This is a difficult book to review because while it is professionally written on an interesting topic, it must also be judged by whether it achieved its objectives, where it was not so successful.

This is a scientific book, and as such a certain amount of the success will be in targeting a readership at the right level of knowledge and communicating its ideas to them. And since I am the one giving an opinion, you need to know my level of knowledge. I chose the book because I thought it might be aimed at me: a moderately intelligent guy with several university degrees and a basic knowledge of calculus and statistics (from 50 years ago). I am also writing a Science Fiction novel on Artificial Intelligence, and I have found it very useful as research.

Unfortunately, I suspect this book is trying to do two things at once for two separate sets of readers. The target readership as stated is “everyone interested in recent developments in AI.”

This includes two groups — educated scientists and the non-scientific socially conscious —  that are too far apart in training and interest for the author to bridge the gap, no matter how well he writes.

This work is in standard textbook format. “Tell them what you’re gonna tell them. Then tell them. Then tell them what you told them.” This keeps both the author and the reader focused on what’s being taught. Like a good lecture, there is also a smattering of gentle humour that keeps readers entertained. There are clear diagrams and useful practical examples to maximize our understanding.

The first two chapters lay out the history and the basic mathematical logic behind the programming of Artificial Intelligence. It is clear and understandable, and does exactly what the author has led us to expect. Because the logic is so well explained, the reader can sort of follow the mathematics, although it becomes increasingly difficult. By the time I had finished that section, I felt I had learned a great deal about AI, how it functions, and how we create increasingly complex versions of it. I had a hazier view of how and why the programming part works.

Chapter 3 is about statistics. Once again, we start with the basics, carefully laid out. If you understand simple statistics, you will find this material useful and fascinating. However, when the topic turns to loss function, the complicated formulas start appearing, and I find myself losing track. Then it moves into calculus. I last studied that art in 1967. Enough said.

Chapter 4 goes back to logic and discussing the practical uses of AI, and this is the real meat of the book. Its problems when used in the judicial system are shown clearly and have a great deal of social import. Its use in advertising involves concepts that have been discussed for many years, so this section of the book really brings us back to solid ground.

The conclusion wraps it up neatly, but the author says he wants readers “…to understand how neural networks function so that you can have an intelligent conversation and formulate your own opinion about the ethics of AI.” By including a great deal of mathematics I don’t understand, he leaves me with the impression I can’t have that conversation or formulate an opinion. The chapter of equations has effectively created a wall preventing this from happening. For many readers, it will be the place they stop reading the book.

One opinion that did come through loud and clear; despite what Science Fiction writers and social media hystericals maybe telling tell us, “…there is still a large gap between human-level intelligence and our best efforts at artificial intelligence.” I’m willing to accept the fact that I can’t understand the math as a pretty good indication that this is so.

So, despite the fascinating information I learned from my reading, I think the author tried to reach too broad an audience, so is bound to disappoint some. This book is highly recommended for those with a scientific education. Many other readers might want to browse through and skip the tough math part, knowing that the ending returns to important material that is easier to grasp.

(4 / 5)