Rebel Genius: Warren S. McCulloch’s Transdisciplinary Life in Science by Tara Abraham
Summary: In “A Logical Calculus of the Ideas Immanent in Nervous Activity” Warren McCulloch and Walter Pitts presented one of the first mathematical/logical models of a neuron. The model is at once naive and also incredibly insightful: it certainly is not fully realistic, and yet it is an attempt to reduce the complexity of the brain and intelligence down to a simple model amenable to logic and mathematics. In many ways the model is a founding paper for the connectionist approach to understanding the mind, though one can see through its connections to boolean algebra the thread of computationalist ideas as well. This book is a very academic biography of one of the authors of this important paper, Warren McCulloch.
Rating: This is the sort of book you get assigned to read in a history of sciences course. The strength of the book is in its examination of the challenge of transdisciplinary work, here defined not just as a mingling of disciplines, but as one discipline being used as a stronger tool in another (here logic and math being applied across the divide into neuroscience). I found the sections discussing how McCulloch’s work was perceived of across the disciplines interesting. Unfortunately the book is very light on a detailing of the actual contributions of McCulloch. I left the book having some idea of who McCulloch was, and the events that transpired to put him where he sits in the pantheon of early neuroscientists, but didn’t come away with a deep understanding of the details of his work, or how it compared and contrasted with that of other early AI pioneers (like Hebbs and Weiner).
Speculation: Mappings from one field into another often bring about great change in the target field. Consider these mappings as reductions in the computational complexity sense. In computational complexity reductions lead to complexity classes and the realization that for some of these classes there are complete problems: every problem in the class can be reduced to these complete problems. One wonders if there are similar notions across the disciplines. And what the complete problems we should seek out when at first delving into a new field?