Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Engines of Creation
in 1986 and was articulated technically in his 1992 book, 
Nanosystems
) laid out 
the foundation of nanotechnology and provided the road map still being followed today.
74
Drexler's "molecular assembler" will be able to make almost anything in the world. It has been referred to as a 
"universal assembler," but Drexler and other nanotechnology theorists do not use the word "universal" because the 
products of such a system necessarily have to be subject to the laws of physics and chemistry, so only atomically 
stable structures would be viable. Furthermore, any specific assembler would be restricted to building products from 
its sea of parts, although the feasibility of using individual atoms has been shown. Nevertheless, such an assembler 
could make just about any physical device we would want, including highly efficient computers and subsystems for 
other assemblers. 
Although Drexler did not provide a detailed design for an assembler—such a design has still not been fully 
specified—his thesis did provide extensive feasibility arguments for each of the principal components of a molecular 
assembler, which include the following subsystems: 
•
The 
computer
: to provide the intelligence to control the assembly process. As with all of the device's subsystems, 
the computer needs to be small and simple. As I described in chapter 3, Drexler provides an intriguing 
conceptual description of a mechanical computer with molecular "locks" instead of transistor gates. Each lock 
would require only sixteen cubic nanometers of space and could switch ten billion times per second. This 
proposal remains more competitive than any known electronic technology, although electronic computers built 
from three-dimensional arrays of carbon nanotubes appear to provide even higher densities of computation (that 
is, calculations per second per gram).
75

---

•
The 
instruction architecture
: Drexler and his colleague Ralph Merkle have proposed an SIMD (single instruction 
multiple data) architecture in which a single data store would record the instructions and transmit them to 
trillions of molecular-sized assemblers (each with its own simple computer) simultaneously. I discussed some of 
the limitations of the SIMD architecture in chapter 3, but this design (which is easier to implement than the more 
flexible multiple-instruction multiple-data approach) is sufficient for the computer in a universal nanotechnology 
assembler. With this approach each assembler would not have to store the entire program for creating the desired 
product. A "broadcast" architecture also addresses a key safety concern: the self-replication process could be 
shut down, if it got out of control, by terminating the centralized source of the replication instructions. 
However, as Drexler points out, a nanoscale assembler does not necessarily have to be self-replicating.
76
Given the inherent dangers in self-replication, the ethical standards proposed by the Foresight Institute (a think 
tank founded by Eric Drexler and Christine Peterson) contain prohibitions against unrestricted self-replication, 
especially in a natural environment. 
As I will discuss in chapter 8, this approach should be reasonably effective against inadvertent dangers, 
although it could be circumvented by a determined and knowledgeable adversary. 
•
Instruction transmission
: Transmission of the instructions from the centralized data store to each of the many 
assemblers would be accomplished electronically if the computer is electronic or through mechanical vibrations 
if Drexler's concept of a mechanical computer were used. 
•
The