Book Review: ‘How to Create a Mind — The Secret of Human Thought Revealed’ by Ray Kurzweil
First of all, let’s get some things out of the way: Ray Kurzweil is one of the most brilliant minds of our time. His numerous inventions and his insightful predictions about the future of technology are remarkable. So when he chooses to say something, we should choose to listen… or in this case — read.
The Structure Of The Book
How to Create a Mind consists of 11 chapters each looking at different aspects of the endeavor of creating a digital mind. Chapters 1 and 2 are an introductory reflection on the nature of thinking. They are designed to prepare us for a methodical presentation of the way the human neocortex functions. Chapter 3 presents a theoretical model of the neocortex as a system of pattern recognizers. It is in this chapter that Kurzweil lays his theory about creating a digital neocortex. Chapters 4 through to 6 present us with the way the biological brain functions as well as elements of its evolution. Chapters 7 and 8 explain how humanity could build a digital representation of the neocortex (and therefore — intelligent machines). Chapter 9 presents us with the philosophical conundrums of ‘consciousness’ and ‘free will.’ Chapter 10 explains one of the main themes in Kurzweil’s line of thought — the law of accelerating returns and chapter 11 presents answers to criticisms of his earlier works. With the epilogue, Ray Kurzweil takes us on an ever-accelerating journey that represents the fate of humanity according to the law of accelerating returns.
The Main Theme
The book streams with optimism and at the time of writing of this review (April, 2017) it was a much needed respite from the wave of doom and gloom sweeping across the world.
In his description, Kurzweil methodically explores every aspect of the biological and the digital brains. He references numerous results of other researches in support of his hypothesis: that a digital brain with human-like intelligence can be built and that such a brain would be as conscious and as exercising its free will as humans are (Kurzweil also hints that we are going to be as attached to synthetic beings just as we can be attached to organic ones).
The main theme of the book is developed in chapter 3, where Ray Kurzweil presents level-headed, methodical and scientific deconstruction of the elements of human intelligence. The author’s thought cuts through the wonderful, imperfect, fuzzy human mess that is the brain and finds a simple, fundamental structural element that is propagated 300 million times to form our neo-cortex. That structural element is the basic pattern recognizer (a structure consisting of about 100 neurons). He explores the way that the mind emerges as a result of the connectedness of these 300 million pattern recognizers. The author traces the multiplication of complexity of human thought as patterns of patterns form from the most elementary, environmental stimuli to create the emanation of human thought — the arts and sciences.
Secondary Ideas
Besides the main theme of the book (the building of a synthetic mind), the author weaves in several other secondary ideas. One of them is that the earliest types of neural networks were limited in their capability (specifically Rosenblatt’s perceptrons). The reason for this was the perceptron’s inability to solve the invariance problem (distinguishing whether two images were connected or not). The author quotes the book ‘Perceptrons’ by Marvin Minsky and explains that due to essentially an overreaction, much of the funding for the field of ‘connectionism’ has been withdrawn. This explains the long pause of developments in machine learning between the 70s and its resurgence now (when advancements in technology allow for much shorter iteration cycles). Although performing well at the autoassociation task (recognizing a pattern when a part of it is missing), the author focuses on the shortcomings of neural nets when it comes to the invariance problem. Kurzweil uses this shortcoming to justify his use of hidden hierarchical Markov models (HHMMs) — a technology he has used in his previous work on speech recognition. Ray Kurzweil is a firm believer in the capabilities of the HHMMs and he emphasizes their advantages throughout the book.
The way the ‘old’ brain (as referred to in the book) coordinates with the neocortex is another secondary theme developed in the book. The interesting, everyday take away from it is that our experiences are shaped by virtue of the information trajectories in our brains. When an emotional stimulus arises in our old brain, it passes through the neocortex where it sublimates and gets a confabulation (an explanation of the desires of the old brain in terms of the experiences of the new brain).
Example (familiar for many, I believe) is if your boss screams at you. This information gets translated in your neocortex as a possible worsening of your work situation and therefore a possible dismissal. In turn this is linked to a possible financial hardship, therefore a possible decline in standard of living. This gets passed on to the old brain as a ‘threat to survival.’ That’s why you may react physiologically to such an event the same way you would react to a real life-threatening situation (e.g. a lion chasing you). This explains why people are sometimes overly-agitated by certain topics or why everyone tends to feel a struggle in their heads between strong emotional reactions and a cool rational response (i.e. the impulses of the old brain and the rational thought of the neocortex).
In conclusion, as controversial as this book could be to some, I consider it a book with a very high return on invested time reading it. What is even more amazing is how accessible Ray Kurzweil has presented the complex concepts the book covers. Needless to say, I’m keenly looking forward to his next book.
