What do you think?
Rate this book
A Thousand Brains: A New Theory of Intelligence
An author, neuroscientist, and computer engineer unveils a theory of intelligence, of understanding 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.
288 pages, Hardcover
First published March 2, 2021
1,679 people are currently reading
25.2k people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 817 reviews
Read
January 24, 2022Few subjects have captured the imaginations of science fiction writers like artificial intelligence. If you’re interested in learning more about what it might take to create a true AI, this book offers a fascinating theory. Hawkins may be best known as the co-inventor of the PalmPilot, but he’s spent decades thinking about the connections between neuroscience and machine learning, and there’s no better introduction to his thinking than this book.
October 26, 2020
A Thousand Brains is a mixed bag. The first half describes Hawkins' theory of how the neocortex works, and there's a great deal there that is appealing. As a theory, it's relatively simple and self-consistent, and it fits the available data, at least in broad strokes. The theory operates at an intermediate level of abstraction: it's higher-level than the bottom-up feature analysis of David Marr, and it's lower-level than the top-down Society of Mind of Marvin Minsky. To borrow a phrase from Silicon Valley, it's a "Middle Out" approach.
I'm actually rather offended that Hawkins' never mentions Minsky. Hawkins practically steals the language of Minsky's Society of Mind (thousand brains, reference frames, etc.), but neglects to give credit, even in the annotated bibliography (which, let's admit it, is mainly an advertisement for Hawkins' company's papers, where presumably the actual details of this theory can be found). It seems like Hawkins either has a massive chip on his shoulder or an over-inflated ego.
This is all very frustrating, because Hawkins' hand-wavy model is probably correct, and the first half of the book is a good read (albeit, very light on technical details). Unfortunately, the second half of the book is a waste of time. Hawkins switches into prognostication mode and makes a bunch of fairly obvious arguments about the implications of machine intelligence. Anyone interested in these topics should read Yuval Noah Harari -- Homo Deus is a much richer (and better written) examination of the same topics.
I recommend the first half of the book to anyone who is interested in how to understand and reproduce human-level intelligence. But save some time and skip the second half.
I'm actually rather offended that Hawkins' never mentions Minsky. Hawkins practically steals the language of Minsky's Society of Mind (thousand brains, reference frames, etc.), but neglects to give credit, even in the annotated bibliography (which, let's admit it, is mainly an advertisement for Hawkins' company's papers, where presumably the actual details of this theory can be found). It seems like Hawkins either has a massive chip on his shoulder or an over-inflated ego.
This is all very frustrating, because Hawkins' hand-wavy model is probably correct, and the first half of the book is a good read (albeit, very light on technical details). Unfortunately, the second half of the book is a waste of time. Hawkins switches into prognostication mode and makes a bunch of fairly obvious arguments about the implications of machine intelligence. Anyone interested in these topics should read Yuval Noah Harari -- Homo Deus is a much richer (and better written) examination of the same topics.
I recommend the first half of the book to anyone who is interested in how to understand and reproduce human-level intelligence. But save some time and skip the second half.
Shelved as '1-tbr-owned-but-not-yet-read'
December 16, 2022Intelligence, 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.
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.
September 6, 2022
I like this analogy:
Imagine 50 people are invited to an evening party. Everyone arrives at the party at a randomly chosen time. When they get there, they open the door and step inside. What are the chances they see a party going on or an empty room?
It depends on how long they each stay, If all the partygoers stay for one minute before leaving, then almost everyone who shows up will see an empty room and conclude that no one else came to the party. If the partygoers stay an hour or two, then the party will be a success, with lots of people in the room at the same time.
I recommend reading this book, especially for someone who wants to know more about the brain and AI.
Simple to learn and understand.
Imagine 50 people are invited to an evening party. Everyone arrives at the party at a randomly chosen time. When they get there, they open the door and step inside. What are the chances they see a party going on or an empty room?
It depends on how long they each stay, If all the partygoers stay for one minute before leaving, then almost everyone who shows up will see an empty room and conclude that no one else came to the party. If the partygoers stay an hour or two, then the party will be a success, with lots of people in the room at the same time.
I recommend reading this book, especially for someone who wants to know more about the brain and AI.
Simple to learn and understand.
