Molecular Manufacturing: Where’s the progress?

by Eric Drexler on 2009/12/19

Part 2 of a series on the history and prospects of advanced nanotechnology concepts,
prompted by the upcoming 50th anniversary of Feynman’s historic talk,
“There’s Plenty of Room at the Bottom”.


John Stewart Mill
John Stewart Mill
Debugging defects in human thought

As cognitive psychologists know, we human beings suffer from multiple, systematic cognitive biases, aberrations of intellectual vision that can be corrected, in part, when we recognize their operation in our own minds. One is a bias toward expecting causes to resemble their effects, the implicit, default assumption that like causes like.

This bias has, I think, impeded full recognition of the most important line of progress toward molecular manufacturing, and has likewise impeded recognition of where that progress can lead. This is a line of progress that I’ve advocated since the beginning, and that I helped to kick-start: engineering complex, self-assembling molecular systems as an indirect approach to implementing systems that will look very different. That difference, of course, is what triggers the cognitive bug.

The cognitive bug

John Stewart Mill described the like-causes-like bias as

…the most deeply-rooted, perhaps, of all which we have enumerated: one which not only reigned supreme in the ancient world, but still possesses almost undisputed dominion over many of the most cultivated minds….This is, that the conditions of a phenomenon must, or at least probably will, resemble the phenomenon itself.

Some cognitive psychologists (Nisbet and Ross) view this as an aspect of the famed and treacherous representativeness heuristic.

Current progress (obscured by that cognitive bug)

This bias tends to distort perceptions of progress in molecular manufacturing because, at this stage of development, steps toward high-throughput atomically precise manufacturing (HT-APM) do not look like HT-APM. Filtering one’s comprehension through the representativeness heuristic can make such progress virtually invisible, no matter how large it may be.

The progress I have in mind centers on advances in atomically precise fabrication by chemical and biological means, and these advances now have reached a level that places the implementation of first-generation artificial APM systems within reach. These, however, also won’t resemble slick, large-scale, general-purpose HT-APM systems. Instead, they will support the implementation of more-capable second-generation APM systems that can support a fast design-build-test cycle and thereby enable a well-focused and well-organized develpment program to rapidly ascend a ladder of technologies leading to HT-APM.

Available technologies now enable the design and fabrication of intricate, atomically precise nanometer-scale objects made from a versatile engineering polymer, together with intricate, atomically precise, 100-nanometer scale frameworks that can be used to organize these objects to form larger 3D structures. These components can and have been designed to undergo spontaneous, atomically precise self assembly. Together, they provide an increasingly powerful means for organizing atomically precise structures of million-atom size, with the potential of incorporating an even wider range of functional components.

Lines of advance (obscured by the same bug)

Unfortunately, the representativeness heuristic strongly opposes recognition and exploitation of the power of this emerging technology base.

The nanometer-scale objects that I mentioned above have nylon-like backbones that link and organize an extraordinarily diverse set of molecular components to form structural elements, electronic devices, and machines. The problem is that because they are traditionally called “protein molecules” their nature is obscured by a powerful association with food. The frameworks have a similar representativeness-heuristic problem: “DNA” makes one think of genetic information in cells, but structural DNA nanotechnology uses it as a construction material.

We now have in hand the engineering materials for a new, breakthrough class of nanosystems, yet the bug in our minds whispers “meat” and “genes”. And even in more sophisticated minds, the biological origin of the these materials encourages the seductive idea that their engineering is a task that can be left to biologists. Developing complex, functional systems, however, is quite unlike studying complex, functional systems that already exist. In science, nature provides the pattern. In engineering, human beings provide the pattern. The difference in tasks and mindsets is profound.

In my view, these problems of perception and organization are the chief obstacles to more rapid progress in developing molecular machine technologies on the critical path to fulfilling the promise that launched the field of nanotechnology.


