|
|
Chris Thornton
As natural resources become less abundant, we naturally become more interested in, and more adept at utilisation of waste materials. In doing this we are bringing to bear a ploy which is of key importance in learning - or so I argue in this paper. In the `Truth from Trash' model, learning is viewed as a process which uses environmental feedback to assemble fortuitous sensory predispositions (sensory `trash') into useful, information vehicles, i.e., `truthful' indicators of salient phenomena. The main aim will be to show how a computer implementation of the model has been used to enhance (through learning) the strategic abilities of a simulated, football playing mobot.