A Thousand Brains: A New Theory of Intelligence

This subject is so interesting to me that I bought Jeff Hawkins' first book, and now I find I've sprung for his second book.The first part of it is interesting. A rather long read, but interesting and entertaining; logical, but not really much progress in figuring out how the brain works since the first book, "On Intelligence". This book just has the feel of a scam to prey on people who want to know more about the brain, intelligence. Hawkins gives us I guess the parts he thinks he wants to tell regular people about, since these concepts if they are real are worth lots of money, so it's like we get a small smattering of a few definitions ... not a lot of meat. ( my apologies to vegans )The problem is the whole book taken together, after the Thousand Brains part, is semi-incoherent in direction or purpose.First he talks about the brain, and the neocortical columns ... not very much there but a dictionary-like definition. Then he says that there are 150,000 of them and that is where the Thousand Brains theory comes into play, which is really a clever branding, not so much a theory, at least one developed very far here.Then Hawkins jumps into talking about how today's AI has problems, which we all know. But he goes on then to pat himself on the back for Palm, which I am sure made tons of money but to me anyway was a rotten product as ahead of its time as to be almost useless, but popular and useful for some people and marketed well. Then he talks about how his Palm IP and device-centric theories, that he seems to want to associate himself with the iPhone and like devices, which he did not really make. Maybe there are some Palm pieces in the iPhone, or Palm lawsuits, I don't know? But because he was a visionary, he then goes on to imply that he is right again about some new kind of AI that he cannot even describe, except that it is better than neural nets. Neural nets which cannot learn. I agree with a lot of his assessments that AI is not all it is cracked up to be. Some good points about humans not really being intelligent ... since we are too stupid to do anything about the fact that we know we are killing ourselves. What is intelligence? He fails to ask or answer that question.I don't know if I want my car to be learning, or conscious, or anything else, acting independently or perhaps making a mistake - as is the primary way humans seem to learn. I don't like it, I don't want it, I don't need it. I don't much like the AI stuff we have today, or at least the way it is used against us, but I sure as hell do not want something with a brain inside the devices I use. But Hawkins does not speculate much or sensibly on that, which could be interesting and I am sure he has some ideas about.Instead he goes on to talk about the fate of humanity in infinite time, preserving intelligence by space devices - like shades that send Morse code to other astronomical civilization that might be looking at the stars in their skies. This is so off the wall and outside the subject matter it really had me questioning Hawkin's sanity or mental integrity. These ideas and speculations were not even very good in that they are common ideas as far back as the first Cosmos series or before. Filler in an already thin on real facts book.I can't say I really regret getting the book, but that doesn't mean I have to think it is great, or be uncritical. Aside from about 1/10th of the book most of it was useless and a waste of time. Sorry if that is harsh - but you took my money for this.3/10 or 1/5 if I have to.

I bought this book expecting great things: an independent perspective and a stimulating review of the latest thinking in cognitive science. Unfortunately, this new book offers little that wasn't already well-written in his previous work.That book, On Intelligence, was a bible of sorts for those of us who take seriously the symbolic nature of mental processing. Ideas like "the way the cortex processes signals from the ear is the same as the way it processes signals from the eyes" and that "memory recall almost always follows a pathway of association." "The cortex," he previously wrote, "creates what are called invariant representations, which handle variations in the world automatically" and "directs behavior to satisfy its predictions." "Perception and behavior are almost one and the same."While not all originally Hawkins' ideas, he presented a compelling narrative in his previous book. "As long as we can decipher the neocortical algorithm and come up with a science of patterns," he wrote, "we can apply it to any system that we want to make intelligent."Exciting stuff! I was hoping to be equally inspired by this new work. Unfortunately, while competently written, it left me wanting more.

Jeff Hawkins's book is really three books.The first is a history of his lifelong quest to understand how the brain and human cognition work. He spent years failing to find an academic discipline that could contain his ideas. He finally gave up, founded two successful computer companies, got rich, and founded Numenta, his private research project.The second is about how the neocortex creates general intelligence, which he defines as the ability to create a constantly updating model of the world. Billions of neurons form trillions of connections that combine into patterns that represent fragments of reality that then "vote" to form an overall picture.The third is a non-mainstream discourse on artificial intelligence. General intelligence, which is in our future, is completely different from today's narrow-purpose AI, which cannot create a constantly evolving world model. Machines, he says, can be "conscious" without having emotions unless we so imbue them, and thus, need be neither moral agents nor necessary threats to humanity. However, machines are perhaps the evolutionary successor to humans, something that Jeff regards as good because they could preserve human knowledge without the archaic limits of the body.This book is the successor to the 2004 book ON INTELLIGENCE, which taught me more about human cognition than anything else I've read, including Kahneman and Tversky. Both books treat deeply technical subjects in a highly readable way.

