From: Phillip Huggan (email@example.com)
Date: Fri Jan 27 2006 - 10:36:35 MST
For starters, all the really interesting AGI discussion points cannot be breached on this public list. SIAI considers it risky to open-source the project and many if not all of the other efforts are structured commercially, so there are IP concern in giving away too many goodies.
Consciousness discussions are relevant because we really don't know yet if Strong-AI or "substrate-ism" is true yet. If Strong-AI works, it still might be easier to go for a system that mimics intelligent behaviour rather than shoot for consciousness. If strong AI assumptions don't turn out to be true, it's nice to empart the idea to AGI engineers or engineers to be that it is still possible to use an AI system to fulfill substrate-isms requirements for actual consciousness or for intelligent behaviour.
Regarding skepticism and ethical dillemnas, AGI goal-systems are the part that actually determines whether the AGI in question kills off the human race or not. It has been noted our ethics and pychologies are evolutionary in origin, and this observation is used to justify study of the subject in looking for deadends and clues to follow towards AGI. But disturbingly, some sl4 posts suggest the AGI should use this ethics too?! If the programmers who code the actual AGI goal systems don't understand ethics very well, the AGI will kill us all. Replace the term "gazelle" with *post-singularity human*, and replace "lion" with *post-singularity enhanced human* and the discussion in a previous thread jumps up the SL-ranking.
Philip Goetz <firstname.lastname@example.org> wrote:
> >- How are we supposed to talk about what sort of goals and ethics we
> >should develop an AI to have, so that it will be friendly, without
> >delving into "mere philosophy"? <SNIP>
Well, I'm willing to give up that discussion if it means we can clear the
list of discussions of consciousness, skepticism, and SL0+1 ethical
What are the most popular cars? Find out at Yahoo! Autos
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:55 MDT