Cybernetic roots and futures of neural computing

Superficially, neural computing is intricately concerned with control and information in man and machine. However, looked at in greater depth, it seems important to keep brain studies at arms length from the making of neural computers. In fact, one could argue, neural computing could lead to a situation which is almost contrary to the aspirations of early cyberneticists. There is a danger in bringing the study of brains and machines under the same theoretical roof from the point of view that one underscores the deficiencies of brains and starts building them into machinery. In this context the author raises and answers several questions: Does close modelling of the neuron lead to the best design of machines. Does the analogue nature of neurons have to be achieved in silicon? and neurons have very large fan-in and implementations of the McCulloch and Pitts model would become very expensive if this were mimicked. Does this mean that neural computing can never aspire to human competence? The author aims to show that while constructive inputs from brain studies will always help in the effort of designing neural computers, the neural systems designer is advised to concentrate on inventing analytic tools that help him to understand the artificial world without using applicability to brain studies as a yardstick