See also:

Recent landmarks in atomically precise self-assembly:

Directions in atomically precise fabrication:

Studies of advanced atomically precise fabrication:

{ 6 comments… read them below or add one }

Guy December 19, 2009 at 11:25 pm UTC

The perceptual bias you point out is especially worrisome as science and technology shrink the gap between what is necessary for a full blown molecular manufacturing platform and what resources are available commercially and informationally. DNA sequence builders are widely available as are several software packages to drive them.It isn’t out of the question for some sufficiently brilliant person to combine Ralph Merkle’s work on diamondoid tool-tip molecules with active DNA structures to produce a rudimentary programmable diamondoid synthesizer. Even if it only built a limited set of diamondoid structures it would still be a giant step that conceivably could be pulled off in a basement lab or on a kitchen table.

Joel December 23, 2009 at 12:19 am UTC

I really enjoyed this article although I don’t have much background in nanotech.

It reminds me a bit of the research in search algorithms I am doing my thesis on. Search algorithms often employ an objective function which is a heuristic of progress to some sort of predefined goal. The problem is that, just like you say, the stepping stones along the way to the objective may not resemble the objective at all. When this happens, search tends to converge to some local optima (just as technological progress might slow if resources are allocated to the wrong areas of research).

Anyways, thanks for the great quotes, I had never heard a formal description of the like-causes-like bias!

If you are at all interested, here is a link to a conference paper that expands on the idea a bit (in the area of evolutionary algorithms, not nanotech…):
http://eplex.cs.ucf.edu/papers/lehman_alife08.pdf

Eric Drexler December 23, 2009 at 3:56 am UTC

@ Guy — My reply to your comment grew into this post.

Eric Drexler December 23, 2009 at 4:33 am UTC

@ Joel — The idea of “novelty search” described in your paper is certainly novel, to me, at least. The criticism you mention (that the process looks like exhaustive search) isn’t really answered by the observation that many sequences of actions collapse into a smaller number of net behavioral outcomes. Although this may give a large speedup, process (if I understand it correctly) still looks like an exhaustive search over the collapsed behavior space.

A mixture of strategies might have advantages: Maintain a population of candidate partial solutions, and split the computational investment between using a novelty-seeking approach of the sort you describe (as a way to escape from local optima), while simultaneously using an objective-function based bias to focus more effort on looking for novelty around intermediate solutions that score better.

Tabu search is an effective but quite different method that explores regions away from the best currently known solution.

Steve Witham February 16, 2010 at 8:16 am UTC

Eric–

So we technophiles have our own ick factors to get over.

This is a little bit of a “curse the darkness” version of what you want to say, though. Do you have a purely “light a candle” version? A brochure or poster. Getting Started in MesoNanoTech! What the proteins are like, what the DNA connectors are like, how the connectors are put on the proteins, how complex are the structures that can self-assemble. How do you fish it out of the bath and get it to do something? What equipment and supplies do you need, who sells it, what labs already have the equipment, who’s actually doing it?

(from http://tinyurl.com/6h5ped )
During a break, Myhrvold announced that he had just bought a CAT scanner, on an Internet auction site.

“I put in a minimum bid of twenty-nine hundred dollars,” he said. There was much murmuring and nodding around the room. Myhrvold’s friends, like Myhrvold, seemed to be of the opinion that there is no downside to having a CAT scanner, especially if you can get it for twenty-nine hundred dollars.

Not all biologists are studiers rather than builders. I think to synthetic bio and diybio people, biology is just a prerequisite to getting your hands on stuff. While they think they want to use cells, I’ll bet some of them wouldn’t mind the finished product coming straight out of the sequence-building machine. Michelle Khine’s shrinky dink microfluidic technology isn’t nano but it’s definitely makerly.

They said “ick” when I started to meso ’round with mesonanotech.

Eileen May 2, 2013 at 2:29 am UTC

I just wished to take a couple of minutes and inform you that I
enjoyed the write-up. I truthfully don’t think that many people know exactly how much effort that goes into having a good website. I know that this is sort of random but it really bothers me sometimes. At any rate really good site.

{ 7 trackbacks }

Leave a Comment

Previous post:

Next post: