A Thousand Brains: A New Theory of Intelligence
- Downloads：9480
- Type：Epub+TxT+PDF+Mobi
- Create Date：2022-12-23 09:51:38
- Update Date：2025-09-06
- Status：finish
- Author：Jeff Hawkins
- ISBN：1541675797
- Environment：PC/Android/iPhone/iPad/Kindle
- Download
Summary
A bestselling author， neuroscientist， and computer engineer unveils a theory of intelligence that will revolutionize our understanding of the brain and the future of AI。
For all of neuroscience's advances， we've made little progress on its biggest question: How do simple cells in the brain create intelligence？
Jeff Hawkins and his team discovered that the brain uses maplike structures to build a model of the world—not just one model， but hundreds of thousands of models of everything we know。 This discovery allows Hawkins to answer important questions about how we perceive the world， why we have a sense of self， and the origin of high-level thought。
A Thousand Brains heralds a revolution in the understanding of intelligence。 It is a big-think book， in every sense of the word。
One of the Financial Times' Best Books of 2021
One of Bill Gates' Five Favorite Books of 2021
For all of neuroscience's advances， we've made little progress on its biggest question: How do simple cells in the brain create intelligence？
Jeff Hawkins and his team discovered that the brain uses maplike structures to build a model of the world—not just one model， but hundreds of thousands of models of everything we know。 This discovery allows Hawkins to answer important questions about how we perceive the world， why we have a sense of self， and the origin of high-level thought。
A Thousand Brains heralds a revolution in the understanding of intelligence。 It is a big-think book， in every sense of the word。
One of the Financial Times' Best Books of 2021
One of Bill Gates' Five Favorite Books of 2021
Reviews
Kailuo Wang
，
The theory is rather intriguing， though in my view， it left some important questions。 This is less a review than a compilation of my thoughts regarding the theory。The theory （referred to as ATB below）'s core is that each column learns a model of the complete object; using them， they produce signals based on inputs which then pass to other columns。 ATB is convincing in regards to the theory that it's not a hierarchical structure in which higher-layer columns learn and represent objects based on a The theory is rather intriguing， though in my view， it left some important questions。 This is less a review than a compilation of my thoughts regarding the theory。The theory （referred to as ATB below）'s core is that each column learns a model of the complete object; using them， they produce signals based on inputs which then pass to other columns。 ATB is convincing in regards to the theory that it's not a hierarchical structure in which higher-layer columns learn and represent objects based on abstract features recognized by lower layers ones。 Instead， each column learn through prediction errors in two ways: 1） errors in their own predictions of their inputs based on previous input as well as signals from other columns。 and 2） errors in other columns' predictions based on signals provided by them。However， in my view， ATB's claim that each column models the "complete object" is overly strong in two ways: First， "modeling the complete object" leaves the impression that a） it alone can recognize the specific object as a unique entity in the world， and b） it can predict all possible sensory input values from that object to the column。 Both sound too strong a power for a single column to possess。 More importantly， they don't NEED to be that powerful。 The two prediction error learning mentioned above do not require models of complete objects in each column。 All they need， to be successful in those learnings， is to extract feature signals that help themselves and other columns predict。 Thus， instead of producing a signal pointing to a specific object， the signal could be simply indicating the kind of the object - a rough classification。 Instead of an internal complete object model capable of predicting future inputs from previous input， a column can always incorporate signals from other columns to make predictions of future inputs。 Here is a thought experiment of your hand touching a doll。 One column， taking inputs from finger-tip temperature and touch sensor， might send a signal indicating it's a rubber-made object; another column， taking in the relative positions of your five fingers， might send a signal indicating it's an object with a body shape。Both indicators can help the other column make predictions of future inputs when the fingers move。 The finger-tip column can predict that temperature and texture remain unchanged while fingers move around the doll， because it learned from past experience that when there is an "it's a body" signal， the object has uniform temperature and texture throughout。 On the other side， the fingers position column can predict that this object remains still while being touched。 