We describe a number of simulations using neural networks and genetic algorithms that illustrate how (artificial) organisms are not passive receivers of inputs from the environment but they actively influence the environment in which they live. By using Artificial Life Neural Networks, i.e., neural networks that have a body, live a physical environment, and are members of evolving populations, we show that (a) organisms evolve or develop their own way of encoding the local environment in their sensory organs, and that they evolve or develop behaviors with the only function to be exposed to useful sensory input, i.e., input which is more informative than other input or is necessary for learning, by either (b) changing the physical relation of their body or body part to the external environment, or (c) directly modifying the external environment.
展开▼