Polymer-based synthesis
From Wise Nano
Biology uses polymers--proteins and nucleic acids--folded into functional systems. An appropriate collection of these can produce a complete set of new construction polymers, plus other substances (including lipids) necessary to contain the works: in other words, a metabolizing, homeostatic, reproducing, evolving cell.
Molecular manufacturing does not require this much functionality. External control of environment would eliminate need for homeostasis. Supply of purified feedstock would greatly simplify metabolism. Dedicated fabrication systems would eliminate the need for self-reproduction.
In a cell, the machinery is intimately involved in its own maintenance, metabolism, and continual reconfiguration. This imposes restrictions on the construction: in general, a cell's machinery must be able to be broken down and re-used efficiently. If this requirement is eliminated, then machines can be made stiffer, more deterministic, and more reliable (though less flexible and complex). And this reduces the need for self-repair.
The point of this argument is that a polymer-based manufacturing system, even one capable of building duplicate manufacturing systems, need not be nearly as complex as a cell, or include nearly as many parts. In fact, a mitochondrion would be a closer comparison, though the complexity would probably be closest to a ribosome.
So how might a polymer-based molecular manufacturing system work? For maximum simplicity, it would have to be controlled by an external computer. So it would need actuators. (Control by a chemical tape, as the ribosome uses, is slow and cumbersome, and doesn't scale well.) To achieve exponential manufacturing, it would have to be able to build more manufacturing systems; this means that it would have to catalyze or control the synthesis of the same polymer it was build out of. If a linear polymer could be engineered to fold into actuated machine systems, that might be sufficient. In that case, it would just have to handle one end of a chain, as a ribosome does. Alternatively, it might be built of branched polymer, and use more general actuation to deprotect ligation sites on the workpiece.
The important point is that its action should be programmable, and that its product should be engineerable: in other words, it should be possible to translate from a structural or functional design to a sequence of operations with a minimum of R&D. It should be able to produce a wide range of product reliably; this will greatly accelerate the development of higher-level capacities.
It goes without saying that the product should be digital, atomically precise: it should be able to build something exactly the same configuration as itself (once thermal noise is averaged out).