Again， past experience with the "it's rubber" signal is enough for the column to make the correct prediction。 In this thought experiment， neither column produced an "it's a doll" signal， nor did they need a model for dolls to predict future inputs。 It is all these columns working together， without a higher-level column congregating their signals， that gives rise to the ability to predict all sensory inputs from a doll， i。e。 a prediction model of dolls。Thus， at least， at the perception level， information between columns can be flat。 However， it's hard to imagine higher-level cognitive faculty without some form of hierarchical information processing。 ATB proposed that such a hierarchy corresponds to the hierarchy of objects in the real world。 This might be a bit too speculative。 Columns learn from prediction errors。 They can predict raw sensory inputs; they can also predict signals by other columns produced from sensory inputs。 Thus， learning can happen when there are raw sensory input prediction errors as well as when there are other column signal prediction errors。 Learning in columns can easily be hierarchical - naturally， models （or knowledge） learned from them are hierarchical。 That being said， there is no reason to believe that the hierarchy is a neat pyramid with clear-cut division between layers。 Any column can learn from any other columns as long as their signals are useful。 It's just that learning， and thus models， can happen orders away from raw input signals。 Lastly， speaking of high-level cognitive functions such as reasoning， math， etc。 ATB proposed they are all based on reference frames。 Again it didn't provide a clear picture of the mechanism。 I believe it missed a crucial fact that most of these high-level cognitive functions are enabled by language。 Here is my speculation: high-level concepts and relationships between them exist linguistically in our brains， and cognitive functions based on these concepts and relationships are also encoded in sentences-like linguistic memories。 Our brains have the faculty toStore the models of the world in the sentences like linguistic memory。 E。g。 Deers come to this spot when there is a drought。 Process information based on the models expressed by these sentences。 E。g。 there is a drought now， we should hunt deers at this spot。 High-level human cognitive functions are the enterprise of our braining employing these two faculties。 We don't have dedicated circuitries for each model expressed in linguistic memory， we just need the basic circuitries for language processing。 If you are still here， thanks for reading this very long thread。 These thoughts are also published in the mind。net for easier consumptionhttps://www。themind。net/mapItems/Hypo。。。 。。。more
Steven
，
Was ok interesting theory about primitive brain and newer parts of brain and worked in with ideas of future AI intelligence 。 Not sure I agreed with all his arguments but had some interesting ideas。
Robin Messenger
，
The main thrust of this book deserves a 5 star rating， however the author unnecessarily oversteps the scope of his expertise in parts。 This book outlines an alternative conceptual framework for understanding the function of the neocortex in humans and other mammels。 The author uses clear arguments， accessible to non-experts， to argue for it。The author does go off the rails a bit in a few places。 Firstly， he confidently presents half-baked musings on Chalmers' hard problem of consciousness， witho The main thrust of this book deserves a 5 star rating， however the author unnecessarily oversteps the scope of his expertise in parts。 This book outlines an alternative conceptual framework for understanding the function of the neocortex in humans and other mammels。 The author uses clear arguments， accessible to non-experts， to argue for it。The author does go off the rails a bit in a few places。 Firstly， he confidently presents half-baked musings on Chalmers' hard problem of consciousness， without seeming to have a good understanding of it。 He seems to conflate the function of a cortical column （creating models and reference frames， etc） with the subjective experience of awareness that goes along with it。 For example， he argues that if a cortical column could create a model of the salient features of qualia， then it would solve the hard problem， without considering that such modeling could exist with or without subjective experience in principle。Secondly， he seems not to have fully thought out his ideas about the role of the "new brain" （neocortex）， and the "old brain"。 He explains that the "old brain" gives us volition -- the drive to take action in the world to affect some outcome -- and the "new brain" is merely a tool to help us more effectively carry out that volition。 He then goes on to proclaim that we can create a better world by suppressing the negative influences of the "old brain"， using the rational "new brain"。 He seems unaware that this recommendation is a contradiction of his own model of the brain。 