Russell Standish wrote:
On Fri, Jun 10, 2005 at 12:22:24AM -0400, Jesse Mazer wrote:
>
> Hal didn't say anything about only sampling the birth moment randomly
> according to the absolute measure, or imply it as far as I understood
him.
>
The RSSA is applied to the "next" OM, so can only predict
probabilities of OMs that have a predecessor. A birth moment has no
predecessor by definition. Usually you would expect to apply the SSA
which applies to an absolute measure over birth moments. If you want
to apply some other principle, then you're welcome, but it would be a
little strange.
Well, an advocate of the pure RSSA approach might simply say that it's not a
well-defined question to ask what the probability of one birth moment vs.
another is, that it only makes sense to talk about the probability of
different possible successors to your current OM. Also, if you accept
absolute probabilities, conditional probabilities, and the immortality of
all streams of consciousness as I do, then it seems to me a necessary
consequence of this that streams of consciousness are neither created or
destroyed, so there's no such thing as "birth moments" in your sense. In
terms of the water tank analogy, if you want the amount of water in each
tank to stay constant over time (no change in the absolute probabilities,
since there's no time at the level of the multiverse as a whole), then if
water molecules are not destroyed they can't be created either. As I said
earlier, it seems to me if you want to believe both in immortality and the
idea that your current OM is fairly typical (as suggested by the ASSA), you
need some kind of "immortality with amnesia"--I suggested two ways that
could work in that earlier post.
>
> No, I'm not saying there is no "next" OM, my point was that the two
methods
> can give different probabilities for my next OM--for example, a Jesse
Mazer
> OM and a Russell Standish OM might have about equal absolute measure,
but
> given my current OM, a Jesse Mazer OM would have much higher relative
> measure.
Yes, of course.
OK, is that why you're saying the ASSA and RSSA are incompatible? But my
point is that I think this incompatibility is removed if you always take the
ASSA as applying to your current observer-moment, and the RSSA as applying
to your next observer-moment. This may seem like nonsense because your
"next" observer-moment will become "current" when you are experiencing it,
but I don't think it is. Remember that the information available to you
changes over time, you know nothing more than the content of your current
observer-moment--in terms of the water-tank analogy, the molecules have no
independent "memory" of where they've been in the past, they only know their
current tank (the tank itself may have 'memories' of some sort, but they
will be compatible with multiple possible past tanks). So suppose we
calculate the absolute probability of different possible OMs being my "next"
experience *without* taking into account specific knowledge of what my
current OM is, by doing a sum over the absolute probability of each OM being
my current experience multiplied by the conditional probability that that OM
will be followed by the OM whose probability of being my "next" experience
we want to calculate. If we just had 3 possible OMs labelled A, B, and C,
then we'd do:
P(C is next) = P(A is current)*P(A -> C) + P(B is current)*P(B -> C) + P(C
is current)*P(C -> C). The condition I talked about earlier that there is no
change in the absolute probabilities, that the level of water in each tank
doesn't change because the water flowing out is balanced by the water
flowing in, is equivalent to imposing the requirement that P(C is next) =
P(C is current), which means that the vector of absolute probabilities must
be an eigenvector of the matrix of conditional probabilities with eigenvalue
1:
P(A)*P(A -> A) + P(B)*P(B -> A) + P(C)*P(C -> A) = P(A)
P(A)*P(A -> B) + P(B)*P(B -> B) + P(C)*P(C -> B) = P(B)
P(A)*P(A -> C) + P(B)*P(B -> C) + P(C)*P(C -> C) = P(C)
This is the constraint that I think might help to determine a unique
self-consistent probability distributions for both absolute and conditional
probability, an idea I discussed in the "Request for a glossary of acronyms"
thread. And if you impose this constraint, it means that if you calculate
the probability of experiencing different possible OMs as your "next"
experience without specific knowledge of what your "current" experience is,
then the probability you get for having a given OM as your next experience
is exactly the same as the absolute probability of that OM being your
current experience. So this is why I think there is no incompatibility
between the ASSA and the RSSA, if you define them in the way I do.
Jesse
