From: Ben Houston (ben@exocortex.org)
Date: Tue Jun 18 2002 - 13:53:14 MDT
Dani Eder [danielravennest@yahoo.com] wrote:
> It is my opinion that such a value system would
> be easy for an AI to use, as values would be
> measured in bits.
I think that "value" is a very grey area and to attempt to simplify it
to a single dimensional scale of "information content" is fairly
incorrect. My DVD of absolute random information will thus be worth
more than data representing the human genome because theoretically
according to Claude Shannon my DVD has more information.
Dani Eder [danielravennest@yahoo.com] wrote:
> It is also my opinion that
> such a value system could be 'tuned' to be in
> accord with human values. This would be done
> by asking humans to make value comparisons, and then
> adjusting relative values of different types
> of information (genetic, neuronal, computer-stored)
> to match the human results.
The addition of an undefined and open ended "tuning" procedure to the
first measure could easily result in a measure that is completely
different. I would argue then that any initial measure to which such a
"tuning" procedure was applied could easily be equal to the one you are
proposing.
Dani Eder [danielravennest@yahoo.com] wrote:
> For example, it is easy to value a bus
> full of people more than the life of one
> suicide bomber, and thus decide kill the
> bomber.
The situation over there is truly horrible.
Dani Eder [danielravennest@yahoo.com] wrote:
> Now consider the question of a suicide bomber
> who has infiltrated the Louvre museum. Do you
> kill a human to save a lot of unique artwork?
I believe that deadly force can only be used according to US law if the
individual can potentially cause physical harm to either the officers
involved or others and deadly force is necessary to apprehend him. I
believe that such requirements would probably be met through various
circumstances if in fact a suicidal vandal infiltrated the Louvre and
could not be stopped in a non-lethal way.
Cheers,
-ben
http://www.exocortex.org
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT