Neural networks perspectives and potentials

The evolution of machine-based signal processing exploits the controllable complexity properties of the available physical vehicles. Contemporary integrated electronic device-array complexity appears to have reached the critical size where the functional behavior of some bioneural array networks can be emulated directly. Such nonnumeric analog token model methods were in common use before the digital computer era. The author suggests that, beyond a revitalization and broadening of analog model and circuit knowledge, the present neural network flurry may result in human-sense-like circuit functions of designable levels of sophistication where the analog feature complexity approaches that of the real thing. Candidates for eased-in growth are ASICs (application-specific integrated circuits) where one can combine and evolve the inherent strength of analog and digital design methods. It is concluded that technology deployment appears so far to be nonevolutionary and well within the existing and diverse analog circuit capabilities