This paper is available on the web in Microsoft Word for Windows 2.0 format at ftp://quake.unr.edu/pub/gillett/nearterm.rev.
Writings on molecular nanotechnology (MNT) typically describe the wonders (or alternatively the horrors) that will be available once a mature MNT is available. Left unspecified, however, are any details on the developmental pathways that might be followed to achieve MNT. An enormous investment must be made for a distant and somewhat ill- defined payoff, which is not a situation conducive to rapid development. This is especiallly cogent because historical experience suggests that full MNT development will be considerably more difficult than initial enthusiasm suggests. In particular, system failures related to undesirable emergent properties, which seem difficult to avoid with an ensemble containing an extraordinarily large number of molecular moving parts, seem to have been little considered.
Hence, nearer-term economic drivers must be identified: what intermediate applications might justify the still-substantial capital expense required to develop a "partial" MNT? Refinement and extension of such applications could then provide the financial incentives for further evolutionary development of MNT, much as continued economic incentives have driven the incremental but rapid improvement of computer hardware in the recent past.
A major suite of such early applications lies in the fabrication of nanostructured materials, including such things as perfect crystals, polymers, high-density electronic chips, dispersed metal clusters for nonlinear optics, high-temperature superconductors, thin- film photovoltaics, catalysts, and semipermeable membranes. Such structures involve MNT, because at least ultimately they have "a place for every atom, and every atom in its place." Having no moving parts, however, such materials are likely to be much more reliable and need not be perfectly assembled to work at all. Indeed, rather than being vulnerable to point failures, in many cases they should simply degrade gradually. Undesirable emergent properties also seem less likely.
Molecular assemblers for fabricating such structures, which might be termed "molecular looms", also seem to be easier to develop than full-blown assemblers, because the "loom" could work in a highly controlled environment and need not be general-purpose or even programmable, at least to begin with. Indeed, if sufficiently reliable (i.e., if enough product can be generated between failures), the "loom" need not even be self-replicating. A biological analog of such a loom is the molecular mechanism in a hair follicle that spins out hair. Of course, the experience gained in building such primitive assemblers should then apply directly to building more complex ones.
Finally, nanostructured materials have billions of dollars of already identified applications, such that many short-term financial incentives exist for pursuing improved methods of their synthesis. One example is fixed catalysts, which would greatly profit from molecular design and assembly: consider surface catalysts, which rely on details of atomic structure at an interface, or "volume" catalysts such as zeolites, whose selectivity is due to differences in molecular access to the active site due to steric hindrance. For another example, semipermeable membranes also promise a huge number of applications. They seem critical to cheap practical fuel cells, for example, by allowing the intimate but controlled mixing of fuel and oxidizer, and they also have a staggering variety of applications related to resource extraction and pollution control, as I elaborate elsewhere.