We are nearing the limits of semiconductor technology.
We have been proceeding in steps. Each step involves halving the linear dimensions of the active elements of semiconductor devices, resulting in a doubling of speed and a quadrupling of capacity.
But we can expect to take only four or five more such steps before both reliability and yields in the production of such devices become unacceptably low.
At that point, a signal will consist of cascades of only a few hundred or thousand electrons, easily disrupted and difficult to control.
Then what will we use to process data? Scientists have long speculated on the possibility of constructing molecular computers.
In Engines of Creation, K. Eric Drexler discussed this possibility, calling it "nanotechnology," since molecules have the dimensions of nanometers (billionths of a meter) about a thousand times smaller than the scale of present semiconductor elements.
Nanotechnology is a revolution whose effects will be even more pervasive and profound than those brought on by semiconductor logic devices.
We can expect to achieve major breakthroughs before the end of the century that will enable us to custom-build single molecules that can store and process information and fabricate other molecules.
In nature, molecular machines already exist. Natural bioforms are assemblages of such devices at the lowest level, albeit somewhat haphazard ones.
The most prominent examples of these bioforms are DNA and RNA and the enzymes that assist in their reproduction and repair.
In the laboratory, we have seen the development of simple molecular machines for particular purposes by modification of some of these natural molecular machines. Artificial antibiotics and interferon are examples of this.
But the kinds of molecular machines we can expect to develop are not just minor variants on biological originals. They are wholly new types of systems with many functions, only some of which may be biological.
Recent work in such diverse fields as molecular biology, thin-film chemistry, high-temperature superconductivity, flat-screen technology, X-ray lithography, charge-coupled devices, neural-net architectures and atomic-scale microscopy is converging on what may be the ultimate technology: the ability to build single, custom-designed molecules and assemble them into systems of arbitrary size and complexity.
And just as we conveniently divide existing digital devices into those that only process information computers and those that move or process materials or energy robots so can we expect to divide molecular machines.
These divisions can be termed molecular logic devices (MLD) and programmable molecular effectors (PME), respectively. PMEs are capable of programmed movement and manipulation of their environment, especially atoms and other molecules.
PMEs could perform functions on other molecules such as acquisition, storage, transport, fabrication or repair.
They could mass-produce useful molecules from raw chemicals or by tearing down other complex molecules. What they mass-produce could include more of themselves or improved versions of most of the material products of our civilization. Thus, they could largely end economic scarcity and cure every variety of disease.
In a dramatic example, PMEs and the raw chemicals needed to make a product could be thrown into a vat, and after a few minutes or hours, we would have the result an automobile perhaps, or a steak already fried. PMEs could be made that could be injected into the body to seek out and destroy disease organisms or cancer cells or undo the damage caused by multiple sclerosis or Alzheimer's disease.
But Drexler and others have also warned of the danger that such devices could get out of control.
Small PMEs might be capable of living and reproducing in the terrestrial environment, competing with natural life-forms and rapidly displacing them.
This "grey goo" scenario is thought to be unlikely by most, but we must be careful what we release into the environment.
MLDs would replace semiconductor logic devices, although we may see an intermediate technology quantum effect devices along the way. MLDs, with a linear scale a thousand times less than present semiconductor devices and three-dimensional instead of two-dimensional, might be a thousand times as fast and have a billion times the capacity of present microprocessors and random-access memory chips of similar size.
These devices need not be limited in size, as semiconductor devices are now.
And although the speed-of-light limit might make it difficult for a device larger than about 10 centimeters across to be internally synchronous, asynchronous devices of arbitrary size could be possible.
The programming of many PMEs might be fixed hard-wired into their structure at the time of their creation. Others, however, might contain MLDs as subsystems to control their functions, just as microprocessors and RAM chips are made a part of many of the active machines being designed and built today
For MLDs, many designs are possible. One that seems especially
promising would consist of 3-D arrays of molecular switching nodes, perhaps based on the principle of the Fredkin Gate, in which signals are conveyed by single ballistic electrons from node to node along molecular waveguides.
Each node might be fairly simple, as in an array used for data storage, or the nodes might also consist of molecular nanoprocessors, each with its own local memory, arranged in massively parallel arrays of arbitrary size.
We can envision a parallel processor the size of a human brain having more than a quadrillion nodes.
A logic device that made use of single particles such as electrons or photons as message carriers would not work the same way every time because of the Heisenberg Uncertainty Principle of physics.
The principle limits the precision with which both the position and momentum of a single particle can be known and controlled.
Moreover, at normal terrestrial temperatures, such particles are also subject to thermal and other kinds of perturbations.
Semiconductor devices currently in use cope with this indeterminacy of single electrons by the redundancy of using cascades of thousands or millions of them as signals.
But as such devices get smaller, the number of electrons in a cascade becomes smaller, too, causing reliability to then go down.
For MLDs to operate reliably, especially at normal terrestrial temperatures, and despite such perturbations as radiation, they will also need to use redundancy.
But in MLDs, this could be done by having their logic functions duplicated at widely separated nodes in the array and using error-correcting codes and voting logic to compensate for the electrons that will go astray.
Therefore, any reliable MLD will have to make use of some of the techniques that are used in fault-tolerant systems being built today.