If the "old brain" gives us volition， and the "new brain" helps us to act on that volition， then any role the "new brain" plays in， for example， supressing undesirable actions would only be in service of some other goal of the "old brain"。 For example， if you have the urge to steal someone's sandwich， but your "new brain" tells you not to， it's not because your "new brain" has acheived some victory over your "old brain"， but rather that your "new brain" has the forsight to predict that stealing sandwhiches may lead to negative outcomes （as defined by your "old brain"） for you in the long run。 In the author's own view， the "new brain" merely helps maximize values defined by the "old brain"， and has no values or goals of its own。 Some consistency would have been nice here。Overall， it would be a lot easier to recommend this book if it stuck to the neuroscience， and skipped the tangential musings about the future， AI， morality， etc。 。。。more
Sabin
，
This review is going to be a hard one。 And， I'm afraid， also an incomplete one。 I finished the book eight months ago， but this year they're all getting reviews even if I won’t remember everything I should。This book is about the author's theory on the way brains function at a general level and also aims to explain the emergent property of intelligence in what amounts to little more than compact bundles of nerve cells。I've read my fair share of books on neuroscience and AI， two domains which conti This review is going to be a hard one。 And， I'm afraid， also an incomplete one。 I finished the book eight months ago， but this year they're all getting reviews even if I won’t remember everything I should。This book is about the author's theory on the way brains function at a general level and also aims to explain the emergent property of intelligence in what amounts to little more than compact bundles of nerve cells。I've read my fair share of books on neuroscience and AI， two domains which continue to inform each other more and more lately， and feel that I have a fairly adequate grasp of their main thesis。（The following three hidden paragraphs talk about other books I've read but mostly not bothered to review – as a result of me being a lazy bastard。 I'll probably reference them in this review but they're mainly here for my satisfaction）（view spoiler）[Anil Seth， in Being You， chips aways at the hard problem of consciousness by attempting， in small incremental steps， to bring it into the realm of the possibility of knowledge。 His hope is that by solving problems which fall into the realm of scientific knowledge we can get enough answers so that the hard problem disappears。Micahel Graziano， in Consciousness and the Social Brain applies the concept of a mental model to everything that the mind does， and explains consciousness within this framework。 The fact that we hold in our brains not only a model of us， as the "home team" consciousness and our environment， but also of other people's consciousnesses explains a great number of behaviours and biases inherent to social creatures。 He uses what he calls the attention schema to explain the utility of a process like consciousness， instead of grappling with its philosophical implications。In Human Compatible， Stuart Russel focuses on existential issues related to artificial intelligence and possible ways of ensuring that AI remains under human control。 What he brings to the discussion on neuroscience is its application in AI research and the fact that none of the approaches used in AI until 2018 （just future-proofing my reviews） achieves anything remotely resembling human consciousness。 （hide spoiler）]Jeff Hawkins is describing here an utility oriented theory， one that describes brain mechanisms which can achieve every manner of intelligent processing。 The concept of cortical columns， borrowed from Vernon Mountcastle， for which physical evidence seems to accumulate， splits the neocortex into around 150k autonomous brains which work together to create a unified experience。His claim is that these cortical columns， each about a spaghetti strand in width， contain all the types of neurons required to create a complete experience of the world based on the input received: a frame of reference。 So if one such column was connected to the optical nerve of one eye， it would create a complete experience based on what that nerve transmits。 If it were connected to the auditory nerve， it would create an experience based on what the ears sense。 Trying to put this info together now， I'm thinking that different cortical columns， from different regions of the brain， may connect to the same input， but process the info in different ways。 So we would get a column which deciphers language and one which deciphers tone from the same auditory input。 And the horizontal connections between the columns， which create these different frames of reference， integrate and unify our experience。The idea of multiple frames of reference ties in to Graziano's attention schema， where the brain holds the model of the self but also of other beings perceived as conscious and of the environment at the same time。 