In the last few days of technical discussion, several people have asked for more detailed information on what molecular manufacturing is and how it is supposed to work. One even asked for a short article summarizing this.
The trouble with that is that there's far too much information to put into a short article, a long article, or even a book. The best we can do is to provide an overview and answer questions.
The goal of molecular manufacturing is to use molecularly precise devices, configured as computer-controlled manufacturing systems, to make a wide range of high-performance products, including more manufacturing systems.
There are several reasons why atomically precise, sub-micron machines are desirable:
1) High functional density
2) Very high performance (scaling laws)
3) Maintenance of dimensional tolerance between manufacturing generations
4) Predictable shape -> easier automation of manufacturing
5) Zero-wear sliding interfaces (contrast with MEMS)
6) Zero static friction, very low dynamic friction
To build structures with every atom in its engineered position requires extremely high precision. This precision can be supplied by well-designed molecular machines, since all atoms of any given type are identical.
There are several reasons why multi-scale, heterogeneous nanosystems are desirable:
1) Easier interface with the macro-world
2) Tighter integration of diverse functions
3) Build complete products, not just components or materials
To build heterogeneous nanosystems requires a lot of information to be delivered to the nanoscale. This information can be delivered through computer-controlled actuators. Programmable manufacturing can build manufacturing systems with integrated nanoscale computers for even higher information delivery rates.
This overview is both high-level (stratospheric!) and incomplete. There are other desirable features, such as high-performance materials -- and nanoscale manufacturing systems built out of high-performance materials will find it easier to build high-performance materials. But perhaps there's enough here to get the idea: nanoscale programmable manufacturing systems will have lots of good properties, which will feed back to enable better nanoscale programmable manufacturing systems.
There are two main ways to do molecular synthesis under direct computer control to make engineered precise structures. One is biopolymer synthesis: DNA, protein, etc. The other is proximal probe chemistry.
DNA synthesizers can make any desired sequence of DNA by adding small molecules one by one to a growing chain. In this process, the time sequence of operations is important; the sequential choice of which molecule to add next is what controls the information content of the result. Protein synthesizers use basically the same technique, as does Chris Schafmeister's process for building stiff polymers.
In the biopolymer systems other than Schafmeister's, the structure is not directly built into the molecule. Instead, the molecule folds after manufacture to form the desired structure. DNA structures are fairly easy to predict, at least in comparison with protein structures -- but even protein structures are easier to predict for engineered sequences than for natural proteins.
Nadrian Seeman has recently built a machine that is made of DNA, is programmed by adding more DNA strands, and builds one of four product strands as desired. This is a very impressive proof of concept, although the machine cannot (yet!) build strands as complex as the strands that it is built out of.
Proximal probe chemistry controls the spatial location of reactions. As far back as 1994, the Aono Group in Japan was transferring individual silicon atoms from place to place on a silicon surface, under automated control, with error correction. A variety of individual molecules have been split and joined by scanning probe microscopes. It seems only a matter of time before three-dimensional deposition will be achieved. It should be pointed out that the scanning probe microscopes used to date have all been large, but conceptual designs exist for sub-micron probes (not just the tip, but the positioning system as well).
In proximal probe chemistry, the temporal sequence of operations is less important than the spatial position. The structure will be built directly, according to where the deposition reactions take place.
Is it possible that construction systems other than molecular machines could place enough atoms with enough precision to build useful products? It's too early to rule out that possibility, but there are some reasons to think it would be difficult. Without some kind of computers and actuators at the nanoscale, it would be difficult to deliver enough information to build intricate heterogeneous products. Without direct manipulation of molecules, the range of structures that could be built would be limited, reducing design flexibility. Without precise machines, or at least precise templates, it would be difficult to place the product molecules exactly where they needed to go.
Are there other ways than polymer buildup or proximal probe deposition to build precisely designed molecules using molecular machines? This seems likely. There are several microscopic techniques that can image atoms, and this implies that the atoms might be manipulated, and perhaps a construction technique could be based on this. It's even conceivable that some self-assembly technique could be developed that was flexible, precise, and fast enough to build products approaching the level of intricacy and performance promised by molecular manufacturing.
Do complete molecular manufacturing designs exist? No. Anyone asking for end-to-end laboratory demonstration, or even complete plans, will have to remain unsatisfied for now. The broad goal exists, along with a lot of supporting analysis to indicate that molecular machines really can be powerful enough to be worth working toward, and really can implement manufacturing systems. (We know that the polymer-based systems can work, because that's what your body uses. Positional-controlled chemistry seems more obvious to some people (computer scientists, mechanical engineers) than to others (molecular biologists), but it should work just as well if not better.
For more details, hunt through the literature, starting with Nanosystems. Some good basic articles are "Developing Molecular Manufacturing," Molecular Manufacturing: What, Why and How," and "What Is Molecular Manufacturing?" And feel free to ask questions.