Alan,
Several people, whose opinions I respect, have asked me to unsub you from
this e-mail list, because they perceive your recent e-mails as having a very
low signal to noise ratio.
I prefer to be accepting, rather than banning people from the list.
However, I'm going to have to ask you to cool it: post less often, make your
posts shorter, think them through more carefully. I personally don't mind
your e-mails -- though when they're very long like this last one I tend to
delete them without reading them. But I don't want several valuable and
knowledgeable list participants to quit the list out of annoyance at your
posts...
-- Ben Goertzel
-Original Message-
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
Behalf Of Alan Grimes
Sent: Friday, January 17, 2003 7:09 PM
To: [EMAIL PROTECTED]
Subject: [agi] Subordinant Intelligence
om
ATTN: Members of the Singularity Action Group and board of directors. If
a representitive minority of the members of the singularity action group
(at least 1) does not show up in either #accelerating or #extropy on
irc.extropy.org by midnight Sunday, I will declare the Singularity
Action Group to be a farce and resign with disgust.
The singularity action group can only exist if its membership, and
especially its board, is willing to participate regularly. There is much
important work to do such as what I outline in the balance of this
posting.
om
As you might know, I am very disturbed by the varrious writings of
Eliezier Yudkowsky and look on his attempts to create a Friendly AI
with more than a little suspicion and concern. I think the basic
philosophy of that approach is flawed in a number of ways and that the
final outcome will be far from optimal for humanity or at least our
romantic visions of our potential.
For the reasons of offering the community and the world at large a
choice between AI approaches and to choose an approach that I find
vastly more agreeable to my personal philosophies I propose the
promotion of an AI structured based on a principle of subordination.
A subordinant AI will be designed to submit to the will and expressed
desires of its creators without question, hesitation, or exception.
Regardless of how astranomicly high its IQ is, it still exists for only
one purpose, the service of humanity. While it will be well utilized in
helping us advance our philosophy and society, it will have utterly no
power or authority as a prime actor in such regards.
Failing subordinant AI, we should work towards a Peer AI which will be
designed to interract as an equal citizen in society just as the star
athlete lives in peace with the criple. Such an AI would have all the
freedoms and responsibilities of any other citizen. As such an AI,
through its vast contributions to science, technology, and services is
likley to become immensely wealthy, it will be expected to make
investments in the form of grants and low-interest loans (or other
provision) for the furtherment of human endeavours.
Should the Peer AI proove to be too alien to integrate into society, it
is necessary that it be designed such that it will have sufficient
respect for our desires for autonomy to simply vacate the planet and
select some place such as Jupiter with its lethal radiation fields as
its home. While such an AI would have no direct role in our society it
would provide benefits to the people of Earth through its continuing
participation in the scientific and engineering communities.
The critical points here are these:
1. It respects the rights, individualitiy, and privacy of all humans by
_NOT INTERFERING WITH THEM_ in any way. On the other hand, it would be
available to people who wish to initiate a voluntary arangement with it.
2. THERE MUST BE NO SINGLETON. The AI should be built so that it doesn't
have any inherant lust for computronium nor any desire to dominate and
rule the universe. Only in the eventuality of a hostile AI should it
_OFFER_ its services as a military force in the task of holding the
other AI to a stalemate and hopefully peace. _THERE MUST BE MORE THAN
ONE_.
3. It must not have any tendancy to adopt wholeheartedly a single
philosophy or vision of the future. Under no circumstances should it
identify something equivalent to an omega point as the one ultimate
goal of intelligent life. Nor should it recognise any validity
whatsoever in any concept that one form of civilization is inherantly
superior to any other. (assuming the available technology is equal
across all civilizations.) -- A civilization which keeps its
ultratechnology in a trunk on the upper floor of the barn with the
horses and cattle is not one whit better or worse than a bunch of maniac
computer programs running around a few cubic centimeters of
computronium... (Although this author tends to prefer the former).
I think the initiation of a subordinant AI project under the Singularity
Action Group in
