|Image by Creative Tools licensed under Creative Commons Attribution 2.0 Generic (CC BY 2.0).|
In a recent piece in Locus Magazine, Cory Doctorow discusses the economic and copyright consequences of a manufacturing model in which 3D printing enables production runs of zero, tens, thousands or millions of objects. The article mentions that the limited production capacity of 3D printing companies makes large production runs troublesome, but there is another problem with the economics of this kind of mass production: 3D printing is still quite expensive, and there are not significant economies of scale. The main cost driver for 3D printing is probably machine time, and printing a thousand model rabbits requires a thousand times as much machine time as printing one.
In large scale search – and many other internet services and computational domains – have a similar problem: computing a result can be quite expensive. Perhaps 3D printing services should adopt the same solution used in search (computation): caching. In caching, the results for commonly submitted queries (computations) are stored in a cache. Future results for a cached query can then be generated cheaply by returning copies of the cached result rather than rerunning the expensive computation needed to generate a result from scratch. Without such caching, search and other large scale internet services would be very significantly (3x-10x?) more expensive.
Maybe companies like Shapeways which operate a 3D printing service could use caching to significantly reduce costs? When a sufficient number of orders for an object have been made – or are anticipated – they print a mold, and produce future copies from the mold(s) rather than on the 3D printers. Just as in computational world, such caching could be performed entirely behind the scenes, and just as in the computational world, it could significantly reduce the cost producing commonly requested objects.
Casting is already a limited part of the 3D production process – on Shapeways, silver objects are exclusively produced using lost-wax casting.
As a footnote, perhaps what draws me to both 3D printing and computer science is the shared computational aspect. One of my current 3D projects concerns an object which is entirely procedurally (computationally) generated. So far that has required about two weeks of Erlang programming for a custom extension to Wings3d.