Le 07-nov.-05, à 08:29, Hal Finney writes:
> Bruno writes:
>> OK. But the word "universe" can be misleading here. It is probably
>> less
>> misleading to say that the Universal Dovetailer generates all
>> computations. By assuming comp, this generates also all the (first
>> person) observer-moments (states/worlds/...).
>> The physical reality will emerge from that, but there is no a priori
>> reason to believe the UD generates any particular physical reality,
>> although we have empirical reason that some quantum dovetailer will
>> win
>> the "measure" battle.
>
> Isn't it hard, even assuming comp, to know whether a particular
> computation corresponds to a particular first-person observer-moment?
It is hard. It is even impossible; even provably impossible if you
assume you are a consistent machine.
> Comp says that I am a computation, at some level of abstraction;
> but having faith in that principle will not tell me whether a given
> computation implements me. How can I bridge that gap?
By betting (but that is not knowing). You cannot bridge that gap by any
effective way. It is almost a private" question which concerns you and
your possible doctor.
This shows also that comp does not provide an effective direct way to
make exact predictions (quite like non relativistic quantum field
theory, where an exact prediction would need an exact computation
depending on 2^aleph_0 histories).
But the nice thing is that it is possible to derive a (quantum) logic
from the particular case of the "measure one" on the consistent
extensions.
>
>> If that means that my probable future, when I am in a comp state S,
>> is
>> entirely determined by the collection of computations going through S,
>> with "intrinsical weight" determined by the UD (and thus by
>> theoretical
>> computer science alone), then OK.
>
> Right, and the same question applies. To know if a given computation
> represents one of my first-person probable futures, I have to know
> quite
> a bit.
Even knowing your comp state exactly (which is impossible except in
some lucky betting way) you would still be unable to unravel the
collection of computations going through your states). remember that
the problem of determining if two programs does the same computation
(with some notion of "equality" on computations) is not effectively
solvable. Look at "Rice theorem" in the Cutland's book, or on Google.
All that is explained with all details in my long work (in french,
sorry). I can prove it if you insist ;). You could try to prove it from
the diagonalization posts.
> I need to know how to go from a computation to a first-person
> experience;
With the comp hyp, this is given by the comp first person
indeterminacy, like the notion of "memory machine" is enough to deduce
the appearances of an indeterministic collapse for an observer
described by the SWE (Everett).
> and I need to know details of my own first-person experiences
> so that I can judge whether a computation "matches" my experiences.
>
> That second part is obvious, I guess; I can be assumed to be aware of
> my own experiences.
Even this is probably impossible given the possibility of amnesia.
> But the first part is what is hard, looking at a
> computation and deciding what kind of mind it creates.
Impossible.
> Do your theories
> offer insights into this hard part?
I think so. By explaining why, assuming the comp hyp, those hard parts
need to be hard and actually impossible.
But this has nothing to do with the fact that comp makes the physical
laws emerging globally from all computations. Once you derived the laws
of physics you can apply them, or apply approximations of them, like we
can apply the laws of physics derived empirically. We just cannot prove
them, for we cannot prove comp, unless we are inconsistent.
Bruno
http://iridia.ulb.ac.be/~marchal/
Received on Tue Nov 08 2005 - 05:30:09 PST