On 5/9/2017 10:28 PM, David Nyman wrote:
On 10 May 2017 3:04 a.m., "Bruce Kellett" <bhkell...@optusnet.com.au
<mailto:bhkell...@optusnet.com.au>> wrote:
On 10/05/2017 12:41 am, Bruno Marchal wrote:
On 09 May 2017, at 09:36, Bruce Kellett wrote:
Yes, it does seem that we are each outlining positions and
arguments that do not necessarily intersect at many points. I
will try and answer some of your more direct questions. Why do I
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." The reason is that I
am starting from a slightly different perspective -- I am
looking at the UD as a system in its own right. The questions
seem to concern statistics extracted from the behaviour of this
system. When approaching such a question, I tend to look on the
thermodynamic, or statistical mechanical properties of such a
random system. If you take the UD, with its completely random
operation over all possible (computer) programs, the analogy
that comes to mind is that of thermal equilibrium -- every
possible state has equal probability of occurring. Ergodic
theory is possible also relevant, but I have less familiarity
with that, so tend to stick to ideas deriving from Boltzmann.
Given this state of thermal equilibrium, states of some order --
such as conscious moments -- are going to be unlikely, and
fluctuations that give single conscious moments are
overwhelmingly more likely than more extended fluctuations that
give a sequence of related conscious moments.
Not sure I understand. The UD is not random at all.
The sequence of computational steps may be deterministic, but the
net result in the infinite UD is random. Think of Boltzmann's
case, a gas of a very large number of molecules. The molecules
move and interact according to entirely deterministic laws, but
ergodic theory indicates that after a suitable time, the motions
of the molecules will be effectively random. I think the same must
happen with the dovetailer: although each program is
deterministic, the dovetailing of infinitely many such programs
means that sequences of individual steps are random (or
indistinguishable from random).
What is random is the First Person Indeterminacy on all the
stable continuations of my states, as seen by the first person,
so the first thing to do is to get a mathematical theory of the
first person (which I take to Theaetetus, as Gödel's
incompleteness makes it work again, again Socrates opinion).
Given also the insight from Barbour that each conscious moment
-- time capsule -- is self contained,
I am not sure that such an intuition is correct. A conscious
moment needs at least two universal numbers, but in fine relies
on an infinity of them. Nor do I conceive such a thing as an
observer moment. the semantic of all first person view (the
modalities with "& p" attached to them) are topological.
Consciousness is always on an interval, not on a discrete point
in some time frame.
I was deliberately vague in specifying what was meant by a
"conscious moment". I doubt that it is of zero duration, but the
duration is indeterminate. Beside, there is no concept of time in
the UD, so it is hard to say what a conscious moment might
actually be -- some sequence of computational steps, perhaps --
but how many? A time capsule is certainly self-contained. Whether
these overlap or not to give a sense of continuity is another
question, and would seem merely to extend the notion of a
conscious moment in time somewhat -- but what is time?
and in itself, a complete explanation of our conscious
experience, the computations that pass through our conscious
moments are overwhelmingly likely to be random, with just small
fluctuations from equilibrium. I.e., single conscious moments
with no consistent continuation-- going from white noise to
white noise. This is, of course, Russell's Occam catastrophe in
a different guise. The experience of the agent is not random --
they experience conscious moments with a seemingly coherent
chain of memories giving a comprehensible history -- but there
is no reason to suppose that these memories are veridical.
They will be more or less plausible, with respect to the normal
computations, if they exist of course, but they have to exist if
computationalism is correct.
The computations underlying the conscious moment have, then, to
also compute the physics that renders the memories veridicial --
but that involves memories stretching back tens of years. An awful
lot of computations have to come together to make consciousness
that means anything. Making the probability in the sea of random
noise smaller and smaller all the time.
Sure, but probability of what and from whose point of view? Aren't you
continuing to think of this principally from a third person
perspective (actually merely an abstract "view from nowhere")? Yes,
from that impossible point of view there is no conceivable search
function that could locate the critical computational structures of
this sort; under this interpretation their measure is effectively
zero. Nevertheless we know their presence is in fact assured by
assumption. In point of fact these computations have the recursive
characteristic of exploding into an infinite fractal-like structure
of extremely high frequency (as Brent has recently put it) which would
give them in a certain and possibly critical sense a highly robust and
non-trivial structure. But the key point is that, on the basis of
Bruno's theory of computational subjectivity (again, implied by
assumption of the CTM), they must be *self-locating* from the first
person perspective. This is the key difference that would unleash the
creative subjective potential of the torrential output of the UD, as
distinct from Borges's merely alphabetical Babel which can only ever
be a zero-informational wasteland.
But how can we assess "probability" in such a context? Very
controversial point as you know. Nonetheless, Hoyle gives us an
intuitive heuristic that allows us to think of this in what is
effectively a quasi-frequentist manner (i.e. the relative subjective
frequency of "encountering" any particular momentary perspective over
any finite segment of their abstract serialisation). This heuristic
has both absolute (in the first instance) and relativising (in the
second) self-sampling characteristics. If we think of it in something
like this intuitive way (which IMO is the absolute key to the
argument) then the justification of a measure assessed in the above
manner has to lie in the direction of understanding how and why the
"organised" threads of narrative subjectivity shouldn't be effectively
swamped in a sea of subjective chaos because of competition from
"pathological" quasi-narrative fragments. I've tried to pump our
collective intuition with various analogies to suggest why this
wouldn't necessarily be the case, to supplement Bruno's more rigorous
logico-mathematical argument. Hardly conclusive of course but the
intention is principally to encourage a harder look in this direction.
The sea of pathological dross that must form the overwhelming but
fragmented majority of the "conscious potential" of UD* must somehow
be effectively suppressed from the perspective of the relatively
tiny, but mightily persistent and powerful narrative threads of
veridical consciousness (i.e. those that refer truthfully to an
externality that in turn explicates their perceptions of it, or what
we call physics). A pathway of least effort through the phase space of
possible subjectivity? Russell's solution to the possibility of an
Occam catastrophe that would sink this fragile vessel is simply to
assume that this physics is the unique requirement for its own
observation. Open problem, as Bruno would say?
David
I'm not so concerned about the measure being non-zero. I'm sure fans of
"everything" will just appeal to self-selection: the anthropic principle
applied to the UD. My point is that the computations in the threads
supporting some consistent consciousness will necessarily be computing
also a consistent physics...that there cannot be JUST conscious
thoughts. They must be embedded in a physical world, whether that world
is made of arithmetic or something else. It is this physics environment
that makes it possible to define "consistent continutation" as Bruce
notes. So then Bruno's theory doesn't seem so different from what
Tegmark and other physicists seek in a TOE. From the physics-first
perspective, he has just hyposthesized which computations that are
instantiating a physics also instantiate consciousness.
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 everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