July 30, 2021
When Charles Darwin worked out the theory of evolution in the nineteenth century, he didn’t have all the details in which the theory would ultimately depend. After all, On the Origin of Species was published in 1859—a full 49 years before the term genetics was introduced and 94 years before the discovery of the double-helix structure of DNA.
But while Darwin didn’t know the details of genetic inheritance, he was able to deduce, through countless observations, that all plant and animal life—including all of its apparent diversity—is a manifestation of a single mechanism or algorithm: evolution by natural selection. He would leave the details of the mechanism to be filled in by future scientists.
Major scientific breakthroughs often follow a similar pattern. The details of planetary orbits, for example, are highly complex, but the discovery of heliocentrism provided a simplified framework in which future scientists could work out the complex mathematics.
In the contemporary world, our understanding of intelligence is in need of a similar breakthrough. While we know a lot of details about the workings of the brain—just as Darwin knew a lot of details about different forms of life before he developed the theory of evolution—we are in need of a single mechanism or algorithm that can explain the diverse functions of the human neocortex, the seat of human intelligence in the brain.
Jeff Hawkins, in his latest book A Thousand Brains, proposes just such a theory. The Thousand Brains Theory of Intelligence, as Hawkins calls it, has the potential to do for our understanding of intelligence what the theory of evolution did for our understanding of biology, and what heliocentrism did for our understanding of planetary orbits.
Let’s see what the theory entails, and why it might not only change our understanding of the brain but also fundamentally alter the approach of artificial intelligence (AI) research.
First, consider that the neocortex is, to the naked eye, an undifferentiated mass of tissue. We know that certain regions perform certain functions (vision, hearing, higher-order thinking), but this isn’t obvious by inspecting the organ—it simply all looks the same.
Zooming in to the level of the neuron doesn’t offer much help. All parts of the neocortex contain neurons arranged in what are called “cortical columns,” or vertical strips about the size of a strand of spaghetti. The neocortex is essentially composed of about 150,000 of these cortical columns that process information vertically and communicate across columns horizontally. These cortical columns all have the same component parts (with minor differences), but produce very different functions.
If you think about it, the problem of explaining intelligence appears to be analogous to the problem encountered by Darwin. Faced with the apparent diversity of life, Darwin proposed a single mechanism responsible for explaining all of the variety (natural selection). In the same way, is it possible that the diversity of intelligence can also be explained by a single mechanism? The homogeneity of nervous tissue and cortical columns seems to suggest that the answer is yes.
The first scientist to propose that this common mechanism must exist was the neurophysiologist Vernon Mountcastle in the late 1970s. As Hawkins writes:
“Mountcastle proposed that the reason the [neocortex] regions look similar is that they are all doing the same thing. What makes them different is not their intrinsic function but what they are connected to. If you connect a cortical region to eyes, you get vision; if you connect the same cortical region to ears, you get hearing; and if you connect regions to other regions, you get higher thought, such as language.”
All cortical regions, Mountcastle hypothesized, are doing the same thing; the problem is, Mountcastle didn’t know what that same thing was. What Hawkins is proposing is that not only was Mountcastle right, but the latest research is starting to piece together the specifics of this underlying algorithm of intelligence.
The key to understanding the diversity of intelligence, it turns out, is to think of the brain not as a single unit that processes information in an input-output manner similar to the way computers are currently built, but to think of the brain as being composed of thousands of smaller brains (cortical columns) that each develop independent predictive models of the world and communicate that information across columns in a way that creates a unified experience of perception.
Each one of your 150,000 cortical columns builds models of the world using what Hawkins calls reference frames. When you look out your window at passing cars, for example, the reason you see the cars as occupying a location that is at a distance from you—and not in your eyes where visual perception is actually taking place—is because your cortical columns contain both place cells (to decipher what it is you’re looking at) in addition to grid cells (to decipher the location of objects in the environment relative to your body). Your brain attaches similar reference frames to all objects in the world and even to abstract concepts like democracy and mathematics, according to Hawkins.
The act of thinking is the process of moving through reference frames, and the reason memory-enhancement techniques like the “memory palace” work so well is because knowledge is stored in the brain spatially. Using the memory palace technique, you would create an imaginary location, such as your house, and proceed to “store” important information in each room. As you move through the house in your imagnization, you are able to better recall the information than if you tried to simply memorize the list, and that is because the technique is taking advantage of the way your brain naturally stores information.
