From: Ben Goertzel (ben@goertzel.org)
Date: Sat Jan 26 2002 - 08:17:25 MST
hey there,
> My attempt at a somewhat-new question is:
>
> How could some preeminent scientists, in this day and age, defend the
> notion that existing computer power (or even a lot less than it),
> combined with some rather simple but undiscovered programming technique,
> would be sufficient for general intelligence?
My own belief is a little bit off from this, but not so far off.
I believe that existing computer power is adequate for general intelligence,
and that what is required is a *not all that simple* design (I wouldn't call
it a "programming technique").
I also think that computers of 10 years ago, if not *quite* adequate for
human-level general intelligence, would have been adequate for a lot *more*
general intelligence than they were ever made to manifest.
I think the idea that there's a single simple trick for realising AGI
(artificial general intelligence) is a fallacy. I think that intelligence
requires the integration of a fair number of tricks in a mutually
interadapted
way so as to give rise to the emergent structures of intelligence.
I think that the reason we don't have AGI yet is that the field of AI has
not
been focused on the right things. Because AI comes out of the standard
tradition
of reductionist Western science, and the mind does not easily yield itself
to
this kind of thinking. We do not have good scientific tools for thinking
about
complex emergent structures, and how to coax their emergence from the
physical systems
at hand.
But the good news is, folks, my friends and I have arrived at *a* workable
design
for AGI ;) We call it Novamente, it's the successor to Webmind. Webmind
was basically
workable too, but had a couple problems, such as:
1) the "core" components were not implemented in a way leading to adequate
efficiency
on a von Neumann architecture ("core" refers to lower-level services like
memory
management, scheduling, etc.)
2) the interface between declarative and procedural mind was not worked out
correctly
3) the overall design was too complex which made it hard for even us
supernerds to
manage all the interrelations between the parts
We have now resolved these problems and arrived at a pretty bloody good
design if I do
say so myself. And we are working toward implementing it, but not as fast
as we'd like,
because he have a small team, and we are spending part of our time applying
simple parts
of the system to solving practical problems in the bioinformatics domain,
because we need
to eat...
"Novamente" = "new mind" in Portuguese, it also ="anew/afresh/again".... As
a background
meaning, it = "new lie" ;->
Anyway, to get back to your original question, you have, in me, a *genuine
live specimen*
illustrating the phenomenon that has been puzzling you. Probe and poke
away, and see what
my various delusions and psychological peculiarities will yield to the
scientific instrument of
your mind... ;->
-- Ben G
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT