From: Russell Wallace (russell.wallace@gmail.com)
Date: Thu Nov 10 2005 - 06:59:46 MST
Awhile ago I was arguing with Eliezer over IRC on the topic of extrapolated
volition and moral axioms, where I took the position that you can't
extrapolate or deduce past a certain point; I suggested my criteria of
"safety, diversity, fairness" as the best starting point.
I've realized in the last few days that I was mistaken. While safety,
diversity and fairness are good heuristics - I think it likely that a good
strategy will need to respect them - they don't have to be taken as axioms;
there is a deeper level (though I don't at this time have a definition of
it) of which they can be regarded as derived approximations. (You do still
have to start _somewhere_, of course - moral values aren't logical
consequences of "A or not A" - but you don't have to start from where I
thought you did.)
This in turn invalidates my earlier criticism of the extrapolated volition
part of CEV as "chasing the ghost of universe-dictated morality", which I
now withdraw. (My problems with the _collective_ part of CEV still stand,
however.)
- Russell
This archive was generated by hypermail 2.1.5 : Tue Feb 21 2006 - 04:23:19 MST