Customer reviews

By late 2012, the short term predictions in this book have already happened. Might be worth paying attention to the rest.

0Comment

ByM.J.on September 9, 2012

This book was a great read, very accessible and might prove to be a very important book one day. It's concise and to the point and if you have any interest whatsoever in AI you simply can not miss this it. It's a quick read that will without a doubt have a significant impact on how you view the future of artificial intelligence.

As a testament to it's relevancy today (I'm writing this Sept 2012, seven years after the book was published) he predicts three technological applications that may become available in the short term (5-10 years) due to breakthroughs in the kind of trainable AI this book discusses:

Computer vision and teaching a computer to tell the difference between a cat and a dog (this was successfully demonstrated in a study published in June 2012 - the paper is called "Building High-level Features Using Large Scale Unsupervised Learning" and is available online, or just search for "computer learns to recognize cats" for articles)

PDAs (as they were called back then) will understand naturally spoken instructions like "Move my daughter's basketball game on Sunday to 10 in the morning" (this kind of sentence, copied from the book verbatim, is exactly where Apple's AI application SIRI shines)

Smart/autonomous cars - in Aug 2012, Google announced that their self driving cars have logged 300 K accident free miles in live traffic on public roads, exceeding the average distance a human drives without accident.

The thing to note here is that when he wrote the book these three things had hurdles that we did not know how to solve, and at the time there was no clear linear progression of existing solutions that would guarantee they would be solved. His prediction is that we'll be able to train computers to recognize patterns by themselves which will allow us to eventually solve the problems (and this is exactly how the computer learned to recognize cat faces from youtube videos)

Furthermore, he predicts that AI will become one of the hottest fields within the next 10 years - and with the current explosion of interest in Big Data, Machine Learning, and applications like SIRI it is hard to deny that it lookslike we're right in the midst of seeing just this happen.

The grander implications of the model of this book won't be known for another 10-20 years or more, but 7 years in his general predictions about the field of AI have been very accurate.

12 people found this helpful

Excellent book, but I have one objection

0Comment

ByOlivia Shafferon April 26, 2010

The book is about Hawkins' theory of how the mammalian cortex, especially the human cortex, works. Hawkins thinks it is only by understanding the cortex that we will be able to build truly intelligent machines. Blakeslee has aided him in presenting this theory so that it is accessible by the general public. I am very impressed by the theory of the cortex, but I do not agree that the cortex is the only way to achieve intelligence.

Hawkins defines intelligence as the ability to make predictions. I think this is an excellent definition of intelligence.

He says the cortex makes predictions via memory. The rat in the maze has a memory which includes both the motor activity of turning right and the experience of food. This activates turning right again, which is equivalent to the prediction that if he turns right, food will occur.

The primate visual system, which is the sense best understood, has four cortical areas that are in a hierarchy. In the lowest area, at the back of the head, cells respond to edges in particular locations, sometimes to edges moving in specific directions. In the highest area you can find cells that respond to faces, sometimes particular faces, such as the face of Bill Clinton.

But the microscopic appearance of the cortex is basically the same everywhere. There is not even much difference between motor cortex and sensory cortex. The book makes sense of the connections found in all areas of the cortex.

The cortex is a sheet covering the brain composed of small adjacent columns of cells, each with six layers. Information from a lower cortical area excites the layer 4 of a column. Layer 4 cells excite cells in layers 2 and 3 of the same column, which in turn excite cells in layers 5 and 6. Layers 2 and 3 have connections to the higher cortical area. Layer 5 has motor connections (the visual area affects eye movements) and layer 6 connects to the lower cortical area. Layer 6 goes to the long fibers in layer 1 of the area below, which can excite layers 2 and or 3 in many columns.

So there are two ways of exciting a column. Either by the area below stimulating layer 4, or by the area above stimulating layers 2 and 3. The synapses from the area above are far from the cell bodies of the neurons, but Hawkins suggests that synapses far from the cell body may fire a cell if several synapses are activated simultaneously.

The lowest area, at the back of the head, is not actually the beginning of processing. It receives input from the thalamus, in the middle of the brain (which receives input from the eyes). Cells in the thalamus respond to small circle of light, and the first stage of processing is to convert this response to spots to response to moving edges.

And the highest visual area is not the end of the story. It connects to multisensory areas of the cortex, where vision is combined with hearing and touch, etc.

