This article about genetic algorithm (mostly used by biologists to simulate the evolutionary processes) focuses on the applications these techniques may have in design, and explains how it could help architects in their design processes.
Does genetic algorithm replaces design?
It has now become easy to breed new forms without designing them but deliberate design is still a crucial component. In order to be interesting enough, genetic algorithms must lead to unpredictable designs and bring a large variety of results.
In order to design rich enough search space, some philosophical ideas traced to the work of Gilles Deleuze may be useful: population, intensive and topological thinking. They were not invented by Deleuze himself but he was the first to bring them together, leading to a new conception of the genesis of form.
How to represent the final product and the process itself to be able to apply the genetic algorithm? In architecture, when using a CAD software, this is quite easy: the model represents the body and all the operations needed to design it correspond to its genetic structure, the “virtual DNA”.
Population thinking:
This form of thinking was created in the 1930’s, when biologists brought together Darwin’s theory of evolution and Mendel’s theory of inheritance to synthesize the modern version of the evolutionary theory. It leads to the idea that one needs to study a large reproduction community, instead of a small one, to observe the evolutionary process.
That means, for computer design, that the architect must have enough variables taking place in the generative algorithm, to be able to breed a sufficient amount of different results which could continue to propagate and interact with the rest of the community again and again.
Intensive thinking:
This idea is coming from the field of thermodynamics. Intensive quantities, such as temperature, pressure or speed are indivisible and a difference on intensity tends to cancel itself spontaneously.
In architectural design, intensive elements, such as the distribution of stresses, can be brought into a CAD model (via the genetic algorithms) in order to make sure that structural elements do not lose their function and that the result of this virtual evolution has a meaning for structural concerns. In this way, the software is doing natural selection, and the designer can perform artificial selection for aesthetic purpose in the remaining population forms.
To become a little more interesting, such processes requires the addition of mutation points in order to be intentionally transformed by the designer. But it’s not enough. In comparison with biological evolution, the results obtained by artists with those techniques are relatively poor, as if the space of possible designs was limited.
Topological thinking:
The notion of body plan, or “phylum chordata”, shared by a large variety of living bodies, may explain why biological evolution has generated that quantity of different shapes, contrasting with the results artists have so far obtained with the genetic algorithm. That means we could imagine a kind of “abstract vertebrae” that may be able to yield an elephant as well as a bird. Named as “abstract diagram” or “virtual multiplicity” by Deleuze, this concept may be applied to any entities (organic as well as non-organic).
To understand the idea of abstract diagram, we should relate this concept to geometry. Except from the familiar metric Euclidean geometry, in which extensive quantities such as length, areas and volumes are the basics notions; there is different kind of geometries in which these notions are not basic, since they do not preserve length because of inner geometrical operations such as: projective geometry, differential geometry, and topology which operations (stretching without tearing, folding without gluing,...) preserves only a set of very abstract properties invariant. Those invariants are precisely those we need to think about body plans and abstract diagrams.
“Architects wishing to use this new tool must not only become hackers [...] but must also be able to hack biology, thermodynamics, mathematics and other areas of science”.
Images for 18 Musicians, Live software performance for Steve Reich's Music for 18 Musicians.
Originally performed at the 2004 Ars Electronica Festival in Linz, Austria.
