I'm sitting here at the Foundations of Nanotechnology conference, listening to a talk by Erik Winfree of CalTech. He's doing an excellent job of presenting molecular manufacturing ideas in a way that's acceptable to experimentalists. Not that he's explicitly supporting molecular manufacturing, but his work requires the ideas. While trying to build computational systems, he is working toward general-purpose manufacturing.
He started by invoking the goal of putting information into a chemical soup and getting a product. Basically, the molecular manufacturing vision. He acknowledged that we're not near that today... but the direction of his talk is clearly one of: Let's see how close we can get to that vision.
He called for getting beyond pure self-assembly: "You can do some things with passive self-assembly; you can do them much better if you're able to implement molecular motors."
He talked some about the ability to implement computer algorithms (Turing machines) with chemical reaction networks: possible in theory with arbitrarily low (but not zero) error rates. But it can take a large vat to implement a complex algorithm.
Now he's talking about DNA strands being used for computing. I was hoping he'd be talking more about fabrication of structures, but I guess in a way he is: computation with molecules does, physically, fabricate structures. And the fabrication of structures
Oh, good, now he'll talk about controlling growth using passive assembly. .... Oh, very cool. He's talking about how self-assembling tiles with carefully designed matching edges can implement any algorithmically-describable shape. And, of course, DNA can implement tiles with lots of different edge configurations.He has a slide titled "Theory of Active Self-Assembly (of Molecular Robots)." It includes the statement, "Any algorithmically-describable shape can be assembled _efficiently_." He talked about shapes with a billion components and said they could be assembled exponentially. He acknowledges "This is something for the future" since we're still at the stage of molecular motors rather than molecular robots.
It will be very important to have physical systems that include large-ish blocks that self-assemble reliably. There's a whole lot of incorrect assumptions running around, even among cutting-edge molecular self-assembly people, professional chemists and physicists. Many of them are just not comfortable with this kind of system, and would not expect them to work, and will not be able to work on them until they drop some of their (mis-applied) theory. It will require experimental demonstration to shift the paradigms.
Winfree's work appears likely to produce tangible, even flashy results; to produce some results with today's tools, with gradual steps all the way to programmable molecular robots driving assembly mechanically.
If molecular manufacturing is not developed by a targeted program, it will require a series of paradigm shifts. Paradigm shifts are rare and difficult, but Winfree's work seems likely to fuel a few key shifts.