In my previous post, I listed several things that affect the amount of impact a self-copying manufacturing system will have:
- What types of products could it build?
- How fast does it work?
- How long does it take to build a duplicate system?
- How much does the feedstock cost?
- How easy is it to design new products?
- What inputs other than molecular feedstock are required to build a duplicate system?
Let's take a look at a system that exists today, and was designed to be self-copying: RepRap, the Replicating Rapid-Prototyper.
RepRap can build a wide range of plastic shapes. They can be water-tight, and fairly strong. The surface is not very smooth, compared to molded plastic, and they only come in one or two colors. [Update: surface quality has gotten much better in the last few months, and can now be compared with commercial machines that use the same melted-plastic method. See the second photograph on this RepRap blog entry.]
RepRap can deposit a few cc per hour, or a few tens of kg per year. Thus, you would not use it to build a boat, and it would take a while even to build a pair of shoes. [Update: Forrest Higgs tells me that RepRap, depending on the version, can actually print tens of cc per hour, again comparable to commercial machines.]
It can build the plastic components of a duplicate system; the motors, electronics, steel rods, and various other parts have to be purchased, and the system has to be assembled by hand. Although I couldn't find how long it takes to print the plastic parts, it seems to take several hours to assemble. Even if it took a week to print and assemble, it would be possible to make all the RepRap's you could ever want in well under a year - if you had enough people working on it.
The feedstock is just plastic filament, so doesn't cost very much. [Update: commercial machines charge a high premium for their plastic filament - RepRap may be an order of magnitude cheaper.]
New products can be designed in a variety of software packages.
As I see it, the impact of the RepRap is limited by three things: the limited range of products, the speed of manufacture, the labor and skill required to build and use it. All of these areas can improve, and future versions of RepRap may have greater impact. But it will be a while before RepRap's products compete directly with the products of other rapid prototyping systems, let alone factory-made products. [Update: while any plastic-deposition machine won't produce factory finish, RepRap appears to have advanced substantially in just a few months. If I were in the business of making rapid prototyping machines for profit, I'd be worried.]
RepRap's cost to own already is comparable with other home appliances, but its required skill is suitable only for mechanical hobbyists. This will surely improve. Its range of products and speed of manufacture will also improve. [Update: RepRap can be used for milling as well as deposition, and can thus be used to make electronic circuit boards.] When the convenience and/or fun of being able to make cool stuff outweighs the difficulty of learning to use it, and the cost falls by a factor of three or four, then RepRap will probably find a niche along with other high-end toys like Lego Mindstorms - unless it is outcompeted by low-end factory-made rapid prototypers.
For RepRap to make a major economic splash, it would have to be able to make a large fraction of the products we use today. Even high-end rapid prototyping systems today can't do that.
RepRap, for me, sets a lower bound on what a molecular manufacturing system would have to be able to do in order to really make an impact. Simply being able to make a large fraction of its own specialized parts is not enough; it also has to be easy to use, and make a variety of products (not just parts).
Depending on the underlying materials and fabrication methods, expanding the range of products of a manufacturing system will range from easy to near-impossible. It is conceivable that, for example, a DNA-based rapid prototyping / flexible manufacturing system might simply be unable to be extended to make a wide enough range of products. For this reason, even a system that clearly satisfied the definition of molecular manufacturing might not cause a fast takeoff.
But even a limited molecular manufacturing system might contribute to a fast takeoff. One of the limitations of nanotechnology research is the difficulty of doing anything at the nanoscale. A nanoscale manufacturing system, even a primitive one, could be expected to build tools that can't be built any other way. By increasing the tool set, it could speed development of other molecular manufacturing systems.
There is also, today, a major conceptual barrier: many people have trouble accepting that there's nothing magical about a mechanical system building a copy of itself. If this could be demonstrated in any nanotechnology - perhaps including such features as automatic maintenance of precision and full automation - then it would be easier to free up funding for more advanced molecular manufacturing programs.