What path will be followed to the first assemblers?
Several paths lead to nanotechnology, and work contributing to
one or more of those paths has won several recent Nobel prizes.
Even without the motive of building assemblers, practical and
academic motives have moved technology in directions that bring
Nobel prize for chemistry went to Charles
J. Pedersen, Donald
J. Cram, and Jean-Marie
Lehn for developing relatively simple molecules that perform
functions like those of natural proteins. Pederson synthesized
what are known as "crown ethers," a family of molecules
that selectively bind specific metal ions in solution, holding
them in properly-sized internal hollows. Cram and Lehn have
extended this work, using chemical techniques to synthesize a
wide range of molecules that specifically bind other molecules.
This sort of selective binding is a common protein function.
The molecular machinery of cells self-assembles though the
selective binding of one protein to another. Other molecules that
bind selectively to one another might likewise be used as a basis
for molecular machinery, providing an alternative to proteins for
building first-generation assemblers. The ongoing work of Cram,
Lehn, and their coworkers may be of great importance to the
development of nanotechnology.
If protein design remains too difficult, building initial
molecular machines from non-protein molecules may prove an easier
path. Myron L. Bender and Ronald
Breslow have already made non-protein molecules that function
Nobel prize for physics when to Gerd
Binnig and Heinrich
Rohrer for development of the scanning tunneling microscope
(STM). This device, reported in 1982, uses vibration isolation,
piezoelectric positioning elements, and electronic feedback to
position a sharp needle near a conducting surface with atomic
Assemblers, of course, will work by positioning reactive
molecules to atomic precision to direct chemical reactions.
Several persons familiar with Eric Drexler's work on assemblers
(including Drexler, Conrad Schneiker, Steve Witham, and no doubt
others) independently observed that, as Engines of Creation
notes, mechanisms of the sort used in scanning tunneling
be able to replace molecular machinery in positioning molecular
tools," perhaps helping to build a first-generation
Suitable molecular tools remain to be
developed. In a series of experiments, R. S. Becker, J.
A. Golovchenko, and B. S.
Swartzentruber have produced modifications on a germanium
surface, measured as 0.8 nanometer wide. These features are
thought to represent single atoms of germanium, electrically
evaporated from a bare STM tip, with the large size of the
features resulting from problems with STM resolution. At last
report, they were unable to call their shot (that is, to put the
atom in a pre-selected location), and the process did not work
for the related element, silicon. The evaporation process
requires that the STM tip be retracted from the surface.
Scanning tunneling microscopy also promises to be of use in
characterizing molecules, since it can give atomically-detailed
pictures of various surfaces. This could speed molecular
engineering, helping designers to "see" what they are
doing with greater ease. Little has been demonstrated as yet,
however. It has been used neither to sequence DNA, nor to
characterize unknown chemical structures. George Castro of IBM's Almaden
Research Center reports that experimenters have thus far had
difficulty detecting molecules on surfaces, to say nothing of
determining their structures. Nonetheless STM and related
technologies for microscopy and micro-positioning are well worth
watching as possible aids to the development of nanotechnology.
Nobel prize for chemistry went to Bruce
Merrifield for developing the technique used for synthesizing
the most complex, specific chemical structures now made. This
technique, known as solid phase synthesis (or simply the
Merrifield method) uses a cyclic set of reactions to extend a
polymer chain anchored to a solid substrate. Each cycle adds a
specific kind of monomer, building polymers with a specific
The Merrifield method is at the heart of the machines now used to
manufacture specific proteins and gene fragments by chemical
methods. It is thus central to protein and genetic engineering
(either one of which could, in principle, proceed without the
other). The Merrifield method could be used to make other
polymers, perhaps including non-protein molecules with
protein-like functions, such as specific binding and
self-assembly. By providing multiple paths to complex molecular
systems, the Merrifield method provides multiple paths to
What path will be followed to the first assemblers, and hence to
nanotechnology? It is hard to guess, today. Protein engineering
will clearly suffice, because proteins already serve as the
components of complex molecular machines. Micro-positioning
technologies may help, though development of suitable molecular
tools seems likely to prove the hard part of the task. Molecular
systems like those explored by Cram and Lehn, together with
synthetic techniques based on the Merrifield method provide a
wealth of alternatives having many of the advantages of protein
engineering, but fewer constraints. With that lack of
constraints, however, comes a lack of knowledge, a lack of
examples from nature. A reasonable guess is that several paths
will be followed, and will contribute in a synergistic fashion.
First-generation nanotechnology need not be based on any single
class of molecule or device.
In considering this confusing wealth of possibilities, two points
are important to keep in mind. The first is that multiple
approaches multiply possibilities for success, bringing it
closer: assemblers will arrive by whichever is, in practice, the
fastest (a simple tautology!), hence difficulties with any single
approach need not mean overall delays. The second is that how the
first assemblers are built will make little long-term difference:
crude assemblers will be used to build better assemblers, and the
nature of nanotechnology will soon become independent of the
nature of the initial tools. In short, this kind of uncertainty
about the path ahead--stemming from a wealth of promising
possibilities--gives confidence in the emergence of assemblers,
without obscuring the nature of the subsequent nanotechnology.
Nanotechnology may seem remote. Molecules are invisibly small,
and they differ from the familiar objects of daily life.
Manipulating them with assemblers is essential to nanotechnology,
but assemblers will take years to develop.
Computers, though, can bring nanotechnology closer, letting us
design molecular systems using computer models, years before we
have assemblers able to build them in real life. This
design-ahead process seems sure to occur, but when will it begin?
Roger Gregory of the Xanadu hypertext project argues that the
answer to this is simple: Almost immediately.
If design-ahead were to require expensive facilities and major
funding, it would need to wait for broad acceptance of the
importance of nanotechnology, or even for a sense of its
imminence. This might take years. But Gregory observes that the
early stages of design-ahead need neither funding nor new
facilities: personal computers and motivated hackers are enough.
("Hackers" is used here, not in the media's sense of
computerish juvenile delinquents, but in the original sense of
inventive technologists making computers jump through hoops.) The
growth of amateur molecule-hacking may have major consequences
for the emergence of nanotechnology.
David Nelson, chief technical officer at Apollo Computer, has
plotted trends in computer price and performance. They follow a
classic smooth exponential, with performance at a given price
growing ten-fold every seven years or so. At this rate, good
personal computers today have roughly the power of a
seven-year-old minicomputer or a fourteen-year-old mainframe.
There is every reason to expect this trend to continue for years
to come. (Nanotechnology will eventually put many billions of
today's mainframes into an air-cooled desktop package, but that
is another story.)
Molecular modeling software--able to describe molecules and the
forces that shape them--has advanced over the years while
migrating into less and less expensive machines. After long
residence on machines such as Digital Equipment Corporation's VAX
minicomputers, it has now arrived on personal computers, such as
the Macintosh. Prices are still high and offerings sparse, but
the barrier to amateur molecular design work is being breached.
in Update 6 contains links to current
information on molecular modeling available on the Web.
See also review of Chem 3D Plus by Eric Drexler, and links
therein to current material, in article
in Update 11.
Can these computer models give accurate results? This depends
on one's standard of accuracy, which in turn depends on one's
In engineering, one need only have enough accuracy to distinguish
between designs that do and don't work. In nanoengineering, as in
ordinary engineering, designers will generally (though not
always) aim to maximize such things as stiffness and strength,
while minimizing such things as size, mass, and friction. A
designer can often compensate for an inexact model by aiming for
a large, favorable margin in the uncertain parameters. Software
based on modern molecular mechanics models is fairly accurate
even by scientific standards; it should be good enough to design
a wide range of molecular machines, with substantial confidence
in the results.
Molecular modeling software falls into various classes. At the
low end are programs that just provide a three-dimensional
software sketch pad for patterns of atoms in space. At the high
end are systems of programs that do the sort of molecular
mechanics mentioned above--that derive molecular shapes and
energies from information about the interactions among atoms. (An
example of the latter is MicroChem, a program for
the Macintosh; we expect to have a copy of version 2.0 to review
for the next issue.)
Present systems are expensive and may need some adaptations to
make them more useful for the design of molecular machinery. Once
suitable software is available at a reasonable price, however, we
can expect to see the emergence of a community of molecule
hackers. Interest in nanotechnology is high in the computer
community, and electronic mail and bulletin board systems will
make it easy for designers to swap ideas, designs, and
criticisms. Once the process gets rolling, designs for molecular
widgets--such as gears, bearings, shafts, levers, and logic
gates--should accumulate at a good pace, spawning a lively
informal competition to design the best. As computers and
software improve, the complexity of feasible designs will grow.
A few examples of more recent molecular device
The spread of amateur nanomachine design will spread an
understanding of molecular machines and nanotechnology. It will
spread the idea of design-ahead by demonstrating it in action.
Within the nanotechnology community, it will provide a channel
for creative activity having concrete results, ranging from
pictures suitable for video animation to studies suitable for
journal publication. It will give people a chance to pioneer
future technologies today, while gaining the knowledge, skills,
and experience needed to enter the field professionally, when
serious research funding begins to grow. By helping people to
visualize nanotechnology, it will aid foresight and preparation.
How fast will home molecule hacking get off the ground? It is
hard to say, but the activity seems fun, valuable, and worth
promoting. The hardware is here, and the software is within
The Foresight Update plans to review
software tools useful for molecular design. If you come across
reviews or advertisements, or are yourself familiar with such
tools, please send us information.
The Foresight Institute will co-sponsor several events at MIT
this January in cooperation with the MIT Nanotechnology Study
Group. FI president Eric Drexler
will lead a four-day seminar on foreseeable breakthroughs
entitled "Nanotechnology and the Limits of the
Possible." Topics will be covered roughly as
follows--Monday: technical basis of nanotechnology, nanomachines,
replicators. Tuesday: nanocomputers, thinking machines.
