I guess I have to write something really smart now, on a book with such a title… This book comes highly recommended, by people like James Watson of DNA fame, to name just one. The author is not the typical scientist though and that makes the book rather intriguing. Jeff Hawkins is the man behind the PalmPilot or the Treo smartphone, quite impressive achievements in itself. But Hawkins makes it clear that understanding the brain has always been on his agenda, or should I say on his mind. As I am the lucky winner, and subsequent happy user, of a Palm m515, I was curious and could not resist buying this book. It is a popular science book and very very accessible. A bit too accessible at times and the book is rather slow at embarking with the real stuff. I wonder if the book would not have benefited more from the collaboration with a Scientific-American style illustrator than with a co-author.
So what kind of intelligence are we talking about? From the computer background of the author an obvious first choice might be the “artificial intelligence” kind of intelligence (GOFAI). But the author is clearly interested in the biology and/or anatomy of our brains and states that this is too often ignored. In other words, more conventional computer power will not do the trick. But he also states that simply adding more “neural networks” the connectionist approach, so successful in pattern recognition, will not yield intelligence. The problem Hawkins wants to tackle is the big one, explaining how the brain works, emotion, creativity, consciousness and everything else included. And knowing that “We will be able to build genuinely intelligent machines”. The thesis of Hawkins, condensed in one phrase, could sound like this: intelligence emerges from the neocortex, a memory-based continuous prediction machine.
Hawkins describes how in 1979 he was inspired by a Scientific American paper by Francis Crick (also from DNA fame), an influential paper that made it obvious that understanding the brain and consciousness was a wide open field. Another inspiration for Hawkins, from the same period, was the “Chinese room“, the famous thought experiment by John Searle. Hawkins has a rather peculiar reading of this though experiment. He agrees with Searle that the Chinese room does not have intelligence, apparently refuting any computational theory of mind, but the rest of the book is the description a one particular computational theory of intelligence, a fully materialistic one, dominated by our insights of the structure and function of the neocortex. These philosophical problems can be safely ignored however, they are only what got Hawkins started, and after some interesting history of neuroscience the major inspiration of Hawkins Vernon Mountcastle makes an appearance.
Next we get to the main course of the book. I will not try to summarize it comprehensively. The thesis of Hawkins, condensed in one phrase, could sound like this: intelligence emerges from the neocortex, a memory-based continuous prediction machine. Hawkins quite convincingly describes how the anatomy of the different layers and column in the cortex together with some clever feedback mechanisms and connections with other parts of the brain explain a lot about our intelligence. This is the most technical part of the book and it would benefit greatly from enhanced graphics. Some fancy, interactive, on-line technology would come in handy, I guess there should be enough room on the books web-site .
I like the core of Hawkins thesis, to use a clichÃ©, it provided me with quite an “aha” feeling. Even better Hawkins framework may explain what happens during an “aha” experience. In the appendix Hawkins makes predictions about this (nr8) and gives a lead about how this has to be tested experimentally. This is the strongest aspect of the book! This is real science in action.
In the penultimate chapter Hawkins wraps it up by tackling a lot of subjects and questions and linking them together with his central theme the ‘memory-prediction framework’. This sometimes works as in his discussion of awareness (the easy to explain part of consciousness). But other topics are treated in a rather superficial or strained way as when he claims that language is a fully learned capacity, Pinker has convinced me it isn’t. Hawkins take on creativity didn’t convince me either. He basically describes how we might detect art but there is no clue about how it is created, or how machines would become truly creative. We should however not request an explanation of everything at once, and Hawkins is probably over-stretching things in this chapter. (maybe, just maybe, having a business, product-launch, background is a handicap here)
In the last chapter Hawkins gets into Kurtzweil-light mode and describes/predicts the intelligent machine of the future. Hawkins has an optimistic view, end his quest does not necessarily result in scary machines or Bill Joy type gloom and doom. Hawkins envisions that emotions and experiences are optional as this would make machines uncomfortably human-like. This kind of engineering sound to me as a connectionist approach next generation. Exciting? Certainly. Intelligent? Maybe. I have no doubt these boxes will play chess even better and faster. I fail to discern the quantum leap toward intelligence. Maybe Hawkins wants the cake and eat it. Following Damasio and others, none of which are mentioned or referenced in the book, intelligence without the emotions is probably noting more than automated chess playing. But is this the kind of intelligence that got Hawkins excited? In the end I will have to read Vernon Mountcastles book to fully position Hawkins and make up my mind about this intriguing book. But I recommend it and hope more outsiders write up a synthesis of their bold ideas.