The Thousand Brains Theory of Intelligence, then, says that your brain is composed of 150,000 cortical columns that build models of the world using reference frames and that thought is a form of movement through information and ideas stored in relation to these spatial frames. Rather than simply processing information, your brain is creating thousands of independent models of the world, using all available senses, and integrating them together into a cohesive whole.
If Hawkins is right, this new way to think about intelligence will open the door to a new approach to the creation of artificial intelligence, which currently lacks true intelligence in the sense of the mental flexibility we see in humans. Future AI research, according to Hawkins, will need to replace neural networks with reference frames if we are to ever truly achieve general AI.
So is Hawkins right? As he admits, there is still much we do not know about the brain, and further research will be needed to confirm the Thousand Brains Theory. Nevertheless, Hawkins is confident that, while the details of the theory will inevitably be altered as new information is discovered, the general outline of the theory will hold up to scrutiny as the only explanation that can fully explain intelligence.
In the second part of the book, Hawkins tackles the future of AI and the prospect of machines attaining consciousness. Hawkins convincingly demonstrates that current AI is not really intelligent at all, and that the prospect of creating truly intelligent machines will depend on our ability to replicate cortical columns (which we can’t currently do). Hawkins also dismisses the exaggerated fears of AI as an existential threat, noting that motivations must be programmed into the machines and therefore there is no reason to assume inherent malevolent intent. Like Steven Pinker and others, Hawkins is confident that robust safety protocols—which are core components of the engineering profession—will prevent AI from getting out of control.
It is when Hawkins turns to the topic of consciousness that he is least persuasive; in fact, I don’t think it’s an exaggeration to say that the chapter on consciousness is embarrassingly simplistic and philosophically naive. He presents no counterarguments to his position, discusses none of the philosophical literature relevant to the topic, dismisses the hard problem of consciousness outright and without argument, and confidently proclaims “I have no doubts that machines that work on the same principles as the brain will be conscious.”
It’s interesting to witness Hawkins display the appropriate humility when discussing brain science—a topic he is thoroughly familiar with—yet when the topic turns to something he clearly knows very little about, consciousness, he becomes dogmatic and self-assured. This is a common phenomenon you see in contemporary scientists living in a reductionist age. If science can’t currently (or even conceivably) solve the problem, it’s better to just pretend it doesn’t exist.
Hawkins’ very own theory demonstrates the hard problem of consciousness. Remember, Hawkins is claiming that the brain is composed of thousands of cortical columns, which are all essentially doing the same thing. Each column is composed of neurons and electro-chemical activity. If this is the case, how is it that some columns create the perception of color—the actual experience of seeing red, for example—while some columns create the perception of sound, if every column is doing the same thing physically? Hawkins offers no explanations for how the same swirl of electricity in every cortical column could possibly produce different forms of conscious perception, or even how nerve impulses and chemicals traveling across synapses can result in actual three-dimensional experience. He wants the reader to simply pretend the problem doesn’t exist and to take his word for it that there’s nothing to explain. He even at one point shifts the burden of proof onto the skeptical reader. Hawkins writes:
“If you believe that consciousness cannot be explained by scientific investigation and the known laws of physics, then you might argue that I have shown that storing and recalling the states of a brain is necessary, but I have not proven that is sufficient. If you take this view, then the burden is on you to show why it is not sufficient.”
We currently have no satisfactory scientific explanation of consciousness. Further, we have not, to date, been able to create consciousness in machines. Therefore, if Hawkins is making the case that we can create consciousness in machines, you would think the burden of proof would be on him to explain how consciousness arises in the brain (winning him a Nobel prize in the process) and how exactly this could be replicated in machines. But he can’t, so he just assumes that artificial consciousness is possible and then tells his skeptics to prove him wrong. Of course, if it’s not possible to create consciousness in machines, this will be very difficult to prove, other than continuing to not be able to do it.
Hawkins’ shallow treatment of the philosophy is actually unfortunate, as it tarnishes what would otherwise have been a fascinating and thought-provoking book. It seems that Hawkins might be on the right track to explaining intelligence and providing the groundwork for more effective AI, but the reader should not mistake the Thousand Brains Theory for an all-encompassing theory of consciousness and the brain, because it is not.
