232 Wybo Houkes
solutions to the scalability problem: even for relatively simple problems, such as construct-
ing patterns of tessellating tiles, the number of rules needed increases dramatically, with
no improvement in the fi tness of the solutions, with increasing complexity (Bentley and
Kumar 1999). Consequently some researchers have introduced a second level of nature
imitation. They do not just employ a distinction between circuit phenotypes and an underly-
ing code but they also seek to imitate the embryogenesis of organisms, that is, the way in
which items develop through the interaction of the genetic code with a constantly changing
environment. This approach does not make use of a large and intricate set of rules—not
even of a set that may evolve during the design process. Instead a random set of starting
confi gurations are used in combination with simple rules, which are activated or ignored
depending on the state of the environment (i.e., the intermediate product of the design
process), applied in parallel instead of sequentially, and which—perhaps most impor-
tant—can be changed or supplemented during the evolutionary process; extra rules may
be added without the intervention of the designer, or the activation conditions of existing
rules may be changed. In this so-called implicit approach, more of the design process is
put beyond the control of the human engineer. Some promising, albeit very preliminary,
results have been reported: an implicit embryogenesis appears to solve complex problems
more quickly, more reliably, and more diversely than explicit approaches (Bentley and
Kumar 1999; Kumar and Bentley 2003a; Gordon and Bentley 2002, 2005).
13.2.3 Preliminary Assessment
What is, for my purposes, most salient about this research is that some researchers in ED
are trying to overcome a specifi c problem, that of scalability, by means of an increasingly
intricate transfer of selectionist concepts and models to the domain of artifacts. Thus
electrical engineers do not transfer concepts and model mechanisms because there is no
argument not to do so, or because they envisage a unifi cation of biology and engineering
science. Instead they have noted that there is at least a structural similarity between a
problem in the domain of artifacts and one in the domain of organisms. This scalability
problem is the driving force behind the interdomain transfer of concepts and the modeling
of mechanisms.
The researchers themselves most often describe their efforts as “biologically inspired”
or based on “metaphors.”^6 If this were true, organisms would hold no privileged position
over other objects that may be a source of inspiration—circuit designers might just as well
have looked at the clouds. Yet the choice for this particular source is far from arbitrary.
The reason for looking at organisms rather than clouds is, as said, the similarity of the
scalability problems encountered by nature and electrical engineers. Moreover, the interest
of biologists in ED programs is almost comparable with the interest of engineers in genetic
mechanisms.^7 This is explained straightforwardly by assuming that the prospects of ED
programs increase the more accurately nature’s strategies for solving the scalability