Re: Singularity Objections: Singularity, Intelligence

From: Thomas Buckner (
Date: Thu Jan 31 2008 - 16:37:34 MST

--- Thomas McCabe <> wrote:

> * A smarter being is also more complex, and
> thus cannot
> necessarily improve itself any faster than the
> previous stage -- no
> exponential spiral.
> o Does anyone have counter-evidence?
> This looks like a real
> possibility. - Tom

I'm not saying I can answer this, but I can think
of a couple of analogies which might throw light
on the shape of a rebuttal.

Analogy 1: Metazoa. During the Cambrian Explosion
some 535 million years ago, we see (relatively)
sudden development of every modern metazoan
(multi-celled animal) phylum, plus several that
don't exist now. Previous fauna were odd and
mostly soft-bodied things, and before that the
seas were filled with single-celled creatues.
Suddenly in the Cambrian we have all sorts of
novelties: legs, gills, shells and claws, eyes,
you name it. How could nature get complex so
The real advance was in the software. Eukaryotes
had learned to live in colonies, and used such
software as HoX genes to build segmented bodies.
>From a row of identical simple segments,
specialized segments with arms, ribs, and legs
are an 'obvious' next step. Many wildly
diferent-looking beasts may actually have been
close relatives.
But now look at us metazoa: we really are largely
stuck with out basic body plans as originally
developed half a billion years ago. We can't
rearrange our organs or grow six extra legs on a
whim, because the systems are now too complex,
too interlinked. We've got surviving to do.

Oh, wait: the situation just changed again: we
humans, with biotechnology, are no longer forced
to play the biological hand we were dealt. We've
developed ways to meddle with biology that no
creature has ever had. Singularity thinking
assumes AI with a similar level of power to
analyze and alter itself, from scratch if need

Analogy 2: NASA. This is in answer to a later
objection in the original E-mail, which I can't
quote because it got truncated. Rough paraphrase:
Intelligence is not a single measure of a single

Consider NASA as if it were an AI. This hive mind
was created in the late 1950's by joining
thousands of human brains to achieve a short,
discrete list of supergoals (Land a man on the
Moon before the Soviets do; Build satellites and
landers to investigate other planets and moons;
Build satellites to view Earth phenomena from
above; etc.) Each node (human brain) was chosen
for high performance in the areas of research
most apppropriate to the supergoals; nodes were
linked by advanced communications and supplied
with advanced equipment, and expected to develop
even more advanced equipment to achieve the
supergoals. Once the supergoal 'Land men on the
moon' was achieved, problems developed in the
hive mind. The superusers (U.S. government and
taxpayers) issued bad or conflicting new
supergoals, harming the effectiveness of this AI
which had proven to be amazingly effective under
the right conditions. A major successor to the
Moon landing supergoal was 'Build and fly a
reusable space plane every two weeks.' This was a
terrible supergoal: the timetable was
unachievable, the space plane idea was a sop to
the Pentagon super-users that was a poor choice
for the 'Launch satellites supergoal,' and so on.
Eventually one of the space planes exploded
because it was flown in cold weather against the
advice of engineers.

My point in this tale is that sooner or later,
the whole complex system yokes together many
nodes of human-level intelligence, with all their
various abilitites and talents, and yet needs to
be able to turn on a dime beause one little node
(a frantic solid-booster engineer in this case)
says so. NASA in 1986 fell right on its face
precisely because of inertia, because the handful
who knew there was a problem had no power to
elevate their concerns to near-supergoal status.
One way to visualize a successful FAI is to
imagine NASA crammed into a box, and to assume
that it made fewer mistakes than the real NASA.

Objection: That sounds hard!
Rebuttal: Yes, but we have to try anyway.

Tom Buckner

Looking for last minute shopping deals?
Find them fast with Yahoo! Search.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:01 MDT