Robot factory predictions
Mike Treder
2006-05-19 00:00:00
URL

Chapter 5 explains the technical advances that both require and allow (according to the author) the new economic structure he proposes. It seems that fully automated factories were about to be developed, using computer-controlled robots to do manufacturing operations.

When automatic factories be­gin to manufacture automatic factories, cost reductions will propagate exponentially from generation to generation. The introduction of computers into the manufacturing process thus has the potential for increasing productivity on a scale never before conceivable. Eventually the cost of finished manufactured goods may fall to only slightly above the cost of unprocessed raw materials.

Albus projected some of the possible consequences. His writing on this point sounds a lot like CRN's:
Robot technology, like computer technology, has military as well as economic implications. Any country that develops the capacity to run its factories around the clock seven days per week with only a few human workers will have a tremen­dous advantage both economically and militarily. If nothing else, this capability would allow military weapons to be pro­duced in virtually unlimited quantities at extremely low costs. But, even assuming that such plants were never used for mili­tary production, the country that possessed such a large sur­plus of efficient production facilities could easily dominate the world economically simply by selling manufacturing capacity at rates far below what countries using less efficient methods could hope to match. .... Whether this event results in unprecedented benefits or economic chaos depends largely on whether we can devise satisfactory answers to the questions: “Who owns these machines? Who controls them, and who gets the wealth they create?”
So what happened? Why didn't automated factories change the world? Why aren't factories fully automated, even now? And why do we expect that general-purpose nanofactories will be easier to develop?

Several factors make nanofactories different from large-scale robot factories. But first, note that automation in factories has in fact brought down the cost of goods quite substantially. If not for the fact that advances in transportation have allowed robots to be outcompeted by inexpensive labor overseas, we would probably be seeing even more robot use.

A large factory is limited in its speed; large machines can only work so fast. If the first automated factory-building factory costs a billion dollars, and it can build a new factory in as little as a year, then it will take ten years before 1,000 factories exist, and each factory will still have a million dollars in amortized capital cost--plus the substantial cost of raw materials. Meanwhile, advances in manufacturing technology will require continual redesign to avoid obsolescence.

Today's machines require a wide array of materials, formed by an even broader array of processes. I'd guess that at least a million different operations, from chemical purification to injection molding to grinding to measuring, are involved in making a large modern factory. Each of those operations would have to be researched and developed in order to automate it. It's no surprise that progress has been incremental.

By contrast, a molecular manufacturing system will use extremely small machines that can work very fast. Basic scaling laws, as well as comparisons with biology and preliminary engineering studies, indicate that a nanofactory should be able to manufacture its own mass in something like an hour. Within a month, the cost of developing a nanofactory could be amortized over not thousands, but billions of factories; furthermore, nanofactory manufacturing capacity would not be scarce. That changes the economics of production more fundamentally than a mere order of magnitude decrease in cost over several years.

The high performance of nanoscale machines (again, due to scaling laws) implies that there will be more demand for nanofactory-built products. It also implies that machines can be over-engineered, making designs less dependent on exact material choices and reducing the number of different materials needed.

Reduction in materials needed implies a reduction in the number of processes needed to make those materials. From the other direction, the discrete and uniform nature of atoms makes control much easier; two parts built by identical operation sequences will be perfectly identical (except for transient variations from thermal noise, and very rare manufacturing errors that will be detectable with limited sensing). The discrete nature of atoms and their bonds also means that properly-designed parts will not wear or require lubrication. All these factors simplify the design task.

It has been speculated that a minimal engineered bacterium might require less than 200 genes. This is for a system that not only manufactures, but metabolizes and self-repairs. It is likely impossible to build a macro-scale hands-free manufacturing system with 200, or even 2000 parts. But there are fewer than 200 atoms in the Periodic Table, and a molecular manufacturing system will be able to take advantage of that fact.

Design of a nanofactory would be extremely difficult today; it might cost on the order of a billion dollars. However, improvements in computers will make it far easier, quite rapidly. I can simulate reactions and components on my laptop today that would have required a small supercomputer a decade ago. Advances in general knowledge of chemistry will continue to improve the models. Rapidly improving tools, driven by other industries, will also make the job easier.

Today, it would be possible to build a fully robotic factory, but it would not be economically rewarding. In just a few years, it will be possible to build a nanofactory--and it will be very rewarding. Someone will do it.