Some thoughts on tipping points in manufacturing improvement...
Reliability, precision, and complexity are key aspects of any manufacturing operation. For example, a block of wood can, with difficulty, be whittled into a cylindrical shape to make a peg to fasten things together. With a lathe, the product can be made far more precisely. Until a couple of centuries ago, precision was a function of human labor and skill. The Industrial Revolution was enabled, at least in part, by methods that allowed precision to be a function of machinery. When machines could make products more precise than themselves, the precision of the product was no longer limited by human skill. This was the first tipping point.
Until recently, reliability has been a function of human attention. A machine might produce 99 out of 100 parts correctly, but to sort out the bad 1% a human had to test them all. A machine made of 1,000 parts could not be built without significant human labor just to test the parts. And it could not operate without more human labor to test the products and keep the machinery adjusted. A complex linkage, such as required for automation, could be developed -- but only with difficulty and skilled maintenance.
In the last few decades, machines have become reliable enough in construction and operation to allow full automation: lights-out factories. (Some of this reliability has been aided by better sensors; that doesn't change the overall picture.) When manufacturing systems can work reliably over their lifespan, productivity -- at least for simple products -- no longer depends on human labor. We're living through this second tipping point right now, with the loss of manufacturing jobs.
Most fabricators today can only build things that are less complex than themselves. No matter how many linkages and cams are designed into them, they will only be able to do a few operations, and will not be able to produce even a fraction of their internal workings. Humans are required to assemble simple products into more complex ones.
The way around this is with software. A simple and repetitive physical design can be combined with a control program of unlimited complexity to make a machine with functionality limited only by the software. A laser printer can easily print its entire blueprint. A computer-numerically-controlled (CNC) milling machine can in principle make objects more complex than any of its parts -- even more complex than the entire machine.
This is not to say that a CNC machine could build a copy of itself -- it can only build a few types of parts, not whole products. Newer rapid-prototyping technologies based on 3D printing are working toward whole-product complex manufacturing, but can't achieve the precision (e.g. smoothness) necessary to duplicate themselves. But we can see where the trend is going. When programmable machines can build entire products as complex as themselves, product complexity will no longer be limited by human dexterity. This will be a third tipping point, at least as important as the other two.
What does this have to do with molecular manufacturing? Just this: that mechanically guided bottom-up molecular synthesis is inherently reliable and precise, and is simple enough to be programmed. A covalent chemical bond is either there or it's not, and most of them are incredibly stable at room temperature. And as far as we can tell, the mechanical operations to make or break a bond can be quite simple. (Smalley and Ratner claim otherwise, but have not given any actual evidence for limits to mechanosynthesis, and have not addressed the relevant literature.)
To recap: when machines can make more precise machines, then precision can increase far beyond human capability. When machines become sufficiently reliable, then automation allows productivity unlimited by human labor. And when machines become programmable, then complexity no longer depends on human dexterity. Molecular manufacturing should be able to capitalize on all these advantages simultaneously. In fact, molecular manufacturing is perhaps the easiest way to reach the third tipping point while still taking advantage of the first two.
And make no mistake: the third tipping point -- complexity unlimited by human dexterity -- will be at least as revolutionary as the other two. This is exactly why we find it reasonable to talk about tabletop nanofactories with quadrillions of fabricators, and microscopic robots with onboard supercomputers. And this is what makes it reasonable to talk about a manufacturing cost that is essentially zero.
The fact that this involves a tipping point -- in fact, not one but three tipping points -- is significant. It means that up to a certain point, we won't see the power of this technology, but the advantages will develop rapidly once that point is reached. This is not how humans expect things to work -- we expect a linear progression. But a tipping point is nonlinear. If we don't recognize it in advance and plan for it, it will take us by surprise.