RE: Different View of IA

From: Ben Goertzel (ben@goertzel.org)
Date: Tue Apr 23 2002 - 07:37:54 MDT


> > Having read some of the archives this list seems very hostile
> to any memes that try to suggest any other scenario than Friendly
> Ai/sysop.

This posited hostility seems to me not to exist.

For example, I am nowhere near as confident as Eliezer that either Friendly
AI or Unfriendly AI is a reasonably likely scenario. My own intuition (in
which I don't place terribly much faith, as I'm just a meat machine, and
like all of us meat machines not well qualified to think about such things)
tends more toward the Incomprehensibly Indifferent AI scenario. Maybe not
an Honest Annie scenario in which our AI's actually shut up and disappear
from us ... but maybe something close....

I have discussed this scenario on this list more than once, and have not met
with a hostile reception. I think this is because

a) there is no evidence or rational argumentation to refute this scenario...
and
b) it does not require any unlikely-sounding special assumption to ensure
its plausibility.

On the other hand, for instance, consider the scenario that Vinge puts forth
in Fire Upon the Deep, that there are limits to intelligence in this part of
the galaxy which prevent true AI from emerging, but that this is not the
case in other parts of the galaxy. This is not *impossible*, and there's no
evidence or rational argumentation to refute it, but it does require special
assumptions that seem to me unlikely to ensure its plausibility. So I don't
take this hypothesis so seriously, and probably others on the list don't
either.

The hard thing about making this sort of judgment, however, is that it's
almost GUARANTEED the future will bring all sorts of things *which would
strike our current mind as currently unlikely* -- probably things that
would strike our current minds as MORE unlikely than Vinge's
fictionally-posited cognitively-meaningful spatial-location variations in
physical law.

For this reason I am continually cautioning humility in prognostication.
Something amazing is very likely to happen, but to think that we can capture
it very thoroughly with our early 21'st century human notions like
"Singularity" and "Friendly AI" is probably a bit too hubristic.

Still, *attempting* to comprehend what's to come seems about the best
alternative we have right now ;>

ben g



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT