While I am happy to see the increasing interest in the subject, I was
somewhat disappointed by your special report on nanotechnology (March/April
1999 Technology Review).
Mark Reed summarized one common thread when he said "There has been no experimental verification for any of Drexler's ideas." Presumably, this includes the proposal to use self replication to reduce manufacturing costs. The fact that the planet is covered by self replicating systems is at odds with Reed's claim.
Self replicating programmable molecular manufacturing systems, a.k.a. assemblers, are not living systems. This difference lets Reed argue that they have never before been built and therefore their feasibility has not been experimentally verified. Of course, this statement applies to anything we have not built. Reed has discovered the universal criticism. Proposals for a lunar landing in 1960? Heavier-then-air flight before the Wright brothers? Babbage's proposal to build a computer before 1850? No experimental verification. Case closed.
It is more difficult to do something new if we refuse to think about it. While thinking is hard, today we have new tools and new methods to aid us: computational chemistry and molecular modeling. Those of us who have both thought about molecular manufacturing and modeled the various components that could be used in such systems have reached a conclusion: we should be able to arrange atoms and molecules in most of the patterns consistent with natural law. It looks hard, and we're not sure how long it will take, but the conclusion is very clear. Feynman reached the same conclusion in 1959 (see http://nano.xerox.com/nanotech/feynman.html). He, too, lacked experimental verification.
We know the goal, we know it's feasible, we don't know how long it will take or exactly which approach is best. Let's investigate the subject: theory and experiment will both contribute, and neither should be denigrated. Each has its particular strengths and its particular weaknesses. By building on the strengths of both, we can develop this new technology more quickly.
Neither Reed's evasion nor the other unhelpful discussions of "real" (i.e., near term experimental) research help us to evaluate this central issue.
The second issue is the utility of positional control at the molecular
scale. Either (1) positioning molecular parts will at some point play a
significant role in future manufacturing systems, or (2) this capability
will never play a significant role in manufacturing. The article would
have us believe the latter is true, but provides as evidence only various
restatements of Reed's evasion.
We have excellent experimental evidence that such positioning is feasible (though this work is still very much in the early stages of development), and we further have excellent theoretical evidence that such control is useful (see, for example, the theoretical and computational studies of the utility of the ethynyl radical in selectively abstracting chosen hydrogens at http://nano.xerox.com/nanotech/Habs/Habs.html). The most straightforward conclusion is that holding, positioning and assembling molecular parts will at some point be a remarkably useful addition to our repertoire of synthetic capabilities (see, for example, http://nano.xerox.com/nanotech/CDAarticle.html).
Despite this, the Technology Review gamely argues that "... after years
of hearing grandiose speculations of a brave new nanoworld, researchers
say it's time to let the science overtake the fantasies." Indeed, Drexler
was the first to argue that self replication combined with molecular scale
positional control would play a major role in manufacturing at some point
in the future. As Technology Review said, he's been saying this for years,
and repeating it quite stubbornly. He's also very carefully said it will
take some time to develop these capabilities, and no one has said we'd
have such capabilities today (see http://nano.xerox.com/nanotech/howlong.html).
Why, then, is this argument being advanced "against" nanotechnology?
Proposing what is new
The most common method of proposing something new is to look at what has already been done and make some incremental changes in it. Because we have done A, and because A is similar to B, we might be able to do B. This common method creates proposals that "should" be experimentally accessible within a few years. If, after a few years, we find we have been unable to accomplish them it is reasonable to ask whether the proposal was at fault. Perhaps A was not "really" similar to B.
There is another very different and very much more powerful approach: examine the basic laws of physics and advance a proposal which is feasible with respect to those laws. The proposal is independent of existing experimental capabilities. Given an adequate understanding of basic physical law we should be able to advance proposals of great value which are beyond the 2 to 5 year time frame that seems typical of most "long term" thinking in this culture.
Drexler called this backward chaining. It is similar to the "Horizon Mission Methodology" proposed by John Anderson at NASA. A more specialized version of this approach is common in the world of chemical synthesis: retrosynthetic analysis (which earned Elias Corey the Nobel Prize). In retrosynthetic analysis, the molecule to be synthesized is "taken apart" in small steps, where the small steps are the reverse (the "retro") of known chemical reactions that could synthesize the target molecule from appropriate precursors. By working backwards from the desired result, retrosynthetic analysis provides a very powerful tool for developing new syntheses of molecules that might previously have been viewed as difficult or impossible to synthesize. (See the introduction to retrosynthetic analysis by Ott).
Whatever you call it, the basic idea is to move beyond simple short term extrapolations of existing capabilities and generate more fundamental (though quite possibly more long term) research targets. Having once determined that a goal is feasible and valuable, it is then possible to work backwards from the goal to determine how to develop it. This is an entirely different approach from the more traditional concept of incremental advance.
The basic advantage of this approach is that it opens up to investigation ideas that are beyond the scope of existing experimental capabilities. The basic disadvantage is that there is no implied claim that the new idea can be developed in the near term: the developmental pathways might well take significantly longer than 5 years. In the case of molecular nanotechnology estimates range from 10 to 50 years, reflecting a basic uncertainty in how long it will take to develop.
Konstantin E. Tsiolkovsky correctly predicted at the turn of the century that flight to the moon was feasible. He based this on Newtonian mechanics, the strength of materials of the day, and the energy that could be produced from chemical fuels. He did not say when or how this would be accomplished, and certainly did not predict that John F. Kennedy would make a lunar landing a national priority. Claims in the 1950's that flight to the moon would never happen were wrong, despite the fact that decades had elapsed since Tsiolkovsky's writings. The passage of time does not falsify a feasibility argument based on fundamental physical law.
Experimentalists are primarily interested in (a) their current experiment and (b) their next experiment. Any proposal that involves neither their current experiment nor their next experiment does not command their attention as experimentalists. Backwards chaining can easily produce goals that will take longer than this, occasionally resulting in the (erroneous) claim that "I cannot do that, therefore it cannot be done." (Reed's evasion is a milder form of the same disease).
We now turn to the accusation that "A few, however, maintain that they have it all but figured out..." and have been "...describing in precise detail how nanomanufacturing will work -- and change the world...." This is, of course, the classic "reductio ad absurdum" approach. What the "few" (actually, quite a few) have figured out is that the ability to inexpensively make most structures consistent with physical law (by using self replication and molecular positional control) will indeed fundamentally change our world. The subsequent accusation that such capabilities are known in "precise detail" is a failure to recognize how theoretical and computational investigations proceed.
If we wish to investigate whether (for example) molecular bearings are feasible, it is helpful to investigate some specific molecular bearings (see, e.g., http://nano.xerox.com/nanotech/bearingProof.html). The general conclusion that molecular bearings are feasible would be more difficult to support if we refrained from investigating and computationally modeling specific instances. This is very different from claiming that the specific bearings will (or will not) be used in future molecular machines. The latter claim, a precise prediction, is very difficult to make with any degree of reliability. The claim that molecular bearings are feasible is a much broader claim which can now be made with great confidence, thanks to theoretical and computational investigations that include analysis and modeling of specific designs.
We therefore reach the following conclusions: there is general agreement that self replication and molecular positional control are both feasible. At some point in the future we should be able to combine these two capabilities to build self replicating systems able to inexpensively synthesize a remarkably wide range of structures. There is general agreement that this development is beyond the capabilities of existing experimental methods. We will therefore have to fundamentally advance our capabilities to achieve this goal. We can either further investigate both this goal and the many possible routes by which it can be achieved, or we can abandon such investigations. As the goal and the major part of the developmental pathways are (today) experimentally inaccessible, such investigations must necessarily draw heavily on theoretical and computational work.
Consider the following statements:
We will never be able to make artificial programmable self replicating
systems or molecular positional devices. Therefore, we can safely ignore
the consequences of such developments and should not think about them.
At some point in the future we will develop artificial self replicating
systems that use molecular positional control to synthesize an unprecedented
range of new products. This will have a major impact on our world, and
we should therefore investigate these systems and their potential consequences
There is little if any support for the first statement, and excellent support
for the second. The course of action we should follow seems obvious: further
A few asides:
Smalley's comments on the infeasibility of a Universal Assembler have been discussed elsewhere (see http://nano.xerox.com/nanotech/hydroCarbonMetabolism.html). Basically, there is general agreement that a device able to make absolutely everything is infeasible. This does not change the fact that we should be able to make "most" structures consistent with physical law. Smalley said: "Most interesting structures that are at least substantial local minima on a potential energy surface can probably be made one way or another." There is agreement on the core issue: we should (as Feynman said) be able to "Put the atoms down where the chemist says, and so you make the substance." (see http://nano.xerox.com/nanotech/feynman.html).
The argument that "real" nanotechnology will be based on self assembly
and not molecular positional control is a reflection of the fact that self
assembly is a very powerful synthetic method that is feasible today and
able to make a reasonably wide range of interesting structures. However,
it is generally recognized that self assembly cannot make anywhere near
the range of structures that are possible. As an example, no one has proposed
a method of self assembling diamond and some chemists have argued that
this is in fact impossible. As it is quite likely that self assembly
will prove critical in the development of molecular nanotechnology, a program
to develop the latter should include major support for the former. Those
who support research in molecular nanotechnology also support research
in self assembly.
The article claims that "practical nanotech is here. It is a modest start." If we define nanotechnology as the ability to inexpensively make, with molecular precision, most structures consistent with physical law; then clearly nanotechnology is not here. The fact that the same term is used in different ways by different people creates confusion, so it's necessary to be careful in how we define our terms. When the term "nanotechnology" might be ambiguous, the more specific "molecular nanotechnology" or even "molecular manufacturing" can be used to clarify what's meant.
As for modesty: the value of aspiring to modest goals is entirely unclear. Nature does not rank what is possible by our opinion of its modesty. A number of immodest goals seem feasible: vastly greater computational power, inexpensive access to space, longer and healthier lives, greater material wealth for all, and the ability to roll back environmental damage. The old saw that "if it seems too good to be true, it probably is" is blinding us to the obvious: truth in the physical sciences simply doesn't care about what we want, one way or the other. If the evidence supports the conclusion that we can all be, by today's standards, rich in material possessions then so be it. As the trend over the past many centuries has been towards greater and greater material wealth, there seems no reason a priori to believe that the continuation of this trend is unlikely. Turning away from a goal because it is too valuable is a very peculiar strategy.
"The Zurich work reflects a deeply entrenched -- and strongly Swiss -- belief in mechanics." This explains it. My Swiss great grandfather (whose birth record I recently examined at the Staatsarchiv in Basel, Switzerland after speaking at IBM's Zurich research lab) is no doubt responsible for my interest in molecular mechanical machines.