The difference is that Hawkins is solely concerned here with rigorously explaining how intelligence emerges， and not consciousness。 Consciousness is like an unavoidable side effect of the architecture of the brain。 Although it would have been interesting， I don't remember him discussing the possibility of consciousness or intelligence arising from different architectures， extraterrestrial or not。Another point which is， to my mind， insufficiently addressed， is artificial intelligence。 His arguments place him with the optimists， that when we do achieve breakthroughs in AI that could potentially harm us， some clever scientists will develop foolproof mechanisms to prevent them from actually doing any harm， unfortunately glossing over the evidence that AI is already harming human civilization。I cannot stress enough， the thousand brains theory is very clever and， most importantly， testable。 So expect it to be put to the test in future experiments and maybe iterated upon in the years to come （as of writing this， in 2022）。 It's just that the book itself is uneven。 In addition， the audiobook version isn't really something to write home about。 。。。more
Josef
，
Nice theory and thinking in mental models is superior。 Need to read a second time。
Arjen
，
I enjoyed parts 1 and 2 of the book， the author has a gift presenting complex brain matter to a lay audience without infantilizing the reader。 His theory of how the brain receives and processes information is fascinating and I hope I’ll be able to remember some of it 。 Section 2， on machine learning， is also convincing; we have nothing to fear of intelligent machines。 But then in part 3 it all came undone for me in the section on false beliefs and the existential risks of human intelligence。The I enjoyed parts 1 and 2 of the book， the author has a gift presenting complex brain matter to a lay audience without infantilizing the reader。 His theory of how the brain receives and processes information is fascinating and I hope I’ll be able to remember some of it 。 Section 2， on machine learning， is also convincing; we have nothing to fear of intelligent machines。 But then in part 3 it all came undone for me in the section on false beliefs and the existential risks of human intelligence。The writing was so balanced until this section: How do climate-change deniers maintain their false belief in the face of substantial physical evidence？ They are like flat-Earth believers: they don’t trust most other people， and they rely only on what they personally observe or what other similarly minded people tell them。 Temperature and water levels can be measured， but climate changes predictions are based on models， the models have assumptions that are， as the word implies， assumed and not measured。 A flat or round earth is not based on models but on measurements。 Then after dissing world religions as a function of old brain instinct， the author goes on to dream of a future of space colonization。 I guess that was the old brain overruling the neocortex， every brain needs its religion， whether its climate change religion， or spacism。It would have been infinitely more interesting to learn how the brain stores information in about one hundred billion neurons and several hundred trillion synapses that can be recalled at any moment。 Any recommendations for a book that covers that？ 。。。more
Lightspeed Petra catching up on reviews
，
Intelligence， to me， is the development of a reaction， probably chemical to start with。 If a single celled creature， say an amoeba， is able to turn away from obvious danger， there must be a time where it is faced with two threats and no where else to go。 Which one is the least dangerous？ All creatures must be able to decide that。 And to me that is where intelligence begins。The book is more about the neurological structures in the brain that give rise to intelligence。 I will get round to it one d Intelligence， to me， is the development of a reaction， probably chemical to start with。 If a single celled creature， say an amoeba， is able to turn away from obvious danger， there must be a time where it is faced with two threats and no where else to go。 Which one is the least dangerous？ All creatures must be able to decide that。 And to me that is where intelligence begins。The book is more about the neurological structures in the brain that give rise to intelligence。 I will get round to it one day。 What I'd like to know as well is why intelligence varies so greatly among people。 I don't mean from standardised intelligence tests which test more for frame-of-reference than anything innate in the brain。 Also some were developed by and at the behest of politicans who were determined to prove the superiority of the white 'race'。 （We really need to start teaching race as the conspiracy theory it is， there is only one human race， and teaching as we do has been remarkably unsuccessful）。 What I mean is why are some people real geniuses and others totally dense？ Why does it vary so much even within a family？ Are identical twins equal in intelligence？ From reading reviews of the book， it doesn't seem it is likely to answer these questions。 。。。more
