On Mon, 13 Sep 1999 GSLevy.domain.name.hidden wrote:
> In a message dated 99-09-12 22:42:10 EDT, Jacques Mallah write:
> << So let me talk about what does make sense, the ASSA. Assume that
> most people are Chinese. If you are a person that doesn't know his
> nationality, you would be justified in guessing that you are probably
> Chinese, based on the ASSA that your observation is drawn from a set with
> uniform weighting per capita. If everyone made this guess, most of them
> would be right, and of course some of them would be wrong unless the
> probability was 100%.
> >>
>
> Actually there are more mice than chimese. Therefore using Jacques'
> absolutist argument it makes pefectly good sense to expect to be a mouse. Or
> whatever... There are more ants than mice... therefore we should be an ant...
> or there are more sentient aliens with five appendages on alpha centauri
> therefore we should be a sentient alien with five appendages....
If you didn't know your species that is correct.. However, if you
are even asking the question, the conditional probability that you are a
mouse or an ant given that you are asking the question is zero. The
Centaurian case is (with some assumptions) just like the Chinese case.
> It seems that
> somehow we are missing the boat by not framing the question according to our
> own frame of reference... which would make the issue a relative issue.
No, if you did that you would miss the boat. The boat that allows
predictions of stuff like the observed laws of physics.
> One way out for Jacques is to assume that humans are the only sentient
> creatures in the whole universe... actually the whole plenitude.
Why did you even write such bullshit and try to tar me with it?
> Talk about an anti-Copernican, anthropocentric point of view!
Then I would have to talk about your view.
I think it's time for a little review of basics. To judge the
effect of a piece of evidence we start with a prior Bayesian probability
distribution as a function of different models. We update the probability
distribution by taking the new probability to be proportional to the old
probability times the conditional probability of getting the new
observation if that model were true.
A piece of evidence is surprising if it would cause a big shift in
our Bayesian probability distribution. This is relative, of course.
Usually a coincidence is surprising. For example if you told me x is a
randomly selected real number between 1 and 10, and I noticed x=pi, I
would suspect that your random number generator had a nonuniform
probability distribution.
OK, getting back to the ASSA vs. the RSSA. Given that you are not
Chinese. (Perhaps Wei Dai might have to sit this one out.) First of all
we can't make any comparison at all on this basis unless the RSSA makes a
prediction of the conditional probability that you are Chinese if given
that the RSSA is true. It's like comparing apples with raisins and
saying "this apple is too big." One must compare apples with apples.
- - - - - - -
Jacques Mallah (jqm1584.domain.name.hidden)
Graduate Student / Many Worlder / Devil's Advocate
"I know what no one else knows" - 'Runaway Train', Soul Asylum
My URL:
http://pages.nyu.edu/~jqm1584/
Received on Mon Sep 13 1999 - 13:05:24 PDT