John Von Neumann - Pioneer in the Mathematical Theory of Computing

Whether they think that artificial intelligence is impossible or inevitable, most people have highly polarized views on it. John von Neumann, genius, mathematician, and inventor of the nearly ubiquitous computer architecture that bears his name, blazed trails for both camps.
This short book, which was written originally for Yale's Silliman lectures, but published posthumously, summarizes his views on machine and biological intelligence with unprecedented clarity and precision. His understanding of neuroscience was that of a brilliant and strongly motivated amateur at the end of the 1950s--good enough to take on the problem, but by no means matching his comprehension of the machines to which he had devoted much of his professional life. Still, his take on intracranial computation is stunningly prescient--he looks beyond the then-fashionable digital metaphors to suggest a semi-analog strategy that uses parallel processing to make up for its deficiency in speed. Prominent neuroscientific thinkers Paul M. Churchland and Patricia S. Churchland provide a brief, enlightening foreword to this second edition, placing the author's thinking in context and grounding the reader in the scientific milieu that gave rise to The Computer and the Brain. Although his computer architecture slowly is growing obsolete, von Neumann has given us a more lasting legacy in his thinking about thinking. --Rob Lightner

I am not endorsing this speaker's final conclusion but his explanation is clear.  If you want to read more about self-replicating machines click here.