Sadish Ravi
，
Interesting read。 Lot of learnings and new theories about how the brain works in the first half of the book。 Explained very well。 The second half is not that engaging and Is more around AI， human intelligence and them coming together。 Wish the second half went more deeper into the workings of the brain。
Felix Delong
，
The theory is awesome and enlightening。 But maybe the author shouldn't leave neurobiology and go into AI and existential risks。。。 while I liked that part， it went too far from the topic。 The theory is awesome and enlightening。 But maybe the author shouldn't leave neurobiology and go into AI and existential risks。。。 while I liked that part， it went too far from the topic。 。。。more
R
，
Initially it was interesting about how the brain works。。。 further down he delves into machine learning which was boring to read。。。。。 anyway it was somewhat interesting。。。。 but i wouldn't recommend it as a eye-opening stuff Initially it was interesting about how the brain works。。。 further down he delves into machine learning which was boring to read。。。。。 anyway it was somewhat interesting。。。。 but i wouldn't recommend it as a eye-opening stuff 。。。more
Muhammad Noor
，
I liked parts of the book， a revelation。 However there are sections in which I disagree with the conclusions that Hawkins have come up with。
Iolanda Ciobanu
，
A complicated theory explained for everyone， also giving a glimpse in the future of technology。
Jerry Wall
，
This review has been hidden because it contains spoilers。 To view it， click here。 Book fairly current on how does the brain work？New brain （neocortex） and old brain （all the rest of the reptilian or later brain。"the brain is the only thing we know of that is intelligent。" p。 124The truth is， we perceive our our model of the world， not the world itself。。 p。 175False beliefs exist and have 3 traits。1。 Cannot directly experience2。 Ignore contrary evidence。3。 Viral （rapid） spread。 p。 193 Book fairly current on how does the brain work？New brain （neocortex） and old brain （all the rest of the reptilian or later brain。"the brain is the only thing we know of that is intelligent。" p。 124The truth is， we perceive our our model of the world， not the world itself。。 p。 175False beliefs exist and have 3 traits。1。 Cannot directly experience2。 Ignore contrary evidence。3。 Viral （rapid） spread。 p。 193 。。。more
Marcello Povoa
，
I just finished reading "A Thousand Brains: A New Theory of Intelligence" by Jeff Hawkins。 I had my eye on this book since its recent release， and soon Mr。 Bill Gates and the Financial Times recommended it — which made reading inevitable。 Understanding how the human mind works is still one of the great mysteries of science。 The book shows a theory of how the brain produces intelligence， and reflects on the application of this knowledge to machines （Artificial Intelligence）。 For those who work in I just finished reading "A Thousand Brains: A New Theory of Intelligence" by Jeff Hawkins。 I had my eye on this book since its recent release， and soon Mr。 Bill Gates and the Financial Times recommended it — which made reading inevitable。 Understanding how the human mind works is still one of the great mysteries of science。 The book shows a theory of how the brain produces intelligence， and reflects on the application of this knowledge to machines （Artificial Intelligence）。 For those who work in the digital world， Jeff Hawkins is an old acquaintance: he was the founder of Palm， and a pioneer in "handheld computers"， even before the emergence of smartphones。 However， his passion has always been neuroscience。 Despite his success as an entrepreneur in Silicon Valley， he voluntarily left Palm and founded Numenta， a company focused on retro-engineering the brain to determine how it works — and applying that intelligence to machine learning。 With simple and direct language， Mr。 Hawkins explains the revolutionary theory of how our brain constructs reality， and then shows the implications of this knowledge for building truly intelligent， autonomous and even conscious machines。 He then reflects on the consequences for human evolution and his very existence。 In a century in which digital knowledge advances exponentially， it is fundamental to know ideas that generate foundation for future bets。 This is a bold and ahead of its time book， and thus likely to be very influential in the future。 Brilliant and highly recommended。 Obs。: as far as I know， there is still no translation into Portuguese。 #neuroscience #artificialintelligence #evolutionarybiology #machinelearning 。。。more
Sergio Alonso De Leon
，
