[ExI] Self improvement

[Richard Loosemore]

Keith Henson wrote:
> On Fri, Apr 22, 2011 at 9:31 AM, Eugen Leitl <eugen at leitl.org> wrote:
> 
> snip
> 
>> The difference is that "friendly" is meaningless until defined.
>>
>> Please define "friendly" in formal terms, including context.
> 
> In *some* circumstances humans are friendly (cooperative) toward other humans.
> 
> There are non-zero sum game reasons rooted in evolution as to why
> humans (conditionally) act this way.
> 
> The switch between alternative human behaviors has also been under
> evolutionary selection for a long time.
> 
> I have been discussion this for years on this list and other places.
> 
> The works of Azar Gat and Steven Leblanc are critical to understanding
> this subject.

All of which may be true, but of no relevance to the question.

If evolution designs a set of motivation modules that includes a selfish 
aggression module and an empathy module, and if I then go right ahead 
and design and AGI that has an empathy module but the aggression module 
is entirely missing, with that one stroke all the evolutionary arguments 
become moot.



Richard Loosemore