This book can be considered a continuation of the previous book : On Intelligence - 2004.Obviously a lot happened since then. Hawkins founded Numenta in 2005.Hawkins proposed The Thousand Brains Theory of Intelligence in 2018 - it is a a new framework for intelligence and cortical computation. Since then he was working tirelesly to put that concept in a book, it took him 18 months to write it. This book contains result of 15 years of research that was done by Jeff with his Numenta team since his previous book. He believes the current deep learning technology will not be able to create intelligent machines. On the ther hand he thinks his theory is the right path to that goal. I find him very convincing. I think this book is going to be very influential in the AI community in this decade and even longer.

The first thing to say: the book is about a new theory of how the brain works from a neuroscientific perspective, but it's written (on purpose) in such an accessible language that anyone can enjoy it.The second and equally important thing to say: the book is divided into three parts. The first one is dedicated to the theory, the second one is dedicated to the future of AI, and the third part is dedicated to the possible future of humankind.You only need to read the first part to understand the theory and, if you think it's worth exploring more, you can augment that with the more complex, academic papers listed in the Recommended Reading section at the end of the book.This part of the book is the one I'd recommend to anybody interested in the latest progress in neuroscience. Contrarian data scientists and machine learning engineers interested in designing AI should read the book, too.My rating is focused on this first part of the book only. It's a 4/5 because I'd have preferred Hawkins dedicating more time to the more technical details (maybe in an optional appendix) rather than digressing in the way he did with parts two and three.It's also worth saying that the theory is incomplete and, even after reading the research papers that support it, you might still have many unanswered questions. That doesn't mean that the idea is not valuable and thought-provoking. Just that it needs more time to be fully confirmed experimentally and further expanded where it still lacks clarity.The second part is only partially related to the theory of the Thousand Brains. I'm afraid I have to disagree with some assumptions related to the risks of truly intelligent machines. Hawkins can envision very complex scenarios about the inner workings of the brain. Still, he doesn't want to contemplate more straightforward scenarios where, for example, interconnected intelligent machines learn about human discoveries in real-time and, in practicality, learn faster than humans about how the implications of those discoveries and adopted them more quickly than us.Overall, I don't think it's worth focusing too much on this section, and nobody should use it as a lens to judge the theory itself. My rating completely ignores it.The third part of the book is almost entirely unrelated to the theory of the Thousand Brains. It contains many interesting, occasionally controversial, perspectives and ideas about how our species could evolve and what should be our purpose in the future. However, that content (together with part two or by itself) could be better suited in a dedicated book.I understand Hawkins' desire to share his views on the topics he touches and use the book to call for action. But, as it is, it's more a distraction from the theory than anything else.Just like for part two, my rating completely ignores this content.Overall, this is a must-read book if you are interested in novel ideas about how the brain works and what intelligence is all about. However, I recommend focusing on part one, then continue by reading the research papers, and only after, as a sort of unrelated bonus, read parts two and three.***How I review books***5 stars - an exceptional book that expands my reasoning, not just my knowledge.4 stars - a great book that significantly expands my knowledge.3 stars - a book with some interesting information and some major flaws that didn't really make an impact in my life.(Notice that there's a time in life for certain books. It's possible this was not the time for this book and the review rating would change 10 years from now)2 stars - a book that gave me nothing and took my time.1 star - a book so poorly written that I couldn't even finish reading.

The book was sold as new from Amazon EU S.a.r.L and arrived with pristine packaging but very clear evidence of water damage. Pages not stuck together, still readable but pages throughut book cockled and some light mould growth within the front cover. Clearly it was dried prior to posting and the damage must have been evident to the person (if there was one) that filled the package prior to posting. Its the first time that I have ever had a damaged product from Amazon. Im not going to retun it because life is too short to walk all the way back to the post office. I will keep the book but it is not woth the purchase price and I am very dissapointed.

No matter how much I thought I knew about the brain, as a medical doctor I learnt the anatomy of the brain, I am mesmerised with this book. It has opened a door that I did not out existed. Incredibly fascinating and full of knowledge. Very highly recommended

I found the theory expounded upon in this book so interesting and exciting. I can't stop thinking about cortical columns and how they work, as well as how a person thinks things through, how prediction is so central to cognition and the notion of voting. Hawkins' opinions on AI were very interesting and enhanced the neuroscience that preceded it. The rest of the book felt distracting. I think Jeff Hawkins would have done well to put this into a sequel, as it covered very different - if associated ground to the first part of the book.