From: Anders Sandberg (asa@nada.kth.se)
Date: Thu Jun 07 2001 - 09:28:51 MDT
torsdagen den 7 juni 2001 00:37 Francois-Rene Rideau wrote:
> >: Ben Goertzel, "RE: vegetarianism and transhumanism"
> >
> > Animal welfare is a branch of ethics, i.e. there is no given right or
> > wrong answer. The extropian list isn't really the place to debate it
> > IMHO.
>
> I question the claim that ethical problems can have no answer;
> this claim is the relativist claim of barbary itself:
> "nothing is more justified than anything else".
>
> Ethics can definitely provide answers to some (obviously not all) problems.
> Libertarians are people who take ethics very seriously,
> and thus so do self-respecting extropians.
>
> Without ethics, you cannot be transhuman. You're not even human.
As I see it, ethics is really part of your policy function (I look at it from
a reinforcement learning perspective). In any given situation you want to be
able to do what provides the best outcome given your values, and for this you
use your policy function to decide what to do. This function is then updated
by experience (and possibly imagining future scenarios) so that you maximise
your value.
Obviously, it is very hard to create a perfect policy function that can deal
with situations you have never experienced. But good policy functions
generalise well: from the cases you have experienced (or heard about) you can
estimate how good and bad different actions are. Ethics is in my opinion the
general "backbone" of the function, the general principles which you use when
there are no compelling special cases (Robin had a similar idea in about
ethical intuitions vs. general principles in one of his papers).
Without efficient, generalisable and consistent policy functions life becomes
so impractical that it doesn't work. It is not black and white, you could
probably get by using a highly irregular and inconsistent policy function,
but it would not be as efficient for achiving your goals as having a function
more in tune with reality. On the other hand, you cannot survive well by just
using general principles, there is a certain amount of special cases and
exceptions to take into account when living successfully. The big problem is
finding the right level of regularisation. It is computationallye expensive
to approximate (i.e. you need a lot of experience) and (due to its
uncomputable nature in the case of a complex world like ours) impossible to
give error bounds on. Hence, a lot of disagreement.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:08:00 MST