Why was I so excited about the FNANO10 conference on self-assembly, given that self-assembly is not molecular manufacturing?
Self-assembly is a way of making large structures out of small pieces, by designing the pieces so that random ("Brownian") motion will jiggle them into place. DNA self-assembles very nicely into quite large structures - as big as 100 nanometers, almost bacteria-sized - almost big enough to see with an ordinary microscope.
A problem with self-assembly is that the pieces have to use their own structure or other properties to template their assembly. That limits and complicates the design. So templates, created by other means, are sometimes used to guide the self-assembly.
With a template, larger and more intricate structures can be built than by pure self-assembly. For example, self-assembled monolayers, using a flat surface as a template, can produce square centimeters of high-precision arrangements of molecules. Lithographed templates can arrange DNA structures in arbitrary orientations over large areas (this is cutting-edge stuff).
Molecular manufacturing uses nanoscale tools to guide the fabrication of more tools. Once you can computer-control those tools to make a programmable range of shapes, you can make more tools (both in quantity and variety) than you started with.
Self-assembly is already using templates, and templated self-assembly is pretty darn close to molecular manufacturing. Once the templates become programmable and are built using the same processes and building blocks that they guide... then that is molecular manufacturing.
With so many advances on self-assembly, it seems pretty clear that just a few years from now, we'll have primitive molecular manufacturing. More steps will be needed, of course, to design a full nanofactory and get it working. But the conceptual and practical hurdles are falling fast.