The very highest area is not cortex at all, but the hippocampus.

Perception always involves prediction. When we look at a face, our fixation point is constantly shifting, and we predict what the result of the next fixation will be.

According to Hawkins, when an area of the cortex knows what it is perceiving, it sends to the area below information on the name of the sequence, and where we are in the sequence. If the next item in the sequence agrees with what the higher area thought it should be, the lower area sends no information back up. But if something unexpected occurs, it transmits information up. If the higher area can interpret the event, it revises its output to the lower area, and sends nothing to the area above it.

But truly unexpected events will percolate all the way up to the hippocampus. It is the hippocampus that processes the truly novel, eventually storing the once novel sequence in the cortex. If the hippocampus on both sides is destroyed, the person may still be intelligent, but can learn nothing new (at least, no new declarative memory).

When building an artificial auto-associative memory, which can learn sequences, it is necessary to build in a delay so that the next item will be predicted when it will occur. Hawkins suggests that the necessary delay is embodied in the feedback loop between layer 5 and the nonspecific areas of the thalamus. A cell in a nonspecific thalamic area may stimulate many cortical cells.

I think this theory of how the cortex works makes a lot of sense, and I am grateful to Hawkins and Blakeslee for writing it in a book that is accessible to people with limited AI and neuroscience.

But I am not convinced that the mammalian cortex is the only way to achieve intelligence. Hawkins suggests that the rat walks and sniffs with its "reptilian brain", but needs the cortex to learn the correct turn in the maze. But alligators can learn mazes using only their reptilian brains. I would have been quite surprised if they could not.

Even bees can predict, using a brain of one cubic millimeter. Not only can they learn to locate a bowl of sugar water, if you move the bowl a little further away each day, the bee will go to the correct predicted location rather than to the last experienced location.

And large-brained birds achieve primate levels of intelligence without a cortex. The part of the forebrain that is enlarged in highly intelligent birds has a nuclear rather than a laminar (layered) structure. The parrot Alex had language and intelligence equivalent to a two year old human, and Aesop's fable of the crow that figured out to get what he wanted from the surface of the water by dropping stones in the water and raising the water level, has been replicated in crows presented with the problem.

10 people found this helpful

How and why the brain works

0Comment

ByRobert MorrisHALL OF FAMETOP 500 REVIEWERVINE VOICEon September 7, 2011

I read this book when it was first published (in 2004) and recently re-read it while preparing for an interview of one of countless thought leaders who have acknowledged their great debt to Jeff Hawkins for what they have learned from him and, especially, for what they learned from this book. Written with Sandra Blakeslee, this book provides a superb discussion of topics that include

o Artificial intelligence
o Neural networks
o The structure and functions of the human brain
o A "new framework of intelligence" (more about that later)
o How the cortex works
o Consciousness and creativity
o Hawkins' thoughts about the future of intelligence

As Hawkins explains, his goal "is to explain [his] new theory of intelligence and how the brain works in a way that anybody will understand." However, I hasten to add, this is not a book written for dummies and idiots who wish to "fool" people into thinking they know and understand more than in fact they do.

Early on, Hawkins acknowledges his skepticism about artificial intelligence (AI) for reasons that are best explained within his narrative, in context. However, it can be said now that after extensive research, Hawkins concluded that three separate but related components are essential to understanding the brain: "My first criterion was the inclusion of time in brain function...The second criterion was the inclusion of feedback...The third criterion was that any theory or model of the brain should account for the physical architecture of the brain." AI capabilities, Hawkins notes, are severely limited in terms of (a) creating programs that replicate what the human mind can do, (b) must be perfect to work at all, and (c) AI "might lead to useful products, but it isn't going to build truly intelligence machines." At least not until we gain a much better understanding of the human brain.

The material in Chapter 7, "Consciousness and Creativity," is of special interest to me as I continue to read recently published books that offer breakthrough insights on creativity, innovation, and the processes by which to develop them. (The authors of many of those books, to borrow from a 12th century French monk, Bernard of Chartres, are standing on Dawkins' "shoulders." It must be getting crowded up there.) Hawkins asserts that creativity does not require high intelligence and giftedness, and defines creativity as "making predictions by analogy, something that occurs everywhere in cortex and something you do continually while awake. Creativity occurs along a continuum...At a fundamental level, everyday acts of perception are similar to rare flights of brilliance. It's just that the everyday acts are so common we don't notice them." I call this phenomenon "the invisibility of the obvious."

