subjectively
speaking, there's a kind of struggle always in process
between remembering and forgetting. So on the basis
suggested above, and from the abstract point of view of
Hoyle's singular agent (or equally Bruno's virgin
machine), inconsistent paths might plausibly tend to
result, in effect, in a net (unintelligible) forgetting
and contrariwise, self-consistent paths might equally
plausibly result in a net (intelligible) remembering.
I'm speaking of consistent and hence intelligible
"personal histories" here. But perhaps you would
substitute "implausibly" above. Anyway, your comments
as ever particularly appreciated.
I think the problem here is the use of the word
"consistent". You refer to "internally consistent
computations" and "consistent and hence intelligible
'personal histories'." But what is the measure of such
consistency? You cannot use the idea of 'consistent
according to some physical laws', because it is those
laws that you are supposedly deriving -- they cannot
form part of the derivation. I don't think any notion of
logical consistency can fill the bill here. It is
logically consistent that my present conscious moment,
with its rich record of memories of a physical world,
stretching back to childhood, is all an illusion of the
momentary point in a computational history: the
continuation of this computation back into the past, and
forward into the future, could be just white noise! That
is not logically inconsistent, or comutationally
inconsistent. It is inconsistent only with the physical
laws of conservation and persistence. But at this point,
you do not have such laws!
In fact, just as Boltzmann realized in the Boltzmann
brain problem, states of complete randomness both before
and after our current conscious moment are
overwhelmingly more likley than that our present moment
is immersed in a physics that involves exceptionless
conservation laws, so that the past and future can both
be evolved from our present state by the application of
persistent and pervasive physical laws.
Unless you can give some meaning to the concept of
"consistent" that does not just beg the question, then I
think Boltzmann's problem will destroy your search for
some 'measure' that makes our experience of physical
laws (any physical laws, not just those we actually
observe) overwhelmingly likely.
Thanks for this. However I'm not sure you've fully addressed
my "path integral" point, for what it's worth. Feynman's
idea, if I've got the gist of it, was that an electron could
be considered as taking every possible path from A to B, but
that the direct or short paths could be considered as
mutually reinforcing and the indirect or longer paths as
mutually cancelling.
Feynman's ideas relies on a physical theory within which one
can calculate the phase change along each possible path. The
upshot is that paths far away from the path of least action
have phases that cancel in the quantum superposition sense.
Note that the crucial input into this picture is that there
is an underlying physical theory, in terms of which one can
calculate the phase changes along each path. Also, it is
important to remember that Feynman's path integral is only
one means of calculating probability amplitudes in QM --
there are many other means of calculating these, and all give
the same results.
Hence the derivation of the principle of least action. So
the analogy, more or less, that I have in mind is that
Boltzmann-type random subjective states would,
computationally speaking, mutually reinforce identical
states supervening on the generator of "consistent" physical
continuations (bear with me for a moment on the applicable
sense of "consistent" here). IOW "If I am a machine I cannot
know which machine I am". So as long as the generator of
those consistent states is encapsulated by UD* - which is
equivalent to saying as long as the computable evolution of
physical states is so encapsulated (which it is by
assumption) - then we can plausibly suppose that the net
subjective consequences would be indistinguishable.
I agree that conscious states, whether Boltzmann brains or
parts of a longer calculation that does not start and end in
white noise, are, insofar as thise states are conscious, they
are indistinguishable. But the Boltzmann brain-type states
cannot reinforce the path that leads to coherent physics,
because the continuations are disjoint.
As to your most reasonable request for a non question
begging notion of consistent in this context, my tentative
answer rests on my remarks about the "struggle between
remembering and forgetting". Here's where I use Hoyle's
pigeon hole analogy, which is pretty much equivalent to
Barbour's time capsule one (as he acknowledges in TEOT)
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.
I think that white noise -> consciousness -> white noise must be
of an infinitsimal measure for programs generating consciousness
moments, because it needs a lot of informations to go from
consciousness moment to white noise, more than from
consciousness->consciousness-next moment, hence there must be
shorter program that goes from one instant to the next than from
one thing to totally something else...
