Brownian motion and robots
From Wise Nano
All biological structures, including machines like ATP synthase or the flagellar motor, are heavily influenced by Brownian motion. Because they are floppy, they continually rearrange themselves a bit, forming less a shape than a cloud. This leaves lots of room for these structures to interact with each other in complicated and complex ways. These structures are well suited to discovery by evolutionary methods.
Robotic nanomachines will be built out of relatively stiff materials, intended to hold their shape: each atom in such structures may wiggle by much less than one atomic radius. The complexity of interactions between shapes will be reduced to the point that humans will be able to engineer their function.
Our experience with computers has shown that loss of complexity may be more than compensated by ease of engineering. Most computers, even million-dollar supercomputers, are built with inefficient general-purpose hardware rather than special-purpose designs. And no one uses analog computers anymore, even though analog circuits can theoretically be much more compact than digital. The same is arguably true of nanomachinery: when it comes time for humans to build it for human purposes, we will find it easier to engineer simple designs than to evolve and test complex designs.
One concern often raised about Brownian motion is that it will make robotics unreliable. At first sight, it seems that parts will have to be built impossibly stiff to avoid wiggling away from each other. But simple compression structures can keep them in line. Because atomic bonds are so flexible, some slop is inevitable--but for the same reason, it is unimportant. The whole system can wiggle as long as its parts don't slip past each other improperly, and parts can certainly be built stiffly enough to prevent that.
Another criticism is that robots will be too stiff, unable to wiggle in a way that would allow using biology's clever engineering tricks. But it is easy to make things less stiff; a nanoscale part can be made hinged or flexible simply by building it thinner. If robot engineers want to use biology tricks, they will be able to.
One function proposed for stiff robots is "grabbing" biomolecules and other floppy molecules, in a way analogous to antibodies. The stiff binding site may be carefully designed to correspond to an optimal tight binding condition, making this state very attractive to the target molecule. At first sight, there would seem to be a mismatch between stiff and floppy: the floppy system can reconfigure itself even after it is bound, so would seem to be entropically favored. But counting up the total degrees of freedom before and after binding shows that the stiff/floppy system has fewer DOF both before and after, and the change is the same in both systems. If the floppy molecule is not floppy enough to fit into a stiff final cavity, a hinged section (driven by Brownian motion) can be built into the cavity. In short, it appears that there shouldn't be a performance loss from using stiff binding cavities.