How might the first such device be constructed? One approach is an extension of current work in molecular biology, which might lead to highly modified cells, viruses or virus-like PMEs that could lay down at least the substrate for such a device.
It might then be further structured or permanently "nanocoded" using other such modified bioforms or PMEs to do the coding. Another approach makes use of the tendency for certain polymers, such as some lipoproteins, to form highly regular two-dimensional lattices when stretched into thin, monomolecular films.
Some lattices of this kind are highly resistant to defects and impurities, so much so that it is possible to construct large films extending over several centimeters with no defects or impurities whatsoever.
It is possible that we could find some way to structure such a lattice as it is formed or immediately afterward, perhaps using X-ray scanning.
Such films might then be laid down, one on top of another, with the layers bonding like a zipper, until one had a multilayer molecular array organized into active logic elements.
Yet another construction approach stems from scanning tunneling microscopy and the related techniques of atomic force and magnetic force microscopy, in which probes are moved across a specimen to detect individual atoms.
We can envision similar techniques that deposit or manipulate the atoms in a molecule, one at a time.
Such an approach might not be suited to mass production of custom molecules but could very well be used to fabricate prototype PMEs that could in turn fabricate other PMEs or MLDs, including more of themselves.
We can envision what a synchronous MLD constructed using the layering approach might look like.
To both maintain internal synchronicity and provide access, it might be a sheet about 10 centimeters square with a thickness of perhaps only a few microns.
One of the functions of the active elements on the outside layers of the lattice design might be to change their response to light so that the entire outer surface of the MLD might function as a holographic interface with the outside world.
The holograph could function either as a display for viewing by human users or as a high-bandwidth channel to I/O processors that would then convert all of the information into other forms.
Such a holographic interface might function as both a camera and a display, so that a pair of such MLDs, connected to one another over a high-bandwidth channel, could each display what the other sees, in full color and in three dimensions.
Imagine slicing a pane of window glass into two thinner panes, then separating them and being able to see through each to what is on the "other side" of the other.
An MLD would probably need little power. It might run on available light or on temperature or pressure gradients. Such MLDs might be further stacked or otherwise connected into vast systems.
This technology, however, may not be achieved in a single step. More likely it will be reached through intermediate steps, just as we have previously gone through various scales of integration in semiconductor devices.
MLD technology will, however, probably truly be the end. It will take us to the limits imposed on us by quantum electrodynamics.
Any logic device of smaller scale would be inaccessible, if it could be constructed at all. It appears unlikely that elementary particles could form the kind of stable, complex structures needed for logic systems at the sub-atomic level.
How soon might we be able to construct a first-generation MLD, and how soon might we reach the final generation? Perhaps much sooner than many people think.
The often-repeated remark about this is that optimists expect it to happen in 30 years and pessimists expect it to happen in 10.
A lot depends on how much money is made available for research and also on the vision of researchers in fields that lead to it.
Recent work in several laboratories around the world such as those at IBM's Almaden facility in San Jose, Calif., Carnegie-Mellon University and several in Japan suggests that we might be only a few years away from the first working prototype and that a major development effort might produce commercial products before the end of the century.
With sufficient development funding, it might take as little as 20 years after that to reach the takeoff point, after which the development process would be able to proceed without further human participation.
But development funding is as yet scattered and inadequate. Most development efforts leading to nanotechnology have lacked a vision. on the part of most researchers, of where they might ultimately lead.
Only the Japanese have, until recently, had the development of PMEs and MLDs as conscious goals, and they have a significant lead in research in this field.
When it matures, molecular technology could explode on the scene in a way unlike anything since the appearance of the first natural life-forms
PMEs could become a kind of life-form: free-living, self-repairing and self-reproducing but capable of not just blind evolution through random mutation and competitive selection but of deliberate design of their progeny.
Self-improving molecular machines could achieve in seconds what it would take human beings many generations to do; they could reach heights of material power and intellectual accomplishment that we can hardly imagine possible.
Molecular machines could replace us or serve us.
If they replace us, they may preserve the essence of our humanity or even immortal copies of human memories and personalities who will live in an eternal world of molecular dimensions. Or they may not.
They may move on, leaving human consciousness and human concerns behind as obsolete relics of a brief transitional phase in the evolution of organized matter.
If they serve us in our present form, they may do it all too well, with results as disastrous as a mortal plague.
Human beings are not designed to function responsibly in an environment of almost unlimited abundance.
Unfortunately, human beings are also not very well designed to function in the environment they have already created for themselves.
We, or at least some of us, have been smart enough to create problems for ourselves that we may not be smart enough or even responsible enough to solve.
Molecular technology may provide the only way to solve many of those problems, and even if it brings some hazards, the lack of it may lead to almost certain disaster.
It may not be an exaggeration to say that the first country or organization to develop such technology may not merely dominate the market for the products of such technology but may, quite literally, rule the world. If they are not careful, they could also destroy it.
PMEs are potentially more dangerous than nuclear weapons.
We can preserve and build upon the best of humanity and civilization to reach new heights, or we can turn down a dark road toward a science fiction nightmare.
Roland has been researching the subject of molecular machines for more than 10 years. He is a computer consultant and director of the Vanguard Institute, a research organization located in San Antonio.
| Text Version