I am among those who are curious to know the answers to questions such as "Why are some people more creative than others?" "Can you train yourself to be more creative?" "What is consciousness?" and "What is imagination?" Hawkins has formulated answers to these and other questions and shares them in this chapter. Much of the structure of the "new framework of intelligence" to which I referred earlier is in place by the conclusion of this chapter. Then Hawkins concludes the book by looking to the future and offers with eleven predictions. Here's #8: "Sudden understanding should result in a precise cascading of predictive activity that flows down the cortical hierarchy." In other words, revelations (whatever their nature and scope) help us, not only to connect dots but to connect those that are most important.

One person found this helpful

0Comment

An important book

Byalgo41on October 18, 2007

First, let me comment on the writing. There is an obvious attempt to make things as clear as possible to the layman, with almost too many illustrations of some of the points, and only as much technical language as is necessary. Some of the imagery is great. Never-the-less, getting through the long chapter on "How the Cortex Works" is a chore. At the same time, the Appendix, whose primary purpose is to lay out a research program, is clear, concise, and very informative; in other words, the appendix should have been incorporated into the chapter, and some of the chapter details left for an Appendix. It doesn't help that while hierarchy is emphasized, within the core unit of the cortex, the 6 layered "column", the flow of information is not primarily upward or downward.

A key observation is that tasks which are complex or impossible to solve by computer, such as determining if a cat is pictured in a photograph, can be accomplished by the brain in less than 100 steps (we know this from the time it takes neurons to fire). Another is that inside the brain it is dark and silent: the cortex is always simply processing spatial/temporal patterns of impulses, whether these originated: outside of the cortex, as sounds, images, etc.; feedback from the body's own activity such as moving or lifting an object; thoughts generated within the cortex. All regions of the cortex look much the same, as best we can tell - they do functionally different things, but Hawkins infers they use the same basic algorithm(s) everywhere. Another observation is that information must be stored in invariant form, so that an a face can be recognized despite the lighting, angle, and so on, which all drastically affect the actual "pixels" which are recorded on the retina.

Hawkins sees the cortex as an auto-associative memory, which stores patterns in an invariant, hierarchical form, and can recall a complete pattern from part of the pattern, and even if the inputs are somewhat distorted (which is why we never notice blind spots in the retina). The cortex is constantly using this capability to predict what pattern it will see next, and to compare it to actual patterns: if the prediction is incorrect, then this information is moved up the hierarchy and learning may occur as new and often more general kinds of classifications (invariant representations) are made dynamically.

Patterns can correspond to concepts as well as the output of the physical senses. In fact, my appreciation of Hawkins' book was greatly enhanced by having previously read Jerome Feldman's " From Molecule to Metaphor", which seems to take a very different approach to the brain in explaining how the child masters language (and is very different as to the actual mechanics, suggesting the use of what Hawkins calls backward propagation neural networks rather than auto-associative networks, the latter making more sense to me). What Feldman makes clear is how we bootstrap learning using analogy (comparable to invariant patterns), so that abstract concepts can be seen as originally built from analogy to models of physical movement and grasping and then get increasingly abstract. Interestingly, just as Feldman starts with concepts of motor control as the basis of language, Hawkins points out a predicted pattern can also correspond to a series of instructions for muscular movement. Hawkins defines creativity as "prediction by analogy" (p.183).

One interesting prediction that Hawkins makes is that neurons will be found to be smarter than mere aggregators which fire only if the sum of positive minus negative inputs exceed some threshold; instead, he thinks at least some neurons also have the capability to fire if certain inputs fire together without respect to an aggregate threshold. He also speculates on why sounds seem different than images, acknowledging that this may have to do with the non-cortical areas of the brain, just as these areas are so important to emotions.

3 people found this helpful

The Intelligence War between Humans and Computers

0Comment

ByJacob Johnsonon September 29, 2013

As a student at Georgia Tech studying Biomedical engineering, I decided to take an Intro to Neuroscience class this fall semester. One of the assignments that was given to us during the first week of school was to pick a Neuroscience related book, read it, and write a review on Amazon. Having not a clue of what book to choose, I searched through Google and Amazon in order to find an enjoyable read that would prove to be both intriguing and insightful. Stumbling upon Jeff Hawkins book, On Intelligence, I had no idea what to expect. From the summaries, I could only guess that this book would relate the intelligence processed between humans and computers. In addition, I discovered that Hawkins was the original inventor of the once famed Palm Pilot that took America by storm just prior to the new millennium.

