Re: superintelligence and ethical egoism

From: Gordon Worley (
Date: Sat Jun 02 2001 - 08:33:03 MDT

At 10:04 PM -0700 6/1/01, Samantha Atkins wrote:
>Gordon Worley wrote:
> > Here is another thing: an egoistic AI would only be interested in
>> other individuals, not species. Ve could care less what happens to
>> humans, be ve might take an interest in a few humans and see to it
>> that they are uploaded before all of the nearby space is turned into
> > computronium.
>On what grounds would an AI be only interested in individuals?
>I would think it would take some interest in the fate of the
>species that brought it into being.

This is because egoism is a philosophy based on the idea that all
action and thought comes from the individual, not a group. An
egoistic AI would, therefore, only be interested in the actions and
thoughts of indivudual humans, not the entire species, because that's
how ve thinks. An altruistic AI (in the traditional sense, not the
Friendly sense) would conversely be interested in helping the largest
numbers possible (i.e. the whole species of several species). An
egoistic AI would fully encompass the philosophy of egoism (and
there's no good reason why ve wouldn't be the ideal egoist short of
bad code) and make all decisions based on the goal of being egositic
as the ultimate goal, just as a Friendly AI makes all decisions based
on the ultimate goal of being Friendly (well, there seems to be some
dispute about this, but if you want just replace the previous example
with whatever equivilant model you think will work best :^)).

To respond to your thought, an egoistic AI would take an interest in
the *individuals* that did the ground work to make ver, but not in
the species as a whole (other than maybe trying to learn the
species's anatomy, languages, etc. as a side effect of being
interested in some individuals, but only insofar as this is necessary
to interact with the individuals).

Gordon Worley
PGP Fingerprint:  C462 FA84 B811 3501 9010  20D2 6EF3 77F7 BBD3 B003

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT