Re: on formally indescribable merde

From: George Levy <>
Date: Tue, 13 Mar 2001 20:08:06 -0800

Marchal wrote:

> But COMP implies MWI (Note that Schmidhuber and me agree on that,
> >> but we disagree on what *are* the (many) worlds MW).
> More precisely: COMP implies MWI in two senses.
> 1) Everett's sense: SE + COMP gives MW.

He really meant SE + No( privileged observer status) -> MW. One could argue then
that No(privileged observer status) is implied by COMP.

> 2) COMP sense: COMP gives SE + MW (my thesis, look at my CC&Q paper)

I agree that COMP implies first person indeterminacy but, as far as I know, it
does not predict Schoedinger Equation in all its splendor, including Planck's
constant which is a parameter in this equation. So would it be more accurate to
say COMP implies Indeterminacy + MW?

> [BM] I pretend that with comp a world is a
> first person (plural) construct. [Schmidhuber] does not defined them but he
> associates
> them with their generating programs. Observers belongs to worlds, with
> Schmidhuber. I pretend with comp that observers's infered worlds are
> defined by the set of the consistent extensions.

I assume "consistent extensions" means "logically consistent extensions of the
observers". In which case we agree fully on the meaning of worlds. Schmidhuber,
just like Mallah, is stuck with the third person perspective.

> [GL] >Again to clarify... the word "machine" means that consciousness arises
> from
> >simple (Turing-like) computations. This is the COMP hypothesis.
> [BM] In first approximation YES. In second approximation NOT REALLY.
> By comp I mean I survive with an artificial digital brain/body/universe.
> As a counter-intuitive conclusion, consciousness does not really
> supervenes on a computation but on an infinite cloud of "similar"
> computations existing in UD*.

I have reached almost the same conclusion, that our consciousness come about from
an ensemble of more or less identical "points" or states in the plenitude and the
"thickness" of this ensemble is a measure of the Heisenberg uncertainty. The
difference is that you call them "computation." I view them more as instantaneous
static entities which are logically connected to each other. Maybe we could
resolve this issue by saying that I focus on the points of the graph and you, on
the links :-)

> [BM] Locally a brain/body/universe only makes it
> possible for a person (the one conscious) to accelerate himself
> relatively to its most probable possible extensions. Note that this gives
> a role to consciousness : self-speeding up abilities.
> And this is linked to another result by Godel. If you add an undecidable
> true statement to the theory (in which that sentence is undecidable), not
> only an infinity of new formula become decidable, but an infinity of
> provable formula get shorter proofs.

You are accelerating too fast for me... I don't understand this at all.

> >I am not sure I understand "shared computational histories". Why would past
> >computational states be relevant? A current state could be reached from
> >different past points (OMs) unless "merging" is not allowed
> Merging is allowed through amnesia. In some sense personal memories
> help you to stay into no merging histories.

I smell a whiff of third person thinking. I'll say something, then I'll retract
it because I just don't have the words to say it straight. An observer in world A
who has a "false" memory (of something that did not happen in world A), is in
the same mental state as an observer in world B who has a true memory ( of
something that happended in world B). The two observers are in exactly identical
states, even though their "shared histories" are different. The point is that
their mental states are the same, they have the same consciousness, they are
really the same observer and they are really in the same world.
Now I can retract what I said. There is no world A and B and there is no shared
history. The important thing is only the current state of the observer(s). The
extensions to the observer are fuzzy.

> [GL] >...Isn't the current state only of relevance?
> Relevance with respect to what? A state is not enough, you need a
> universal machine to support it.

Precisely. With respect to us!. We are the's a vicious circle...we
are self emergent...Not only is our world anthropically defined, but we ourselves
also are....

> It is more "sharing a common history" like the bifurcation W and M.
> Biological multiplication gives a simple model of tree like
> developpement where individuals share a long common history. The more you
> leave the leaves (!) the more the histories are shared. This is an image
> because both with comp and/or QM, we must take into account merging.
> (I know you agree with that).
> >From inside UD* (i.e. from the average first person point of view of
> machines) I make the conjecture that there is no ultimate well-defined
> trunk for these barnches and leaves. From the third person view there
> is one which is just the trivial one: UD.

Your use of the words "shared histories" partially seem to carry some old
fashioned baggage from the days when time was thought to be linear. Yet you do
accept the concept of bifurcation and merging...

Can conventional mechanics support consciousness spliting and merging? I don't
think so. Here is why.
There are two cases to consider: state machines in the continuum and state
machines in discrete space.

 In phase space with the cardinality of the continuum, each point has a unique
past and a unique future. Another way to say this is that if we could make state
machine with continuous states, each machine state would have a unique antecedent
and a unique future. Except at singularities, no crossing is allowed in continuous
phase space... transition lines could get arbitrairly close to each other but
could never reach each other. .I would look at state machines transitions more
like a fluid flow. Singularities could have no future because any future would be
indeterminate. Given their indeterminate future, their past would also be
questionable if we are willing to accept time reversal.

Phase space transitions in discrete space, however, can merge. Once merged, they
stay merged forever. Splitting is not allowed. State machines can reach the same
state from different states. Once in similar state they stay similar.
otherwise, it would generate indeterminacy.

The point is that merging seems to be possible but splitting is difficult. You
could have splitting of consciousness, if, instead of considering consciousness as
a single point represented by definite states, you are willing to consider a fuzzy
region comprised of a multitude of points. Then any transitions out of this region
would be "splitting" This type of consciousness fuzzy region) would make
indeterminacy MANDATORY! In practice the existence of this fuzzy region is hinted
at by Planck's constant.

> [GL] >Why the word "projection?"
> [BM] Or "common anticipation". Like F=ma, or SE, ... We take it as a trunk,
> when it is perhaps just a branch. See above.

Ok. I call this, "common frame of reference."

Received on Tue Mar 13 2001 - 20:14:51 PST

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:07 PST