Since the very first days of artificial intelligence (AI) in the forties and fifties of our century, there have been two main approaches on how to model and simulate intelligent behaviour. While the symbolic approach, as favoured e.g. by Alan Turing, saw intelligence a process relating discrete concepts and predicates according to certain rules, the connectivistic approach, as supported by e.g. Warren McCulloch and Walter Pitts claimed that the connection and interaction of many small and simple units could show intelligent behaviour; a theory that is strongly encouraged by the fact, that this concept has already shown itself extraordinary successful; in the neural system of the human brain.
The evolutionary principle of mutation and surviving of the fittest, first formulated by Charles Darwin, has also proven to be obviously a rather successful one. Despite of the discovery of the genetic encoding in DNA-strings and the cellular reproduction mechanism, it took rather long, until scientists like e.g. John Holland and Davis Goldberg took up the idea to use the same principle as an optimisation algorithm in computers.
While both methods didn't in fact come up to the high expectations that their biological counterparts might suggest, both have left behind their image of rather academic research and play tools and are nowadays generally accepted and used in a wide range of applications, where traditional methods often prove to be unsatisfactory.