I've been following the UAI discussion on "probabilistic modeling of
God" etc. for a while, being really tempted to join the discussion,
but didn't, partially since I didn't want to increase already
heavy email traffic, partially because I'm still finishing
"the very last things" in my dissertation :). However, the most
recent discussions of Hume's argument, learnability and complexity
makes this temptation impossible to resist. :)
So, here are some questions/thoughts I had in mind...
------------------------------
Kevin S. Van Horn wrote:
>... we must therefore be born with significant prior knowledge
>wired into our brains. The interesting question now is, what exactly
>is that prior knowledge with which we are born, and where does it
>come from?
But don't we already have at least two "working hypotheses"
explaining "where does it come from", such as
Hypothesis 1. evolution process - we are the way we are because we are
ALIVE now - those branches of evolution that didn't fit into
the environment died out; others developed into some other
"equilibrium states" with different properties, such as
bacteria etc. - and we are also just a particular
"equilibrium state" having particular properties in that
complex system called "nature".
Hypothesis 2. God's creation
(this hypothesis was already discussed a lot :)
Personally, I prefer hypothsis 1, especially since I'm sceptical
about the accepted definition of "God" as "a superior being" rather
than "the laws of nature" (but I would accept the latter
definition of "God=laws of nature" - well, then it's only
the difference in terminology :).
Now, as Kathryn Laskey said:
>...I conclude that it is possible to engage in scienfifically
>meaninful discourse about what kinds of evidence would lead one
>to conclude that God does or does not exist, and if God exists,
>what God is like.
My personal opinion is that no evidence is enough to conclude that
(well, sounds like pure agnosticism). We learn "conditional
probabilities" from data given evidence and given our prior beliefs
(e.g. God versus "laws of nature"). Chaging priors gives you another
conditional probability distribution (maybe), but if both models fit
data equally well - then you have no way of distinguishing between
the two hypothesis. For example, you may believe in conversation
with God which "really" happened, or you may believe in some
neurochemistry in your brain - if you get same results as being
happy and productive, does it really matter what "really" happened,
if the results are the same? Well, you may say that "you like"
the first hypothesis better - and choose to believe it.
But that could be also a neurochemical state of your brain that
makes you "like certain thing" :)
Best regards,
Irina Rish