But while Darwin didn’t know the details of genetic inheritance, he was able to deduce, through countless observations, that all plant and animal life—including all of its apparent diversity—is a manifestation of a single mechanism or algorithm: evolution by natural selection. He would leave the details of the mechanism to be filled in by future scientists.
Major scientific breakthroughs often follow a similar pattern. The details of planetary orbits, for example, are highly complex, but the discovery of heliocentrism provided a simplified framework in which future scientists could work out the complex mathematics.
In the contemporary world, our understanding of intelligence is in need of a similar breakthrough. While we know a lot of details about the workings of the brain—just as Darwin knew a lot of details about different forms of life before he developed the theory of evolution—we are in need of a single mechanism or algorithm that can explain the diverse functions of the human neocortex, the seat of human intelligence in the brain.
Jeff Hawkins, in his latest book A Thousand Brains, proposes just such a theory. The Thousand Brains Theory of Intelligence, as Hawkins calls it, has the potential to do for our understanding of intelligence what the theory of evolution did for our understanding of biology, and what heliocentrism did for our understanding of planetary orbits.
Let’s see what the theory entails, and why it might not only change our understanding of the brain but also fundamentally alter the approach of artificial intelligence (AI) research.
First, consider that the neocortex is, to the naked eye, an undifferentiated mass of tissue. We know that certain regions perform certain functions (vision, hearing, higher-order thinking), but this isn’t obvious by inspecting the organ—it simply all looks the same.
Zooming in to the level of the neuron doesn’t offer much help. All parts of the neocortex contain neurons arranged in what are called “cortical columns,” or vertical strips about the size of a strand of spaghetti. The neocortex is essentially composed of about 150,000 of these cortical columns that process information vertically and communicate across columns horizontally. These cortical columns all have the same component parts (with minor differences), but produce very different functions.
If you think about it, the problem of explaining intelligence appears to be analogous to the problem encountered by Darwin. Faced with the apparent diversity of life, Darwin proposed a single mechanism responsible for explaining all of the variety (natural selection). In the same way, is it possible that the diversity of intelligence can also be explained by a single mechanism? The homogeneity of nervous tissue and cortical columns seems to suggest that the answer is yes.
The first scientist to propose that this common mechanism must exist was the neurophysiologist Vernon Mountcastle in the late 1970s. As Hawkins writes:
“Mountcastle proposed that the reason the [neocortex] regions look similar is that they are all doing the same thing. What makes them different is not their intrinsic function but what they are connected to. If you connect a cortical region to eyes, you get vision; if you connect the same cortical region to ears, you get hearing; and if you connect regions to other regions, you get higher thought, such as language.”
All cortical regions, Mountcastle hypothesized, are doing the same thing; the problem is, Mountcastle didn’t know what that same thing was. What Hawkins is proposing is that not only was Mountcastle right, but the latest research is starting to piece together the specifics of this underlying algorithm of intelligence.
The key to understanding the diversity of intelligence, it turns out, is to think of the brain not as a single unit that processes information in an input-output manner similar to the way computers are currently built, but to think of the brain as being composed of thousands of smaller brains (cortical columns) that each develop independent predictive models of the world and communicate that information across columns in a way that creates a unified experience of perception.
Each one of your 150,000 cortical columns builds models of the world using what Hawkins calls reference frames. When you look out your window at passing cars, for example, the reason you see the cars as occupying a location that is at a distance from you—and not in your eyes where visual perception is actually taking place—is because your cortical columns contain both place cells (to decipher what it is you’re looking at) in addition to grid cells (to decipher the location of objects in the environment relative to your body). Your brain attaches similar reference frames to all objects in the world and even to abstract concepts like democracy and mathematics, according to Hawkins.
The act of thinking is the process of moving through reference frames, and the reason memory-enhancement techniques like the “memory palace” work so well is because knowledge is stored in the brain spatially. Using the memory palace technique, you would create an imaginary location, such as your house, and proceed to “store” important information in each room. As you move through the house in your imagnization, you are able to better recall the information than if you tried to simply memorize the list, and that is because the technique is taking advantage of the way your brain naturally stores information.
The Thousand Brains Theory of Intelligence, then, says that your brain is composed of 150,000 cortical columns that build models of the world using reference frames and that thought is a form of movement through information and ideas stored in relation to these spatial frames. Rather than simply processing information, your brain is creating thousands of independent models of the world, using all available senses, and integrating them together into a cohesive whole.
