In a message dated 98-05-17 12:07:10 EDT, Dan Clemmensen wrote:
> You seem to be saying "OK, we may learn how to build the machinery, but
> the proper control systems is a lot harder."
Yes, I am.
> I believe that given an assembler and one of today's desktop computers, It
> should be possible to build a supercomputer factory on your desktop. Well
> over half
> of the current cost of computer components reflects the cost of the fab
> lines, and
> most of the rest reflects operating and materials costs.
In theory, this is true, but see below.
> IMO a similar approach can build any simple diamondoid macro-scale part,
> such as
> window panes, bricks, pre-cut building lumber, many machine parts, etc.
I absolutely agree, and think that many, many macro-scale components will be
fabricated at the molecular level sometime between 2005 - 2025.
If by "advanced
> you mean the abiltity to make things other than diamondoid, such as
> I agree, we need something qualitatively advanced from the "robot factory"
Right, and herein lies my doubts about development of an "anything box" before
2015, at least. I believe it was Robin Hanson who best expressed my
reservations: We know it takes the information processing capacity of either
(1) a large segment of the economy or (2) a very complex living organism to
generate just about any complex device right now. Any device that could
generate a complex device from inputs of only atoms and information would have
to be equally complex and coordinated. Creation and control of such an
"anything box" doesn't seem to be any simpler than modeling an entire economy
or organism at a highly detailed level. I believe this problem is solvable,
but is very, very difficult. It seems to me that it will be some time before
a device under the ultimate control of a single individual -- even a highly
augmented one -- can be created.
> However, the IMO the robot factory generating diamondoid parts is easily
> achievable in
> the 2005-2015 timeframe. This technoligy is sufficient to completely
> obsolete all
> current capital equipment.
Not overnight, I think. The process will be "bottom up" in the sense that
simple component manufacture will fairly quickly succumb to "STMs-on-a-
superchip" manufacturing technology, but much secondary processing and
infrastructure handling will then be transformed more slowly (albeit in a
matter of a decade, IMO).
> If the groundwork in AI based on massively
> parallel computing
> is in place when the assember is developed, then the assembler-built
> superconputers will
> also precipitate the singularity.
Again, I think this presupposes a much, much higher degree of modeling and
control of complex systems than we have yet achieved. Accordingly, I see a
"great leap forward" in hardware that jumps a good deal ahead of "software"
for perhaps a decade.
Greg Burch <Gburch1@aol.com>----<email@example.com>
Attorney ::: Director, Extropy Institute ::: Wilderness Guide
http://users.aol.com/gburch1 -or- http://members.aol.com/gburch1
"Good ideas are not adopted automatically. They must
be driven into practice with courageous impatience."