Originally published February 21, 2002 on KurzweilAI.net.

“The future of computing may not be focused on computation so much as fabrication,” said MIT Professor Neil Gershenfeld, reflecting a common theme at the seminar: building computers from the bottom up and using novel strata will enable Moore’s 1^st law (chip speed is doubling every 18 months) to continue far into the future, while defeating the cost-doubling of chip manufacturing plants that occurs every three years–Moore’s less-popular 2^nd law.

Advances in Nanoelectronics

According to Dr. James Ellenbogen of the MITRE Nanosystems Group, the advances in nanotechnology in the final months of 2001 were like “living history…every other week something new was happening.” He pointed out that two of the most important obstacles facing molecular-scale electronics are achieving signal amplification using molecular devices (basically, wiring them in a stable fashion) and building logic circuits using molecular or nanoscale devices–thus using them for information storage.

Five research groups achieved these gains in Fall 2001 with different approaches (see “New world of nanoelectronics may arrive in the near future, AAAS speakers say“).

Dr. Charles Leiber of Harvard University has created what he calls “a damn good switch” using nanowires. Because of their semiconducting properties, he believes they’e the best candidate for a fundamental building block. The cross-nanowire field effect transistor that his group created from nanowires could be used as biological sensors or scaled up to form a computing architecture without the need for silicon as a substrate–quite possibly in three dimensions.

What these groups have in common is that they have achieved real results in manipulating electrons at a nanoscale to create reliable electronic devices. Although, as Professor Mark Ratner of Northwestern University pointed out, “the interconnects aren’t there yet,” these devices form an arsenal of building blocks for constructingscalable architectures. Instead of crunching components onto silicon, smart devices will be constructed from the bottom-up–achieving levels of precision in manufacturing at an atomic level, as opposed to the conventional top-down approach of etching features onto silicon.

The greatest challenge is to begin putting devices together–a challenge that is overcome when, as Leiber said, “the same intellectual underpinnings are required when working with these building blocks, whether for bionanotechnology or quantum computing.”

In his keynote, Leiber also spoke of the creation of NOR gates–a programmable logic device–using carbon nanotubes, another key building block. Professor Cees Dekker of TU Delft used semiconducting carbon nanotubes to build a single molecule transistor that operates a room temperature. “Four years ago,we barely observed current through nanotubes,” Dekker said. “Now, we have controlled transistors, semiconduction, working circuits.” For Dekker, the future is in the merger of inorganic clusters of computable nanotubes with “biological routines and material, such as DNA.”

Dr. Hendrik Schoen of Bell Labs/Lucent Technologies has demonstrated a compromise between the top-down world of conventional chip lithography and bottom-up nanomanufacturing. Schoen’s approach is to grow layers of conductive molecules–“monolayers”–on silicon using interactions between gold molecules and thiols. This layer can operate as multiple transistors of molecules. The cost savings is clear–as Schoen put it, “no cleanroom, no lithography…[is required] to create transistor action on a molecular scale.”

New Paradigms

In a talk entitled “The Problem Before Us,” James Crutchfield of the Sante Fe Institute opened the AAAS New Computing seminar by citing Moore’s 2^nd Law (the cost of conventional chip fabrication plants doubles every three years) and then asking the question, “what kind of intrinsic processes do natural systems use to store information?”

By turning away from the paradigm of computing introduced by Alan Turing–that all problems can be solved using the logic that forms the basis of modern computing–we may find more flexible, complex and far less expensive methods of computing in naturally occurring phenomena, according to Crutchfield. Instead of Turing Machines, he proposes a new class of theoretical machines called Epsilon Machines, which emulate the pattern-based memory of cellular automata or symmetrical neuronal firing in the human brain.

How do you measure global computation among neurons, asks Crutchfield? By representing the parallel connections between neurons with “causal states.” Beginning with a 1 for a synchronized cluster of neurons and a 0 for a non-synchronized cluster, we have the roots for information storage. “This forms intrinsic computation: models of computation engineering can build on,” said Crutchfield.

Crutchfield’s basic thesis is that these systems abound in nature around us. To harness their informational power, one must ask three questions: how much memory does the system inherently have, how is it organized, and how does it produce further behavior? Neil Gershenfeld followed this reasoning in a presentation that featured oddities ranging from printable bicycles to an Internet browser engineered for a parrot (The Interpet Explorer): to “ask nature to solve a problem instead of building a machine to do it.”

Neil Gershenfeld, MIT Media Lab Professor and author of When Things Start To Think, spoke of bringing “ideas out of the box and into the world” with new methods of fabrication. Computational power will not be as important in the future as the ease and speed of producing computational devices. Gershenfeld spoke about allowing common objects in the physical world to take on intelligence and, perhaps more importantly, bringing computing to developing countries in novel ways–to people who need computing “first and most urgently.”

Examples? A liquid that stores information and can be used as paint, bringing intelligence to walls. Reconfigurable ink that not only creates new methods of displaying digital information but can hold information, forming “printable computers.” By discovering and creating applications with “emergent multi-hop networks” and other programmable characteristics within existing physical materials, Gershenfeld feels that we will realize “more profound consequences than just creating faster computers.”

He points out a dramatic reversal of the digital divide that may occur if those without computers obtain the most advanced computational technologies and supersede the comparatively slow pace of the standard, desktop-computer-literate world.