If Hawkins is right, this new way to think about intelligence will open the door to a new approach to the creation of artificial intelligence, which currently lacks true intelligence in the sense of the mental flexibility we see in humans. Future AI research, according to Hawkins, will need to replace neural networks with reference frames if we are to ever truly achieve general AI.
So is Hawkins right? As he admits, there is still much we do not know about the brain, and further research will be needed to confirm the Thousand Brains Theory. Nevertheless, Hawkins is confident that, while the details of the theory will inevitably be altered as new information is discovered, the general outline of the theory will hold up to scrutiny as the only explanation that can fully explain intelligence.
In the second part of the book, Hawkins tackles the future of AI and the prospect of machines attaining consciousness. Hawkins convincingly demonstrates that current AI is not really intelligent at all, and that the prospect of creating truly intelligent machines will depend on our ability to replicate cortical columns (which we can’t currently do). Hawkins also dismisses the exaggerated fears of AI as an existential threat, noting that motivations must be programmed into the machines and therefore there is no reason to assume inherent malevolent intent. Like Steven Pinker and others, Hawkins is confident that robust safety protocols—which are core components of the engineering profession—will prevent AI from getting out of control.
It is when Hawkins turns to the topic of consciousness that he is least persuasive; in fact, I don’t think it’s an exaggeration to say that the chapter on consciousness is embarrassingly simplistic and philosophically naive. He presents no counterarguments to his position, discusses none of the philosophical literature relevant to the topic, dismisses the hard problem of consciousness outright and without argument, and confidently proclaims “I have no doubts that machines that work on the same principles as the brain will be conscious.”
It’s interesting to witness Hawkins display the appropriate humility when discussing brain science—a topic he is thoroughly familiar with—yet when the topic turns to something he clearly knows very little about, consciousness, he becomes dogmatic and self-assured. This is a common phenomenon you see in contemporary scientists living in a reductionist age. If science can’t currently (or even conceivably) solve the problem, it’s better to just pretend it doesn’t exist.
Hawkins’ very own theory demonstrates the hard problem of consciousness. Remember, Hawkins is claiming that the brain is composed of thousands of cortical columns, which are all essentially doing the same thing. Each column is composed of neurons and electro-chemical activity. If this is the case, how is it that some columns create the perception of color—the actual experience of seeing red, for example—while some columns create the perception of sound, if every column is doing the same thing physically? Hawkins offers no explanations for how the same swirl of electricity in every cortical column could possibly produce different forms of conscious perception, or even how nerve impulses and chemicals traveling across synapses can result in actual three-dimensional experience. He wants the reader to simply pretend the problem doesn’t exist and to take his word for it that there’s nothing to explain. He even at one point shifts the burden of proof onto the skeptical reader. Hawkins writes:
“If you believe that consciousness cannot be explained by scientific investigation and the known laws of physics, then you might argue that I have shown that storing and recalling the states of a brain is necessary, but I have not proven that is sufficient. If you take this view, then the burden is on you to show why it is not sufficient.”
We currently have no satisfactory scientific explanation of consciousness. Further, we have not, to date, been able to create consciousness in machines. Therefore, if Hawkins is making the case that we can create consciousness in machines, you would think the burden of proof would be on him to explain how consciousness arises in the brain (winning him a Nobel prize in the process) and how exactly this could be replicated in machines. But he can’t, so he just assumes that artificial consciousness is possible and then tells his skeptics to prove him wrong. Of course, if it’s not possible to create consciousness in machines, this will be very difficult to prove, other than continuing to not be able to do it.
Hawkins’ shallow treatment of the philosophy is actually unfortunate, as it tarnishes what would otherwise have been a fascinating and thought-provoking book. It seems that Hawkins might be on the right track to explaining intelligence and providing the groundwork for more effective AI, but the reader should not mistake the Thousand Brains Theory for an all-encompassing theory of consciousness and the brain, because it is not.
August 30, 2022
barely 3 ☆
A Thousand Brains: A New Theory of Intelligence wasn't what I had expected. In part, it was because I hadn't known that neuroscientists have not conclusively solved this fundamental puzzle. The current predominant hypothesis--the hierarchy of features--presumes that the brain has a linear, flowchart-like way of processing the data collected by one's senses.
