Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Computing with Molecules. 
In addition to nanotubes, major progress has been made in recent years in computing 
with just one or a few molecules. The idea of computing with molecules was first suggested in the early 1970s by 
IBM's Avi Aviram and Northwestern University's Mark A. Ratner.
14
At that time, we did not have the enabling 
technologies, which required concurrent advances in electronics, physics, chemistry, and even the reverse engineering 
of biological processes for the idea to gain traction. 
In 2002 scientists at the University of Wisconsin and University of Basel created an "atomic memory drive" that 
uses atoms to emulate a hard drive. A single silicon atom could be added or removed from a block of twenty others 
using a scanning tunneling microscope. Using this process, researchers believe, the system could be used to store 
millions of times more data on a disk of comparable size—a density of about 250 terabits of data per square inch—
although the demonstration involved only a small number of bits.
15 
The one-terahertz speed predicted by Peter Burke for molecular circuits looks increasingly accurate, given the 
nanoscale transistor created by scientists at the University of Illinois at Urbana-Champaign. It runs at a frequency of 
604 gigahertz (more than half a terahertz).
16
One type of molecule that researchers have found to have desirable properties for computing is called a 
"rotaxane," which can switch states by changing the energy level of a ringlike structure contained within the molecule. 
Rotaxane memory and electronic switching devices have been demonstrated, and they show the potential of storing 
one hundred gigabits (10
11
bits) per square inch. The potential would be even greater if organized in three dimensions. 
Self-Assembly. 
Self-assembling of nanoscale circuits is another key enabling technique for effective nanoelectronics. 
Self-assembly allows improperly formed components to be discarded automatically and makes it possible for the 
potentially trillions of circuit components to organize themselves, rather than be painstakingly assembled in a top-
down process. It would enable large-scale circuits to be created in test tubes rather than in multibillion-dollar factories, 

---

using chemistry rather than lithography, according to UCLA scientists.
17
Purdue University researchers have already 
demonstrated self-organizing nanotube structures, using the same principle that causes DNA strands to link together in 
stable structures.
18 
Harvard University scientists took a key step forward in June 2004 when they demonstrated another self-
organizing method that can be used on a large scale.
19
The technique starts with photolithography to create an etched 
array of interconnects (connections between computational elements). A large number of nanowire field-effect 
transistors (a common form of transistors) and nanoscale interconnects are then deposited on the array. These then 
connect themselves in the correct pattern. 
In 2004 researchers at the University of Southern California and NASA's Ames Research Center demonstrated a 
method that self-organizes extremely dense circuits in a chemical solution.
20
The technique creates nanowires 
spontaneously and then causes nanoscale memory cells, each able to hold three bits of data, to self-assemble onto the 
wires. The technology has a storage capacity of 258 gigabits of data per square inch (which researchers claim could be 
increased tenfold), compared to 6.5 gigabits on a flash memory card. Also in 2003 IBM demonstrated a working 
memory device using polymers that self-assemble into twenty-nanometer-wide hexagonal structures.
21
It's also important that nanocircuits be self-configuring. The large number of circuit components and their inherent 
fragility (due to their small size) make it inevitable that some portions of a circuit will not function correctly. It will 
not be economically feasible to discard an entire circuit simply because a small number of transistors out of a trillion 
are non functioning. To address this concern, future circuits will continuously monitor their own performance and 
route information around sections that are unreliable in the same manner that information on the Internet is routed 
around nonfunctioning nodes. IBM has been particularly active in this area of research and has already developed 
microprocessor designs that automatically diagnose problems and reconfigure chip resources accordingly.
22