“life is not about having a correct model of the world。 Life is about replication。” “Our reality is similar to the brain-in-a-vat hypothesis; we live in a simulated world， but it is not in a computer—it is in our head。” “We are intelligent not because we can do one thing particularly well， but because we can learn to do practically anything。” “through the sensory nerves。 The nerves only send spikes。 And since we do not perceive spikes， everything we do perceive must be fabricated in the brain。 “life is not about having a correct model of the world。 Life is about replication。” “Our reality is similar to the brain-in-a-vat hypothesis; we live in a simulated world， but it is not in a computer—it is in our head。” “We are intelligent not because we can do one thing particularly well， but because we can learn to do practically anything。” “through the sensory nerves。 The nerves only send spikes。 And since we do not perceive spikes， everything we do perceive must be fabricated in the brain。 Even the most basic feelings of light， sound， and touch are creations of the brain; they only exist in its model of the world。” “At some point in the future， we will accept that any system that learns a model of the world， continuously remembers the states of that model， and recalls the remembered states will be conscious。” 。。。more
Mario Russo
，
This was one of the most interesting books in a while。
Moin Uddin
，
There is a point where artificial intelligence meets neuroscience。 This is the domain that has always captivated people's curiosity。 Countless books on human psychology and neurology that are meant for general audience have made it to the best-selling bookshelf。 In the sense pop neurology has become a new reading fad。 Yet this book sets the bar way high。 In this book， the human brain is intended to be plotted like a route map or a neural network that compare the human brain to that of an integra There is a point where artificial intelligence meets neuroscience。 This is the domain that has always captivated people's curiosity。 Countless books on human psychology and neurology that are meant for general audience have made it to the best-selling bookshelf。 In the sense pop neurology has become a new reading fad。 Yet this book sets the bar way high。 In this book， the human brain is intended to be plotted like a route map or a neural network that compare the human brain to that of an integrated system。 The municipal water supply was used as an excellent model because its employees are naturally knowledgeable about how the system works。 Not everyone is an expert。 However， when taken together as a whole， all the persons can operate the water supply system。 A thousand brains are made out of bits and bytes of information。Recommended reading 。。。more
Kasper Johansen
，
Okay - but feel that it could have been half the size。
Cheng
，
第一部分介绍了作者数十年的大脑科学研究成果：千脑智能理论。主要包括重复的工作单元皮质柱，参考系理论，学习需要运动等概念。第二部分介绍未来的人工智能，对当下的问题认识很清晰，对未来并不如科幻或媒体那样悲观（吸引眼球），最后介绍人类智能的未来，作为有智能的高级物种，我们的未来在哪里？第一部分的内容太多的假设与试验证据的缺乏让我并无多少信心，不过机器智能现在的问题是很显然的，当下的人工智能并不会通向强人工智能，只是作者提出的方案我也同样没有信心。如果智能可以被批量复制，未来实在难以想象。但我还是期待这一天的到来。
Andrew
，
If you’ve ever seen Ratatouille， then you can perhaps remember that one scene where fireworks are exploding in the background， representing the explosion of flavors that Remi experiences as he mixes two different ingredients together。 Well， that’s a great way of representing my experience reading this book too。 However， while part 1 is great， the next two parts are severely lacking and I pretty much skipped over them。 They just seemed silly。 So the following review will cover part 1， but will le If you’ve ever seen Ratatouille， then you can perhaps remember that one scene where fireworks are exploding in the background， representing the explosion of flavors that Remi experiences as he mixes two different ingredients together。 Well， that’s a great way of representing my experience reading this book too。 However， while part 1 is great， the next two parts are severely lacking and I pretty much skipped over them。 They just seemed silly。 So the following review will cover part 1， but will leave the latter two untouched。In order to understand intelligence， then we need to understand what the neocortex does。 （Small point: Hawkins seems to make the mistake of identifying with his conceptual consciousness while dismissing his animate one; he praises the rider while ignoring the elephant， which is wrong， since we’re both rider and elephant， so he seems to be labouring under some sort of Cartesian spell。） Interestingly， the neocortex， which is the youngest part of the brain evolutionarily-speaking， contains complex circuitry that looks pretty much the same， regardless of whether it deals with touch， sight， or language。 He invokes the ghost of Vernon Mountcastle next， who argued that every part of the neocortex works on the same principles， meaning that all the things we think of as intelligence （seeing， touching， speaking， pondering） are fundamentally the same， as they implement the same cognitive algorithm albeit in different sensory modalities。 He said that the fundamental unit of the neocortex was a cortical column: it occupies one square millimeter on the neocortex， is 2。