The author Jeff Hawkins had spent some time in a graduate neuroscience program but he doesn't have a PhD. Instead, after he invented the Palm Pilot, a handheld computer, Hawkins has had the freedom to pursue his curiosity about intelligence. He does have a concrete objective -- by figuring out the basis or mechanics of intelligence, he would apply the lessons to perfecting artificial intelligence ("AI").
So A Thousand Brains was not a primer on different parts of the brain and their functions. There's no illustration of the brain as Hawkins is only concerned about the neocortex, which comprises 70 percent of the brain. The other 30 percent is what the author regards as our "old brain."
Hawkins had cofounded Numenta, a neuroscience research company, which focuses its efforts on the neocortex. In the late 1970s, neuroscientist Vernon Mountcastle's big idea was published in The Mindful Brain: Cortical Organization and the Group-Selective Theory of Higher Brain Function. The fundamental building block or unit of the neocortex is the cortical column, in which all of its cells respond to the same stimulus.
Hawkins regards the cortical column as a foundational part of his "Thousand Brains Theory." His "theory" however is only a hypothesis because it hasn't been validated by experiments from the greater scientific community. Hawkins asserted that his "theory" is based on logical deductions of existing research on mammals' brains and addresses one of the major weaknesses of the hierarchy of features hypothesis -- which is that it's static.
In the same respect that Mountcastle’s hypothesis was groundbreaking in the late 1970s, Hawkins intends that A Thousand Brains will also be revolutionary. In it, he proposes that each cortical column (he assumes that there are 150,000 cortical columns) is a sensory-modeling system. The modeling system is based on reference frames. The neocortex will reach a consensus among its thousands of little models so that a human will have only one perception at a time. This was in Part One.
Hawkins is really laying his professional reputation on the line. Because there are neuroscientists who contest that there are even cortical columns never mind what they supposedly do.
• https://www.ncbi.nlm.nih.gov/pmc/arti...
• https://www.ncbi.nlm.nih.gov/pmc/arti...
Part Two contains Hawkins' musings on the future of AI. Most researchers' goal is to create artifical general intelligence (AGI), a standard in which a machine can think like a human, ie. progress to abstractions. One topic which greatly concerns some influential scientists and technical experts today is whether the attainment of AGI will lead to the extinction of the human race (see https://en.wikipedia.org/wiki/AI_take...). Hawkins does not believe that AI will be a threat to humans.
Part Three includes further pondering from Hawkins on human intelligence, existential threats, and the melding of humans with computers to attain a type of immortality. Regarding the latter, Hawkins opined that a human brain with its nearly 100 billion neurons is too complex to form connections with the current technology.
Hawkins is a big picture, conceptual thinker. Given the state of neuroscience today, it will likely be decades before Hawkins' proposal will be regarded as a game-changer or be consigned to the rubbish pile. But I got the feeling that Hawkins just wanted to take a stance and make it known, which is quite brave or arrogant or both. I have never studied or worked in neuroscience or in AI so this book is all very hypothetical to me. But I am part of his target audience because he stated that this book is for the lay reader. I am not impressed but I didn't heartily dislike it which would be 2.5 stars for "okay" or average. But since GR would include that in the "did not like" category, I'm settling upon 3 stars.
How a brain made of simple cells creates intelligence is a profoundly interesting question, and it remains a mystery.
A Thousand Brains: A New Theory of Intelligence wasn't what I had expected. In part, it was because I hadn't known that neuroscientists have not conclusively solved this fundamental puzzle. The current predominant hypothesis--the hierarchy of features--presumes that the brain has a linear, flowchart-like way of processing the data collected by one's senses.
The author Jeff Hawkins had spent some time in a graduate neuroscience program but he doesn't have a PhD. Instead, after he invented the Palm Pilot, a handheld computer, Hawkins has had the freedom to pursue his curiosity about intelligence. He does have a concrete objective -- by figuring out the basis or mechanics of intelligence, he would apply the lessons to perfecting artificial intelligence ("AI").
So A Thousand Brains was not a primer on different parts of the brain and their functions. There's no illustration of the brain as Hawkins is only concerned about the neocortex, which comprises 70 percent of the brain. The other 30 percent is what the author regards as our "old brain."
... the older parts of the human brain control the basic functions of life. They create our emotions, our desires to survive and procreate, and our innate behaviors.
The new brain, the neocortex, is the organ of intelligence.
