Re: White Rabbits and Algorithmic TOEs

From: Marchal <>
Date: Sat Jan 20 05:55:47 2001

Russell Standish wrote:

>Last night I finally got around to reading Bruno's CCQ paper and
>Juergen's Algo TOE paper. Thanks for the citation guys.
>Bruno, its a bit unfair that you lump me in with Schmidhuber in your
>criticism in footnote 9. (I agree your criticism is fair of
>Schmidhuber's approach BTW). It indicates that you missed the point of
>section 3 of my Occam paper - are you using the second (latest)
>version of this paper? I substantially revised sections 3 & 4.

oh ! Sorry, I will load it again.

>I do not assume we are inhabiting a well defined universe amongst all
>possible - quite the opposite in fact. The UTM used for defining the
>Universal Prior is that of the observer erself. In fact, one doesn't
>need a UTM at all - all that suffices is an observer that equivalences
>descriptions. The measure of a description is simply the ratios of
>equivalence class sizes. (see my new Complexity and Emergence paper
>for my point on effective complexity).

Equivalence description of what ? I guess you explain that in the
new Complexity and Emergence paper. You don't need an UTM ?

>In section 3, the uncountable
>random universes are equivalenced to their indistinguishable law-like
>universe. White Rabbit universes, also have their own cloud of random
>variations, but the whole cloud must have lower measure.

I "feel" that correct ...

>I don't deliberately ignore the distinction between 1st and 3rd person
>viewpoints - in "Occams razor", I'm only dealing with 1st person
>viewpoints. I'm ignoring the possibility of whether there even is a
>3rd person viewpoint.

But for communication, I suppose you agree that you are using 3rd person
description (implicitely).

>Incidently, in your "Hunting the White Rabbit (I)", I like your use of
>the Feynman path metaphor. The destructive interference of long
>Feynman paths in some sense corresponds to the way that random strings
>are equivalenced by an observer.

If you prove that, then you deserve a prize. My feeling (unfortunately
it's just a feeling) is that the 1-person point of view obliges us to
take Levy Characteristic function (in probability theory) more
seriously. It is at this point that Fourier transform could be used
to link information and computation .... I am just dreaming aloud ...

>Juergen's paper makes many interesting contributions into different
>types of measures, definitely a step forward. I would criticise it on
>two grounds: a) assumption that the universe is running on a
>particular machine, ala my Occam's razor critique, and Bruno's
>critique, and b) the assumption of COMP - ie uncomputable objects such
>as random numbers are unecessary for consciousness.

Hola ! Remember that I mean by COMP the fact of surviving with an
artificial digital brain. And then ... I show that uncomputable objects
are necessary or at least unavoidable ... This is the main point where
Schmidhuber and me disagree. (This is also the main consequence of the
UDA thought experiment). The probabilities are defined on infinite
(continuous) set of infinite histories.

>This assumption
>leads to specific predictions, such as beta decay having hidden pseudo
>random correllations, and (forgive if I'm wrong) the prediction that
>quantum computing will fail.
>Note that in my Occam paper, I explicity assume that randomness is
>needed for consciousness, and that the mechanism for producing it
>(with respectable measure) is to embed the observer in a
>Multiverse. This is why I end up with conventional QM. But then, I
>have always been critical of COMP.
>Could Juergen's work lead to falsifiability of COMP?

I am not sure. (Juergen would also doubt that).
But, actually, my work *could* lead to the falsifiability of COMP.
It is that point most people appreciated in my reasoning (for
exemple Gilles Henri's post in this list!).
But concluding that comp is false from my work is well to early,
IMO, and honestly I would not bet on that.

Received on Sat Jan 20 2001 - 05:55:47 PST

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