5 mm thick， and there are roughly 150，000 cortical columns stacked side by side across the whole neocortex。 Each column is fundamentally doing the same thing， regardless of whether it processes touch， sight， or language; but what？Making predictions。 Every cortical column learns a model of the world （from the most complex to the most simple） and bases its predictions on this model。 When the predictions are verified， the model is accurate; a misprediction causes anxiety and the desire to either attend to the error and update the model， or to explain the discrepancy away。 （Usually the latter because we be lazy bastards innit。） The inputs to the brain are constantly changing， both because we move and the world around us moves; so how does the neocortex， which is composed of thousands of nearly identical cortical columns， learn a predictive model of the world through movement？ Background Neuro-info: Neurons have thousands of synapses spaced along their dendrites， 90% of which are too far （distal） to cause the neuron to activate other neurons， but which activate the cell and turn it on for a brief period， before turning it back to normal again。 Why？ These dendrite spikes are predictions; the activation of distal dendrite synapses （enough to tease the neuron but not activate it） means that the neuron has recognised a pattern of activity in some other neurons （made a prediction） which， if it turns out to be the actual pattern experienced， suppresses the activation of non-primed neurons。 How does the neocortex then predict the next input when we move？ A place cell is a “you are here” marker on a map， and tells an animal where it is based on sensory input， but grid cells are aware of relative location。 Place cells and grid cells create maps for every environment。 In the old brain， these cells track the body; in the neocortex， these cells track thousands of locations simultaneously; as a sensory input arrives， its represented in the upper cortical column; this invokes the location in the lower layer; when the movement occurs， then the lower layer （location） changes to the expected new location， which causes a prediction of the next input in the upper layer。 If the original sensory input is ambiguous， then multiple locations get activated; if you move， multiple predictions occur in the upper layer （sensory input） before experience settles on what is actual。 To summarize: every cortical column learns models of objects using the same basic method that the old brain uses to learn models of environments。）Mountcastle’s suggestion that a common cortical algorithm underlies intelligence is fulfilled by Hawkin’s suggestion that all columns create reference frames for each observed object。 What is a reference frame？ An invisible， three-dimensional grid surrounding and attached to something; in more abstract terms， we can think of reference frames as a way to organize any kind of knowledge， not merely physical objects， but also conceptual schemes。 So there are two important points here: all knowledge is organized using reference frames and thinking is a form of moving。 If everything we know is stored in reference frames， then to recall stored knowledge， we have to activate the appropriate locations in the appropriate reference frames。 If this is correct， “then what we commonly call thinking is actually moving through a space， through a reference frame。” Your current thought is determined by the current location in the reference frame。 So the question becomes: what reference frame should people adopt vis-a-vis certain concepts？ （See Dutton’s work on superframes and Hofstadter’s analogical thinking。） A point I almost glossed over is the fact that there are four uses for reference frames， one in the old brain， three in the new。 Reference frames in the brain learn maps of environments。 RF in the what columns of the neocortex learn maps of physical objects; those in the where columns of the neocortex learn maps of the space around our body; those in the non-sensory columns of the neocortex learn maps of concepts。 （Wilber， McGilchrist。） Being an expert is mostly about finding a good reference frame to arrange facts and observations。 THE COMMON CORTICAL ALGORITHM IS BASED ON REFERENCE FRAMES！ They’re the substrate for learning the structure of the world， where things are， and how they move and change。THE THOUSAND BRAINS THEORY: Knowledge is distributed across the brain; take a coffee cup; where is knowledge about this coffee cup stored in the brain？ There are many cortical columns in the visual regions that receive input from the retina。 Each column that is seeing a part of the cup learns a model of the cup and tries to recognize it。 But there is no single model！ What you know about coffee cups exists in thousands of different models， in thousands of different cortical columns。 This is why it’s called the thousand brains theory: knowledge of any particular item is distributed among thousands of complementary models。 How do you then solve the binding problem？ （The problem of how diverse streams of sensory information converge to form a single perception。 （The question is the fruit of LH thinking; in reality， there is no problem。）） Cortical columns vote; your perception is the consensus the columns reach by voting。 If you move your finger along a surface to determine the shape， it will take a while; multiple fingers can tell almost immediately what something is。 How is voting accomplished？ Most connections in a column stay within the column; but there are exceptions; some cells in some layers extend very long distances all across the neocortex; the suggestion here is that they are doing the voting。 How？ The long range cells broadcast their votes of what they think they’re observing; often a column will be uncertain so will send multiple possibilities at the same time， while receiving projections from other columns representing their guesses。 The most common guesses suppress the least common ones until the entire network settles on one answer。 。。。more
Daiana
，
3。5* interesting theory but quite speculative and ignores several factors in relation to the evolution of AI。
José González
，
Not perfect， but really exciting。 Food for thought （a lot of）。
p。mac
，
Very good， but you get the main points early into the book making reading it further unproductive。
Ashwath Sundaresan
，
Loved this book。 Explores a new theory of the brain written in a simple and easy to understand manner。 Particularly loved what the implications of this is on machine intelligence and also fundamental human behaviors。
Prithu Puranjan
，
Loved the book ❤️1st part : working of brain。 Thousand brain theory。 Neocortex vs old brain。 Reference frames to create our models of world。 mountcastle's cortical columns in neocortex。 Synapses and our memory， consciousness。 Thinking being a form of movement 。。related with reference frame。Languages and mathematics with a recursive structure。。。。explained by reference frames。2nd part : AI。 How the current AI isnt intelligence。 Achieving intelligence through thousand brain theory。 Consciousness in Loved the book ❤️1st part : working of brain。 Thousand brain theory。 Neocortex vs old brain。 Reference frames to create our models of world。 mountcastle's cortical columns in neocortex。 Synapses and our memory， consciousness。 Thinking being a form of movement 。。related with reference frame。Languages and mathematics with a recursive structure。。。。explained by reference frames。2nd part : AI。 How the current AI isnt intelligence。 Achieving intelligence through thousand brain theory。 Consciousness in machine with pros cons。 Recipe for designing intelligent machine 。。。 embodiment， parts of the old brain， and the neocortex。Existential risks of machine intelligence 。 Intelligence explosion。 Goal misalignment。3rd part: Human intelligence and its ling time survival methods。 Existence and propogation of false belief seen through thousand brain theory。 Existential risk of human intelligence due to old brain。 Merging brains and machines。 Uploading brain or merging it with computer。 Estate planning for humanity， preserving knowledge in wiki satellites。Genes vs knowledge， reason to prioritize saving knowledge through self replicating autonomous intelligent robots。 。。。more
Tom Beck
，
Very interesting book， very thought provoking。 I enjoy the last chapters about legacy systems。 It’s scary to confirm things that I thought myself ， someone else put into words confirming my instincts。
Pablo
，
This book is packed with knowledge and interesting explanations about the brain。 I’ve really enjoyed part one and two。 At the end， the book took a turn in possibilities that are good to be aware off but less actionable in my opinion。Our view of the world is based on the way our brain interprets signals and what we experience。 How we can get influenced by external factors and believe on things they are not true。Understanding area of development and learning， ideas about the reference frames and h This book is packed with knowledge and interesting explanations about the brain。 I’ve really enjoyed part one and two。 At the end， the book took a turn in possibilities that are good to be aware off but less actionable in my opinion。Our view of the world is based on the way our brain interprets signals and what we experience。 How we can get influenced by external factors and believe on things they are not true。Understanding area of development and learning， ideas about the reference frames and how our mind would decide it’s amazing。A great book to be able to think about thinking and learning。 。。。more
Jan Havel
，
One of the best books I have read in years。 Provides insights in latest research on the brain and offers a revolutionary theory of how intelligence is formed at the level of cerebral columns using a general algorithms of reference frames originating in our ability to move around in a world perceived by senses。
AJ
，
Good short book， very readable。 Good brief intro to thousand brains theory。 Wish it was more technical， but he does provide extra reading at the end for that。
Esben Kranc
，