The products of natural evolution can directly or indirectly help for the first steps to advanced APM systems thus they are of interest here.
Limits of natural evolution
Evolution can only reach the places in design space that are reachable by a sequence of small steps. E.g. bicycles, cars, big sized wheels in general, spaceships, wireless communication via radio waves and microwaves, metal conductors, superconductors, silicon microchips, and many more things cannot be reached by evolution. It often has been argued that advanced diamondoid APM systems are unfeasible because nature would have built them - one of the common misconceptions about atomically precise manufacturing.
Naturally evolved polypeptides and their systems (the "machine" parts of living cells) have a number of limitations that make it desirable to move away from them as fast as possible. Polypeptides are:
- not as simple as possible
- not very stable (actually there are evolution related reasons for them to be on the threshold of stability) [todo: which ones]
- not minimalistic in function - they always have to carry shape recognition structure around with them
- ... and some more ...
Systems of polypeptides in nano-biology are often not decomposable that is they are not modular. Changing one thing changes almost everything. This can make the engineering practice of narrowing down errors or finding logical dependencies a nightmare.
Evolution of technology
Is there something like evolution in technological progress? Let's compare it with traits of natural evolution:
mixing and merging that preserves function
On a short timescales technology progresses mainly through targeted design. New measurement results and accumulated documentation is used to choose the next step. No mixing here.
On a long timescale technological progress is mainly lots of unexpected mergement of seemingly unrelated information. Such information comes randomly from the accumulated documentations or measurements on new systems. This is something called serendipity. Hidden options just waited to pop up. Biological evolution too has hidden option that crop up only if a certain level of sophisticatedness is reached but there are key differences:
- documentation and natural manufacturing systems coincide (there's no manual needed for a machine operating itself)
- the documentations allows us to make smarter and more straight forward use of new options (this lets us overcome the small step limitation mentioned earlier)
undirected brute force trial and error
Except in artificial evolution via [genetic algorithms] the natural (radiation driven) method of brute force trial and error that visits much more dead ends than successful continuation points is not something that is usually done to advance technology. Whether artificial evolution should be counted to deliberate design or evolution seems unclear.
Evolution a is rather bad term for technological progress. Instead: targeted technological design and technological serendipity are more fitting technological progress does not solely equate to anyone of these alone.
Evolution of APM systems
APM systems do not evolve due to their system architecture. It is similar to a microprocessor which with human help can calculate which adaptions to itself are necessary to perform better It is very dissimilar to a microbe which adapts to the environment by itself.
It makes sense to put only one copy of the blueprint for a whole nanofactory into one macroscopic device and not put a copy in every potentially autogenous subunit of it. This still makes the technology base quite disaster proof while avoiding useless amounts of redundant data.
What could happen when radiation hits: (list with grossly decreasing likeliness)
- Some atom of a structural DME gets displaced - its likely that nothing happens
- Some internal atom of a moving DMME becomes displaced - a little more friction - greater damage is unlikely
- Some atom of a sliding or rolling interface surface becomes displaced - more friction - possibly spreading damage (stuck, worst case: burn down)
- Some atoms in a radiation sensitive data-storage become displaced (or electrical bits become flipped):
- > Low level data: same effects as above but copied to every standard part of the same type.
- > High level data: It can cause what in biology is known as fasciation (e.g. repeat this structure 0100b = 4 times becomes repeat this * structure 1100b = 12 times) or similar stuff complicated by mixed in decompression artifacts. Geometrically defined shapes could change too (constructive solid geometry).
Such and many more interesting errors normally won't make the product better but rather dysfunctional. E.g. it's obvious that a bolt won't fit if you change its diameter.
With effort one can augment highly parallel component testing with emulated evolution. To do so one can mix alternative combinations of compatible design choices according to a genetic algorithm (possibly at different levels of abstraction) synthesize a block with every tested unit different (very very many) and see what fairs best.
In most cases testing for improvement in a part that is in regular operation makes no sense - this is better done off site.
- Evolutionary Capacity: Why organisms cannot be like machines (Eric Drexler's blog Metamodern (archive) 2009-08-02)
Biological and Nanomechanical Systems: Contrasts in Evolutionary Capacity (at K. Eric Drexlers website)
(pdf direct link) (wiki-TODO: include abbreviated discussion of that in this article)
- Wikipedia: Evolution
- Maybe related paper: "Causal Entropic Forces" by A. D. Wissner-Gross and C. E. Freer