Question about limited knowledge

From: Tennessee Leeuwenburg (tennessee@tennessee.id.au)
Date: Wed Mar 23 2005 - 18:17:49 MST


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

This is a paraphrase of something I heard asked of Peter Singer, a
moderately well-known consequentialist philosopher.

"In order to be a good utilitarian, do you have to have a good
imagination, in order to have a proper appreciation of the alternatives".

I thought this might have some implications for AI and theory of
knowledge.

If we assume that all beliefs are imprecise, then *even if* we accept
some form of Bayesian calculation as the "best we can do" and that
it's not bad, we still have something left over. Let me introduce some
terminology from Phenomenology :

Immanent : What appears to us as phenomena because of reality
Transcendent : What is true about reality

Some immanent phenomena are also transcendent, but there are some
things which are transendent things which we can never know.

Now, let's assume that physics is a remarkably accurate and reliable
model of what is immanent, and as such we can say that a perfect model
is as close to true knowledge of the transcendent as one needs to get,
because one will never be wrong. If physics is a good approximation
(i.e. very high reliability of prediction of immanent phenomena), then
that's great.

Using that framework, there must be some ways in which we can (a)
establish the statistical reliability of any particular model in
predicting immanent events and (b) some understanding of how much
truth much be transcendant if a particular level of imprecision is
inevitable.

Let's say we know all the immanent factors affecting a toss of a coin,
but we still don't get 100% accuracy. Let's say we have 98% accuracy.
The remaining 2% must lie in transcendant, unknowable and
un-modellable aspects of reality. Perhaps reality is truly "random" or
perhaps it is merely un-knowable, but because it's purely
transcendant, we can't know which is "true".

Considering this in the context of an artificial intelligence, do you
think this affects decision making?

I'm sorry to clog the list with what is merely a question, but as
people who are interested in decision-making, I thought someone here
might have an opinion.

Cheers,
- -Tennessee
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFCQhU9Fp/Peux6TnIRAlunAJwKRlzBeOuXsKtlnqTjJca2+GDbDQCdFgup
d+xG9kXcGfc3CN2rakcRnZM=
=do8L
-----END PGP SIGNATURE-----



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:51 MDT