At 22:19 -0700 8/07/2002, Wei Dai wrote:
>Bruno, unfortunately I'm not able to write very quickly. I hope you don't
>mind if I just respond to some of the topics I think are most important.
>Feel free to recall the other ones later if you think I'm missing
>On Mon, Jul 08, 2002 at 05:22:11PM +0200, Bruno Marchal wrote:
>> I give more than a motivation, I suggest (at least) an obligation!
>> First here, I could ask you how you define a universe. That's not obvious
>> at all.
>I'm not sure what the right way to define a universe is yet. For example I
>don't know if the definition should allow uncomputable universes, and if
>so how. But I have at least two starting points, Schmidhuber's and
>Tegmark's definitions, both of which while flawed at least make sense to
>me. I am still unable to really understand yours.
Have you understand what I call the invariance lemma? (The fact that
if I am multiplied into 1000 copies in a whole space-time decor, real,
virtual or arithmetical) the measure of comp indeterminacy depends
only on "1000").
Perhaps the trouble is that you are not really aware of the mind-body
I don't yet understand how Schmidhuber attach the mind to its computa-
>First I question the necessity of defining that average.
This is where the list splits in two. Those aware of a measure problem
and the other ... It is linked to our all debate between RSSA and ASSA
(Relative Self-Sampling Assumption/Absolute SSA).
>Suppose we just ignored the issue of first person indeterminancy, and
>made no reference to
>it at all. Speak only in third-person terms about the future. Why can't we
>just do that?
Because we want to understand the nature of reality and where does it comes
from. Consciousness is part of the data we must explain, physicals laws
also. With comp the laws emerge from the relation between numbers, but as
seen from inside, so that I don't see how to avoid the modal distinctions in
a search toward a toe. I think you fail to appreciate the "proof" character
of the uda.
>Second, why do you restrict yourself to interviewing sound machines? I'm
>sure I'm not a sound machine, yet I am an conscious observer. So I'm
>confused. Why do you ignore most of the conscious observers who are not
I'm not sure you are not a sound machine especially when proving
things on numbers. In the worst case take my "restriction" as a
simplifying assumption. But I don't think there is any restriction here.
The amazing fact is that the sound machine has (through the Z logics) an
amazingly large non-monotonical layer. Remember that "inconsistency" has
been shown consistent in Peano Arithmetic, ZF theory, etc. Sound machines,
sound at the basic level I interrogate them, can be consistently very unsound
once entangled to deep computational histories. I am taking the full nuance
given by the second incompleteness theorem into account here.
You would be really unsound, through my basic use of the term, only if you
were able to give a finite proof of a false proposition, not just pretend
you could find it.
>When I first thought about the issue of duplication, I was on a similar
>track - there's first person indeterminancy, so let's figure out how to
>quantify it. But then I realized it's not necessary, because we can just
>reason and make decisions in third person terms. Why complicate things
>beyond what's needed?
Because with comp the relation between our first person inspired decisions
and third person (or first person plural) realities remains to be explained.
The comp indeterminacy, by the invariance lemma, pertains on the whole
of UD* (the computationalist form of everything).
Of course you can forget comp, postulate a reality (defined by what you
see and expect to see) and take your decisions. But we are looking for
a toe, not a recipe for life. You don't need quantum gravity for everyday
decision do you? But you can imagine quantum gravity being related to the
search of a TOE, ok? Well, what I try to say is that if you take seriously
the hypothesis that our private experience are invariant for functional
substitution at some level, then the utimate explanation of quantum gravity
is accessible by UTMS pure introspection, and that the toe is a mixture
of machine's machine psychology (G), and machine psychology (G*).
The advantage of my way is that it gives an explanation for the origin
of physical laws and at the same time of physical sensations. It has no direct
use in decision theory, except perhaps by predicting new phenomena, perhaps
exploitable, like any new theory.
No doubt I make simplifications here and there, but what remains is
very complex and unknown. I would be glad if someone find a flaw or some
implicit hypothesis I use unconsciously ...
About Schmidhuber's toe, perhaps you could try to explain me how you keep
being in the same computation in front of the invariance lemma, knowing
that all computations exist.