Logical neural nets for easy VLSI

Summary form only given. Neural nets have a rather curious place in the field of symbolic processing. In fact, one role which they play well is to turn ill-defined, large-scale data (as found at the output of vision and speech transducers) into well defined symbols. Another role is that of associative retrieval: the discovery of missing symbols in partial `records´. It is becoming clear that much of the potential of this approach can only be exploited through the hardware realisation of the nets themselves. The author briefly reviews the various approaches: analogue; digital; optical; logical. A case is made for the last of these