Article Preview
TopWhy Nature?
Recently, an eminent colleague, on hearing that one of us worked in evolutionary computing, responded jokingly ‘Have you heard of sand computing?’, making the argument that a heap of sand, through its chaotic dynamic behaviour solves a problem far beyond the scope of computing systems – but implicitly also making the point that this complex behaviour might not be very useful. Why should we expect that nature can inspire useful systems?
In the case of biologically-inspired systems, there is an obvious argument. Terrestrial biological systems have been evolving for at least 3.5 billion years (Schopf et al., 2002), and potentially far longer under the panspermia hypothesis (Thomson, 1871; Crick & Orgel, 1973). In doing so, they have not only solved specific problems, but also evolved more general problem-solving capabilities. Thus the genetic code has itself been optimised as a representation language for specifying proteins (Freeland, Wu & Keulmann, 2003), the organisation of genes into chromosomes is subject to on-going evolution to permit faster and more effective response to selective pressures (Batada & Hurst, 2007) and even sexual reproduction itself probably arose as an evolutionary response to co-evolutionary pressures from parasites (Hamilton, Axelrod & Tanese, 1990).
So why is this different from ‘sand computing’? Why should we assume that the optimisation and learning problems that biological systems solve bear any relationship to those we wish to solve? Why should we assume that the techniques would be useful for our business, scientific and engineering problems?