Gradual Nanotech, Abrupt Impact

In a recent blog entry, we wrote:

"What level of nanotech development are they referring to? It might be only today's nanoscale technologies; it might be up to the second or third generation of advanced nanotechnology; or it might include the implications of exponential general-purpose nanofactory production. Clearly, the differences are crucial."

This got us thinking about a potentially confusing aspect of our message. On the one hand, we see molecular manufacturing as a natural and inevitable outgrowth of today's nanoscale technologies. On the other hand, we see its impacts as being abrupt and disruptive. Why the apparent discrepancy?

Nanoscale technologies are rapidly gaining more control of the nanoscale, building more complex devices (such as this one that includes an actuator, a bearing, and several levers), doing more intricate chemistry under direct control (scanning probe chemistry just got a lot more flexible), and supplying more information to the manufacturing processes.

The recent announcement of "nanomanufacturing" as a goal by the National Nanotechnology Initiative highlights this trend. Nanomanufacturing is one step below molecular manufacturing. It aims to build precise nanoscale devices, but it does not specify that the devices should be built by nanoscale machines. This is a difference so small that it sounds legalistic, which illustrates that nanotechnology is moving toward molecular manufacturing. However, the difference is actually quite important in terms of implications, because it is nanoscale manufacturing systems that will make molecular manufacturing really take off.

Development of a nanoscale fabrication system capable of building copies of itself will enable two things: a vast increase in the amount of intricacy that can be built into nano-products, and a rapid scaleup of manufacturing capacity. These capabilities will, for the first time, allow complete large products to be built with the advantages of nanotechnology in every component. This is the source of the revolutionary implications.

To build a nanoscale device, information must be delivered to the nanoscale. Today's nanoscale technologies use a variety of clever methods to do this. Complex molecules can be built and mixed together to self-assemble. Reaction parameters such as temperature and reactant concentration can affect the product. Patterns can be built on substrates to guide the placement of nanoparticles.

Each of those methods, however, is limited in the amount of information it can deliver. The total amount for most processes and techniques is probably measured in kilobytes. This is enough to select from among thousands of potential outputs -- very useful in comparison with earlier manufacturing techniques. By contrast, specifying useful products (as opposed to components that will require subsequent traditional manufacturing steps) will require megabytes or gigabytes of data. To deliver that much data to the nanoscale requires nanoscale machines to handle it, converting the data into nanoscale manufacturing operations.

Nanoscale fabrication machinery is important for another reason. It raises the possibility of exponential manufacturing: using a manufacturing system to make double the capacity, then repeating that process a few dozen times until the manufacturing capacity has scaled up to kilogram-scale production of nano-structured components. Ten more doublings reaches ton-scale, twenty reaches kiloton-scale, and so on. Exponential manufacturing will make manufacturing capacity non-scarce: it will be possible to build as many factories as desired.

The other advantages of molecular manufacturing, such as scaling law advantages and the benefits of atomic precision, have been discussed on this blog and in our monthly science essays. There is little doubt that, when molecular manufacturing scales up to kiloton-scale, it will be utterly revolutionary. But why do we expect the impact to be so sudden?

Look again at the doubling numbers. To scale from one nanoscale fabrication system to a kilogram of them requires about sixty doublings. To scale from kilogram to kiloton capacity requires only twenty. This implies that, not long after we get the first fabrication system working, we will be able to build truly awesome computers, among other things. (The NEC Earth Simulator could fit inside a grain of rice, using two watts of power.) But even more importantly, in one-third of the attogram-to-kilogram time, we will be able to scale from one nanofactory to a million. In fact, it'll be faster than that, because the scaleup will require debugging time, while mere duplication can happen as fast as the machines can work.

Nanofactories have additional practical advantages, such as being able to build prototypes of a new product just as easily and inexpensively as they can build production runs. This should speed research and development substantially -- by orders of magnitude, in some cases. An encyclopedic list of these advantages would make this post too long.

In summary, the reasoning goes like this:

1. Nanotech is working toward productive nanosystems.
   
2. Nanofactories must be scaled up from nanoscale systems.
   
3. Before this scaleup, nanotech products will be very limited.
   
4. The scaleup can happen quickly.
   
5. After the scaleup, lots of revolutionary products can be designed quite rapidly and built almost instantly.

Thus, although technical progress up to and including nanofactories may follow a relatively smooth slope, the impacts of that crucial last step will be revolutionary and disruptive.

Tags: nanotechnology nanotech nano science technology ethics weblog blog

Comments

[Mike Deering]

Molecular manufacturing, productive nanosystems, nanomanufacturing, nanoscale fabrication systems: all of these terms focus on using molecular scale automated machines to make products. Making products is one use for molecular scale automated machines. Why have you singled out this one use for molecular scale automated machines while ignoring others? Here is a partial list of other things molecular scale automated machines could be used for:

* repairing or changing the body at the molecular scale.
* collecting and storing energy.
* reforming on command, your whole house and everything in it could be made of intelligent utility fog, maybe even your aircar too.
* breaking down unwanted objects, recycling, chemical and physical processing.
* cleaning, pest control, germ fighting, environmental repair.
* processing incredibly huge amounts of information. Everything, regardless of size, can be intelligent.
* diamondoid life-forms.

Don't these other uses of molecular scale automated machines besides just making products need to be planned for as much as molecular manufacturing? When you get fully mature nanotechnology you get a whole set of new capabilities that you never had before, one of them is exponential manufacturing, but there are others.

[Mike Treder, CRN]

Mike, every one of those "other uses" that you describe relies on products built by nanofactories. It is nano-built products that will enter the human body, that will collect and store energy, and so on.

The radical impacts of molecular manufacturing that CRN warns about result from the drastically advanced products that MM will enable. As you suggest, it is not only the nanofactories that will be so disruptive, it also will be the previously unthinkable tasks that their products might perform.

[Michael Anissimov]

Sophisticated nanotechnology applications like nanomedicine and utility fog won't likely arrive with the first wave of products, either. There is incredibly complex programming to be done, to ensure that the nanobots don't consume you from the inside or the utility fog doesn't suddenly collapse in skyscraper mode.

[Mike Deering]

The first generation of MM products can be designed with current types of design software. These first products chief advantage will be their extreme precision though in functional complexity will be similar to present day products.

The first generation of AGI will be implemented on current or near current computer hardware.

Later generations of MM products will increase in complexity requiring super-human levels of intelligence for generation of their design and also incorporated into the products themselves.

Later generations of super-human AGI's will require computer hardware only producible by MM.

In this way the advancement of MM and SAGI are linked. I believe this linkage to be synergistic to the point of producing a faster rate of development than either would indicate if considered in isolation.