The separation of different kinds of atoms or molecules from each other is a fundamental technological problem. Current techniques of resource extraction, which use the ancient paradigm of the differential partitioning of elements into coexisting phases, are simple but extremely wasteful and require feedstocks (`ores') that are already anomalously enriched. This is impractical for pollution control and desalination, which require extraction of low concentrations; instead, atomistic separation, typically by differential motion through semipermeable membranes, is used. The present application of such membranes is seriously limited, however, mostly because of limitations in their fabrication by conventional bulk techniques. The capabilities of biological systems, such as vertebrate kidneys, are vastly better, largely because they are intrinsically structured at a molecular scale. Nanofabrication of semipermeable membranes promises capabilities on the order of those of biological systems, and this in turn could provide much financial incentive for the development of molecular assemblers, as well established markets exist already. Continued incentives would exist, moreover, as markets expanded with decreasing costs, leading to such further applications as remediation of polluted sites, cheap desalination, and resource extraction from very low-grade sources.