It is rare that I come across and academic-related book and find myself fully engaged. However, On Intelligence seemed to do exactly that. Hawkins provides an alternate perception of how the brain works and gives us a secret mechanism that can be unlocked to predict the future. This new approached immediately substituted my view on the way the brain functions. Having always thought that the brain was simply a computer executing various commands, I quickly learned that the real function is its ability to make predictions about the future. Explaining the intricate details of the individual mechanisms of the cortex, we quickly learn how the brain can build intelligent machines. "Prediction is not just one of the things your brain does. It is the primary function of the neo-cortex, and the foundation of intelligence". What was even more fascinating was the concept of how we can make these predictions base on memories that are built in our brain overtime. On Intelligence, was a truly an eye-opening book for me and forced me to turn to the next page after each chapter.

Hawkins gives his own outline of the book in order for the reader to better understand the goal of his writing. Starting with a background on some of the previous attempts at understanding intelligence and how those theories have failed, Hawkins develops an essential theory of what he calls the memory-prediction framework. He uses thought experiments to illustrate the extensiveness of prediction as well as evolutionary comparisons to explain the brains function as it relates to intelligence. Although Hawkins doesn't dwell to long on the biological functions and mechanisms of how these processes are performed, he does spend quite a bit of time on the methods of operation of how the cortex utilizes a hierarchy of invariant memories and sequences to make futuristic predictions. The shift is then made towards the final chapters of the book in which Hawkins starts tying the connections between intelligence and the creative side of the brain. He explains how this source of power can be used to our advantage and determines the success rate for years to come. Lastly, Hawkins makes some predictions of involving the fears of intelligent machines, "Throughout the twenty-first century, intelligent machines will emerge from the realm of science fiction into fact". This final section made me ponder the reality of all of those science fiction books and movies where AI can develop the impossible characteristics of emotions, creativity and self-learning. Most of these plots nearly always have a negative outcome.

Hawkins choice of style and structure of the book proves to be an interesting one. He immediately delivers his personal experiences by explaining the method in which he came upon his theory. The first chapter is dedicated to his personal history and how he became interested and involved in the area of neuroscience. Failing to be accepting into MIT, Hawkins explains the scientific establishment has always rejected the link between neuroscience and artificial intelligence. After explaining his past and his previous thoughts and ideas, Hawkins dives into the main section of the book that includes his primary theory. As stated before, he develops the concept that the brain is a mechanism that has the ability to predict the future. Hawkins predicts that in all areas of the cortex, "anticipatory cells" can be found that fire only in anticipation of a sensory event. He then goes on to explain the primary functions of each of the cortexes throughout corresponding chapters.

Hawkins is able to attract the reader so quickly by his ways of illustrating specific examples or inquiring about memory tests for the reader to actually perform. He uses these concepts and analogies in order to better explain the framework for his ideas and concepts of cortex function. I think these methods were extremely crucial in achieving the goal set out by Hawkins; to enlighten all audiences interested in how the brain functions. Although many neuroscientists today can agree that very little is actually known about the processes and mechanisms of neural functions with validation, Hawkins provides his alternative approach to possibly diminish this vast gap of our understanding. Although some readers may be overwhelmed by the enormous amounts of information confined within this book, I believe it still gets the message across in a significant way as to avoid the loss of the concept and still spark the interest for future understanding.

My recommendation for future readers would be to do some outside research prior to reading this book. Although you can come in with absolutely no knowledge of brain function, it would not hurt to get some basic knowledge of the different systems, such as the visual, auditory and sensory systems work in individually and in unison. This will most likely reduce the overwhelming factor previously mentioned and truly aid in your fascination of the theories described. Perhaps check out some other neuroscience books as well. However, when it comes to the primary focus of this book and how it relates to artificial intelligence, you will not find a better read. For that case, I have given Jeff Hawkin's masterpiece, On Intelligence, 5 out of 5 stars.

