Hi Abram, Him Kim,
Kim, while answering Abram, I realised I was doing the KIM 2.3, you
can read it before KIM 2.2 without problem I think, in any case tell
me if you have follow the argument. I don't answer the questions, so
you or Abram, or anyone else can answer.
Abram, The answer to your post is really the step 3 of the UDA
reasoning. It is the justification of the first person indeterminacy,
and the definition of (relatively) normal machine.
On 20 Dec 2008, at 04:46, Abram Demski wrote:
>
> Bruno,
>
> From what assumptions could a probability ultimately be derived?
From the assumption that when I do an experience or an experiment, I
will observe a result.
And from the hope I will be able to interpret that result in my or our
favorite current theory from which I can *deduce* the probability laws.
This is akin to a self-consistency assumption.
> It
> seems that a coherent theory of the probability of future events is
> needed (otherwise the passing of time could be white noise), but I do
> not see where such probabilities could come out of more basic
> assumptions.
UDA is a non constructive proof that in the MEC theory, we have to
derive the probabilities from the discourse of the "normal machine",
which I will define below (anticipating on the KIM 2 thread).
AUDA is a path toward a constructive derivation of the probability
laws. The basic idea is simple: let us ask the question directly to
the universal machine.
In QM, without collapse, Everett (+ Gleason theorem) has convinced me
that
1. There is no probabilities in the theory.
2. Quantum and classical probabilities are justified in the "normal"
self-observing machines by the SWE only.
But there is a "hic". A "little" problem.
That derivation assumes MEC (or weakenings). And MEC forces the
probabilities to be derived from all type of computations, no way to
chose a particular universal machine at the start, any must do. This
is really what UDA shows. The good news is that such an extraction can
then justify both the quanta and the qualia. Quanta are (should be
here) particular case of (sharable) qualia.
> To reason about the future, we assume that we are in a
> randomly chosen computation--
Right now I don't feel like being on a randomly chosen computation.
"I" "belong(s)" on all computations which have reached my actual
state(s) (singular for the 1-state, and plural for the 3-states, or
the 1-plural states, see below).
My next state will be chosen partially randomly among many consistent
continuations.
> but then we are already using some
> probability distribution.
At some level it is the Gaussian distribution. See the definition of
the "normal machine" below.
>
>
> Evolution is at the root of our ability to predict probabilistically.
> We use one probability distribution over another because it helps us
> survive. However, this is not good enough of an answer in the
> multiverse: every possible form survives anyway.
Once you bet on everything you have to accept also, among many
realities, those who does not "survive", the cul-de-sac.
At the level of reasoning in comp this is equivalent with a "self-
consistency" assumption, which is implicit in teleportation experiment.
The multiverse idea is not so different from the Darwinian idea of
"all the species", of course restricted to the relatively consistent
one.
Relatively to what? Relatively to their most probable histories/
computation.
Below, to define the notion of "normal machine", I will construct a
case where in a sense "every possible form survives". despite this
fact, the probabilities will emerge clearly once we distinguish the
third person discourses from the first person discourses.
> To keep talking about
> evolution, we would need to talk about which forms are more common in
> the multiverse. But to count how common forms are, we would need some
> measure over the multiverse, which could give us a probability
> distribution in the first place. So every possible argument seems
> circular.
I don't think so. See below.
>
>
> So, I don't need all the details of your derivation of a probability
> (though I'm interested); but what assumptions can you get a
> probability distribution from?
From the assumption of mechanism. Digital mechanism. The assumption
that you survive with a relative probability 1 in case of scanning,
annihilation and reconstitution done at some level.
Then you are duplicable. I can "cut" you and "paste" you in two
identical rooms, except that I have put a "one" in a closed box in one
room, and I have put a "zero" in a similar and similarly disposed box,
in the other room. Ah! To help for some probable cutting the air
argument, I put a cup of delicious coffee near the boxes in the two
rooms. All right?
(If you don't mind I will of course assume you really want a cup of
coffee, at the moment of the experience).
Now the third person discourses is the content of a diary (or the
memory) of an observer which, typically does not enter the
teleportation boxes. The first person discourse(s) is (are) any
content of the diary (diaries) of those who does enter the
teleportation boxes, and get out of reconstitution boxes. Note that
first and third person can develop discourses about third and first
person discourses.
Here is the protocol of the experiment, which is told to you in
advance. You will be scanned, annihilated and reconstituted
simultaneously in the two rooms. And I am asking you now some question:
Assuming MEC (that is "probability 1" in case of simple teleportation).
What is the probability that you will survive this duplication
experiment/experience(s) ?
What is the probability that you will drink a cup of coffee in that
experiment/experience(s) ?
What is the probability that you will drink a cup of coffee and think
"OK I have survived, that coffee is good, I don't know what number
belongs to that box, obviously it cannot be both one and zero!"
What is the probability that you will write in your diary "1) I have
survived, 2) the coffee is good, 3) the number in the box is zero.
I don't want to give a definition of what is a normal machine, just
that you feel the point.
Let me give you another illustration which exploits the freedom of
thought a bit more. Indeed, let me duplicate you, or better,
polyplicate you into 2^(16180*10000)*(60*90)*24.
I explain. I multiply you by 2^(16180*10000) in front of a
(16180*10000) pixels screen, with each possible images (black and
white) on it.
And I reiterate every 1/24 of a second that multiplication, and this
during 90 minutes, that is 90*60 seconds.
What do you predict you will feel, as personal, subjective, first
person experience. What do you think is more probable, among:
I will feel seeing a white screen
I will feel seeing a black screen
I will feel seeing a movie
I will feel seeing a good movie
I will feel seeing "2001 Space Odyssey"
I will feel seeing "2001 Space Odyssey" with the subtitle of Caligula
I will feel seeing a random-noise-movie (what you see on TV when there
is no emission).
Of course there is a sense to say: I will see all possible movies, but
this means you are talking about yourself at the third person point of
view, and here what is asked, is what do you expect to experience or
live in your future if you are invited to practice it. By MEC, you
survive, and any of your first person experience is unique, on which
of the above you would bet? What bet will you do if I tell you that I
will accompany you in the multiplication. And send you to hell if your
bet is wrong. What bet you will do if you want to optimize the chance
of not going to hell?
The rest of the UDA reasoning shows that "this" thought experience
happens all the time in arithmetic, but the "probabilities", which
eventually could be credibilities or other uncertainty measure, are
constrained by computer science/number theory. It extends the notion
of normality from the protocol above to the whole Universal Deployment
(which I will (re)define in KIM 3).
Exercise: define the notion of "normal first person experience" for
machine in the protocol above. With the UD protocol, things are so
much complex that I will interview the Universal Machine directly to
provide hints ...
Bruno
http://iridia.ulb.ac.be/~marchal/
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list.domain.name.hidden
To unsubscribe from this group, send email to everything-list+unsubscribe.domain.name.hidden
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---
Received on Sun Dec 21 2008 - 13:45:14 PST