From: Robert J. Bradbury (bradbury@www.aeiveos.com)
Date: Wed Oct 13 1999 - 10:25:08 MDT
On Wed, 13 Oct 1999, O'Regan, Emlyn wrote:
> > Unfortunately nothing else is following Moore's Law, if something else
> > was you wouldn't have to ask, it would be impossible to overlook.
> > That's not to say other things are not exponential, but they're much
> > further back toward the flatter part of the curve than Moore is.
> >
> > John K Clark jonkc@att.net
> >
> >
> Reads a bit like nonsense I'm afraid. There is no flatter bit of an
> exponential curve; it looks the same forward and back (I'm taking some
> liberties), no matter where you stand on the curve. Multiplying by a
> constant amount per constant time period (like multiplying transistors per
> square inch on an IC by 2 every 18 months), says that the rate of change is
> constant.
Sorry Emlyn, but I have to disagree with "flatter bit". If you graph
an exponential curve across a wide range the first part of it looks
very flat. Thats because while the rate of change doesn't change
over time the absolute quantity being changed does! When Moore's
law was doubling 200,000 transistors in 18 months it was very
different from the situation of doubling 20,000,000 transistors
in 18 motnhs. In the first case you go from a multi-chip CPU
to a single chip RISC CPU. In the 2nd case you would go from
~5-7 ALU/FPU (CPUs) to 10-15 tightly coupled parrallel CPUs.
(You could have many more if you could solve the data transfer and
data locality problems).
>
> This relates to the concept of a singularity perhaps. The notion that
> technology and other stuff (that's the technological term) increase so
> much that things go batso (also jargon) - I assume that the rate of
> change of the various exponential curves involved is constant.
This is not likely to be true unless we solve the design problem.
If the CPU designers had had to deal with doing useful things
with all those transistors (instead of simply making bigger
memory caches) we would have had significantly underutilized
the capacity available (a similar analogy is how much Bill
Gates can physically consume compared to what he can afford
to consume for dinner). The question is have you "fully
utilized" your capacity in an "optimal" fashion.
I expect the singularity (in terms of market penetration of
actively used nanotechnology constructs) will go through a big
jump as we all move to Bill Gates lifestyle. If you believe
my analysis and we are "polite" about it, as I've suggested
we should be, that may take 5-10 years. Then we may have
a slowdown while we figure out whether or not we really want
to dismantle the solar system and construct an SI or redesign
all of the cities on Earth or who knows what. Unless we
produce AIs with some well developed aesthetic sense and/or
significantly increase inter-human communication bandwidth
(which requires solving the mind-to-mind mapping problem)
we are going to probably be relatively slow in deciding
what we should do next (due to a lack of general agreement
on what looks good and the "right" general direction and
the relatively slow process we currently use to communicate
our perceptions of and rationale for the best solutions).
> I also
> assume that people can adapt (and indeed have adapted) to a constant
> rate of change - look at the computer hardware industry. Acknowledgement
> of Moore's Law (and design > with it in mind) shows that people can
> cope with constant change, as long as it is predictable.
But it will not be predictable. If you can manipulate the mass
of a planet in a day and what comes out of it can be virtually
anything that can be built, I have a tough time seeing how
that is predictable (at least on the scales we are used to).
The phase space of Moore's Law is predictable because it is
confined. The phase space of what can is enabled by nanotech
is much much larger.
> So what we are concerned with re: singularity is not increasingly
> accelerated change; that's most likely not to happen.
I disagree, the absolute amount of change possible per unit time is
radically different from what we are adapted for. We will either have to
adapt to the rate of change or underutilize the potential capacity.
> What is more important is the qualitative thresholds reached in what
> we can accomplish as this predictable constant increase in technology
> advances. Moore's law (and other exponential increase) is not a direct
> threat,
No because on the current scale it is ~ 1/500 of a human brain capacity.
But when you are going from 1 to 2 to 4 to 8 brain capacity at a rate
significantly faster than we can build and educate new wet brains then
at least from some perspectives you *do* have a threat!
> but it may (should) lead to such things as nanotech, strong AI,
> intelligence augmentation, etc, which all constitute qualitative
> changes to the human landscape. Then new curves are created...
Agreed. Just as steamboats and airplanes enabled lifestyles different
from those that preceeded those developments, the things you mention
will alter the landscape as well.
The fundamental question is whether or not we have nanotech designs
(either evolved or built by enough well-educated nano-mechanical-
chemical engineers) and/or whether AI/IA can be relatively
self-bootstrapping. If we have those things then the growth
curve will be fast and smooth. If we don't have those things
then the growth curve will be bumpy and somewhat slower.
Robert
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:05:30 MST