4 people found this helpful

0Comment

An exceptional review

ByMichael J. EdelmanTOP 500 REVIEWERVINE VOICEon March 23, 2007

There's really nothing new in Hawkin's book- but this is still a very useful (not to mention fascinating) volume. The basic notion of mind as associative memory, in which learning, perception and cognition are all part of the same process, can be certainly be traced back to Pribram, if not earlier, and perhaps even Hebb (whom Hawkins cites). And the notion of a generalized system of perception without modlaity specific mechanisms is certainly as old.

What Hawkins does is bring together a lot of information from areas that haven't talked to each other much, as well as theory and experiments that has been neglected by modern AI and cognitive theorists. His advantage is that he comes into the debate on mind and brain without, as they say, a dog in this fight. Unlike so many AI researchers, cognitive theorists and philosophers, he's not wedded to a paradigm that he's based an acdemic career on. He's obviously read not only the psychology, neurobiology and AI literature, but also the early work of people like Weiner, Lettvin, McCullough, Pitts, and others who came at the problem from an enginnering background, and saw the generalizability of neuonal networks where physiologists might have been inclned to see organs and specialization.

I can't say I agree one hundred percent with everything Hawkins proposes, and I think he is perhaps a bit too dismissive of the philisophical issues. But if you're interested in any of the fields I've mentioned, I think you'll find this to be an excellent read.

6 people found this helpful

Jeff Hawkins has probably dug deeper than anyone else on ...

0Comment

ByAmazon Customeron August 26, 2015

Jeff Hawkins has probably dug deeper than anyone else on the topic of a common cortical model, and how to exploit that idea to build working systems. The common-sense balance he strikes between emulation of the brain and functional equivalence is rare in this field, and it's the key to his approach and its convincing nature. His ideas are explained clearly in this book, and his passion for solving this problem is apparent. I encourage anyone interested in AGI or AI to read this book.

An Intelligent and Readable Book on Intelligence

0Comment

ByOliveron January 12, 2007

Jeff Hawkins is the founder of Palm Computer, and the inventor of the Palm Pilot and Treo. After making his fortune, Hawkins turned his attention to neuroscience. Given that history, I was afraid that this book was only published because Hawkins is rich, successful and presumed smart. In fact, Hawkins is smart. More importantly, he has some very good ideas about how the brain works, and he presents them in a clear and concise way. This is an excellent book.

Hawkins presents a theory of how the brain makes predictions. Questions that are easily solved are solved at a lower level. If they cannot be solved, they move up to the next level -- something like. I'll let Hawkins explain it. He does a much better job.

"On Intelligence" could easily have been titled "How the Mind Works." In fact, that title is taken by another wonderful scientist and writer, Steven Pinker. The two books have very little in common after that. I highly recommend both.

2 people found this helpful

Unified Theory of Intelligence with Machine Applications

0Comment

ByMark Leeon November 16, 2005

Jeff Hawkins is "crazy about brains." In this readable book, the electronic engineer combines his training as a computer designer and his self-education into the seat of human intelligence and posits how he believes we can make machine intelligence.

Hawkins believes the major historical mistake of AI is in having ignored the human brain's design and structure. What we need isn't more power (because today's computers run much faster than the electrochemical synaptic reset times of the human brain), but better design. The components of Hawkins' synthetic brain would include the incorporation of time as a function, the recognition of the importance of feedback, and a reckoning of the brain's architecture.

Hawkins is also critical of older AI models which suggested that behavior is the primary indicator of intelligence. He observes that we can be intelligent, quietly, in a dark room. One of my criticisms is that Hawkins observes that we probably have built in to our human brains old code no longer needed; remnants of "legacy code." I'd suggest, though, that one man's "legacy code" might really contain essential, cryptic subroutines. Regardless, Hawkins has great respect for the natural development that has resulted in the human brain.

In short, Hawkins develops his theme as the brain being a repository of data and streams of new input with resulting feedback from which and in which the brain seeks patterns. It's the difference between established patterns, acceptable variants, and new material that makes up the bulk of what our brain does. And it is the anticipation of patterns and acceptable variants that makes up intelligence. I have been a disciple of Doug Hofstadter (e.g. Godel, Escher, Bach: An Eternal Golden Braid) and his "patterns and recursion" look at intelligence for quite some time, so taking a few more steps as required by Hawkins wasn't particularly difficult for me.

The chapter on the function of the cortex was the most difficult and enjoyable for me, with his conclusions and look forward being the icing on the cake. All in all a very enjoyable look at one man's vision for the future of intelligent machines in one nice, tidy, unified presentation.

10 people found this helpful