Wednesday: applications such as cell repair machines for advanced
medicine and life extension; space hardware. Thursday:
consequences for war and peace, liberty; the challenge of
Drexler will also lead a seminar on "Hypertext Publishing and the
Evolution of Knowledge, " exploring FI's assertion that
a suitable hypertext publishing medium (not just "a
hypertext system") can speed the evolution of knowledge by
aiding the expression, transmission, and evaluation of ideas, and
that development of such a medium is a goal of first-rank
importance. Technical and implementation issues will be
addressed. See "Upcoming
Events" for details on these meetings.
Since May, nanotechnology coverage has appeared in the Washington
Post Magazine, The Media Lab (a book by
Stewart Brand), OMNI (an excerpt from The
Media Lab), Bloomsbury Review, The
World & I, Analog (in a lead editorial
and a follow-on article), Space World, and SFWA
Bulletin. Articles or columns are planned for the January Scientific
American and an unspecified issue of Discover.
MIT Hypertext Lecture/Discussion, leader Eric
Drexler, Jan. 11, 2 pm, room 16-310, free. Co-sponsored by FI;
see writeup in this issue.
MIT Nanotechnology Lecture/Discussion Series,
leader Eric Drexler, Jan. 11-14, 7:30 PM, Room 66-110, free.
Co-sponsored by FI; see writeup
in this issue.
"Where is the Bottom?" MIT lecture by Prof. Ed Fredkin,
speculations on physical processes at the smallest possible
length, Jan. 21, 7:30 PM, AI Lab 8th floor playroom, 545 Tech
Square, free. Sponsored by MIT NSG.
Technological Literacy, the Third National
Science, Technology, Society Conference, Feb. 5-7, Arlington, VA.
Registration $80. Sponsored/supported by AAAS and NSF, among
others. Contact AAAS, 202-326-6500.
Space Development Conference, May 27-30,
Stouffer Concourse Hotel, Denver, CO. Co-sponsored by FI.
Includes nanotechnology programming; see writeup in this issue.
Registration $60 through May 1. Contact Box 300572, Denver, CO
An initial Foresight Institute Board of Advisors has been
formed, consisting of Marvin Minsky, MIT professor and co-founder
of the field of artificial intelligence; Gerald Feinberg,
Columbia professor of physics; and Stewart Brand, founder of the
Whole Earth Review and a director of the Point Foundation.
FI is still searching for an Executive Director. Friends of
the Institute are requested to read the ad in this issue and
wrack their brains for ideas on who can fill this role. If you
think of a likely prospect, please call Chris Peterson at FI:
The L5 Society and its successor, the National Space Society, have been
increasing annual conference coverage of nanotechnology every
year since 1985; the Foresight Institute will be co-sponsoring
their next meeting in Denver over Memorial Day weekend. As this
issue goes to press, plans are still in flux, but we expect at
least two formal sessions on nanotechnology as well as the usual
informal discussions. We urge those of you who would like to meet
the FI leadership--and other FI participants--to attend. Both
technical and non-technical information will be presented. (See
the "Upcoming Events"
section for details.)
Explicit coverage at last year's conference in Pittsburgh began
with a session chaired by FI president Eric Drexler, with speakers Marvin
Minsky (MIT professor, co-founder of the artificial
intelligence field, and FI Advisor), Hans Moravec (head of
Carnegie Mellon's Robotics Lab) and James Bennett
(FI Director and VP of American Rocket Company). Drexler gave an
introduction to nanotechnology; Minsky gave a fascinating,
Bennett discussed the impact of nanotechnology on space
development: today space is valued for its properties of
microgravity and Earth observation, but with nanotechnology,
these features will become less important. Space will then be
valued for its resources, which nanotechnology will help us to
use: energy, materials, and room to grow.
Moravec examined trends in computation, pointing out that today's
supercomputers have a computational capacity somewhere between
that of a mouse-brain and that of an insect-brain. But an
extrapolation of the smooth curve of progress in
computation--starting at the turn of the century with mechanical
calculators, through vacuum tubes, transistors and so
on--indicates that we should expect supercomputers equivalent to
a human brain shortly after the year 2000, and personal computers
of that power shortly after 2020.
In a later session on nanotechnology, Drexler went into detail on
mechanical nanocomputers (which if built could result in
megabrain-equivalent PCs) and cell repair technology.
Nanotechnology-oriented discussion also dominated two panels
covering the financial and legal aspects of space
settlement--reasonably enough, since the time frames for the two
developments may be comparable. It's clear that the
nanotechnology meme is making rapid progress in the space
community, thanks to conference organizers such as Dale Amon (Chair,
'87), Jill Steele (Chair, '88), and Laura Powers (Program Chair,