After Digital

James A. Anderson

Published by
Oxford University Press
ISBN 9780199357789
RRP £29.99
Reviewed by Jude Umeh FBCS, CITP

9 out of 10

After Digital is primarily an academic book featuring very dense coverage and focus on the topic of brain like computing. Not Artificial Intelligence, Machine Learning (or even robotic self-driving cars and IoT) per se, but brain-like computing, pure and simple. It is less about digital transformation but more about cognitive science and neurology. Once you can get past what this book is not, you’ll find it provides some authoritative perspective on a much-reported topic, backed up with academic rigour as well as a refreshingly scathing critique of the over-hyped AI rollercoaster. The excellent counter arguments at the end of Chapter 16 sums it up quite nicely.

In spite of the academic tone of the work, you can tell that this book was written by a real practitioner and is not just theoretical mumbo jumbo. The author, James Anderson, has been there and done that. A tenured university professor with a pedigree of AI related ventures (such as, and various publications, including this book, to his name. He is very much invested in his work and doesn’t hide his hardware specialism, focus and bias, which is evident in the choices and approach taken for this book.

The book is a somewhat lengthy but well organised work, designed to drive home the author's arguments, several hypotheses, and conclusions. The early chapters are clearly devoted to distinguishing key facets of the subject, for example, hardware versus software computing; analog versus digital approaches to computing; and AI versus neurology and cognitive science. The second set of chapters deal with how the brain works by delving into the logic and learning by association; a study of cerebral cortex structures and neurological experiments over the years (especially studies on animal vision and evolution); as well as a look at brain theory, historical constraints and the role of numbers. Altogether very dense material.

The third set of chapters describe the progression of cognitive science, brain-like computing hypotheses, and the author’s hardware-based approach, as well as detailed comparison to biological intelligence. The final two chapters provide a near term and longer term perspective on the future of brain-like computing and its potential applications. Very insightful.

I particularly enjoyed the aforementioned critique of the AI hype machine, notwithstanding the relative and amazing successes of breakthrough AI projects such as: AlphaGo, Big Blue, or even Watson. The author's Ersatz Brain approach to computer cognition is based on certain hardware, architecture and scaling conjectures which point toward a recursive, almost fractal-like, cognitive capability, (built on a multi-layered scalable pattern) that embodies and delivers increasing complexity and sophistication as you scale it up.

The only downside is that this book will be a tough read for anyone without basic exposure or interest in some of the key topics it discusses, such as computer science, neurology, cognitive science, evolutionary biology and even optics. That said, it is an excellent book with diligent focus on the currently under-hyped topic of brain-like computing, which in my opinion, will only become more important as we approach more useful general purpose AI capabilities and systems.

Further information: Oxford University Press

August 2018