Hawkins had cofounded Numenta, a neuroscience research company, which focuses its efforts on the neocortex. In the late 1970s, neuroscientist Vernon Mountcastle's big idea was published in The Mindful Brain: Cortical Organization and the Group-Selective Theory of Higher Brain Function. The fundamental building block or unit of the neocortex is the cortical column, in which all of its cells respond to the same stimulus.
Hawkins regards the cortical column as a foundational part of his "Thousand Brains Theory." His "theory" however is only a hypothesis because it hasn't been validated by experiments from the greater scientific community. Hawkins asserted that his "theory" is based on logical deductions of existing research on mammals' brains and addresses one of the major weaknesses of the hierarchy of features hypothesis -- which is that it's static.
In the same respect that Mountcastle’s hypothesis was groundbreaking in the late 1970s, Hawkins intends that A Thousand Brains will also be revolutionary. In it, he proposes that each cortical column (he assumes that there are 150,000 cortical columns) is a sensory-modeling system. The modeling system is based on reference frames. The neocortex will reach a consensus among its thousands of little models so that a human will have only one perception at a time. This was in Part One.
Hawkins is really laying his professional reputation on the line. Because there are neuroscientists who contest that there are even cortical columns never mind what they supposedly do.
• https://www.ncbi.nlm.nih.gov/pmc/arti...
• https://www.ncbi.nlm.nih.gov/pmc/arti...
Part Two contains Hawkins' musings on the future of AI. Most researchers' goal is to create artifical general intelligence (AGI), a standard in which a machine can think like a human, ie. progress to abstractions. One topic which greatly concerns some influential scientists and technical experts today is whether the attainment of AGI will lead to the extinction of the human race (see https://en.wikipedia.org/wiki/AI_take...). Hawkins does not believe that AI will be a threat to humans.
Part Three includes further pondering from Hawkins on human intelligence, existential threats, and the melding of humans with computers to attain a type of immortality. Regarding the latter, Hawkins opined that a human brain with its nearly 100 billion neurons is too complex to form connections with the current technology.
Hawkins is a big picture, conceptual thinker. Given the state of neuroscience today, it will likely be decades before Hawkins' proposal will be regarded as a game-changer or be consigned to the rubbish pile. But I got the feeling that Hawkins just wanted to take a stance and make it known, which is quite brave or arrogant or both. I have never studied or worked in neuroscience or in AI so this book is all very hypothetical to me. But I am part of his target audience because he stated that this book is for the lay reader. I am not impressed but I didn't heartily dislike it which would be 2.5 stars for "okay" or average. But since GR would include that in the "did not like" category, I'm settling upon 3 stars.
August 12, 2021
I have mixed feelings about this book. Basically, I think that the broad stroke explanation for how the cortex works is a great summary and conveys the core ideas of reference frames, hierarchies, prediction, and action-based learning quickly and at an easy level of understanding. As a neuroscientist, I was craving a little more depth for some of the ideas. For instance, Hawkins proposes a mechanism of various models of the world, instantiated in cortical columns, as "voting" on what will become our singular bound perception. I would have liked more details about how this voting happens. There are lots of neuroscientists with proposals, and I think that mechanisms through oscillatory coherence to build consensus are favored right now. In the same way, how reference frames become connected to the body or to objects is a bit hand-wavy and could use so more depth. And goal-representation, which in my mind is a core feature of intelligence, is barely touched. Still, those are minor criticisms and this would be a very useful primer on some of the core principles that we think are important for how the brain is organized.
I have separate more major criticisms for the first and second parts of the book. In the first part of the book, Hawkins discusses theory. Jeff Hawkins is very smart. He thinks and writes clearly. He has original ideas, and he is very creative. But sometimes, I feel like he has his own breakthroughs or epiphanies that make him super excited but then fails to recognize that others have had the same ideas before him. This tendency sticks out in a book that tries to reduce technical implementations of theory into general, high-level principles. He emphasizes his own thinking and "aha" moments, but the result is that it sounds like he is taking credit for older ideas. Virtually all of the big ideas presented in the book are older than Jeff Hawkins' work: the idea of reference frames in the cortex, the idea of the cortex checking predictions, the idea that cortical paths that are object-oriented or location-oriented are based on different inputs, the idea that the cortex is flexible and that columnar units across the cortex do similar things, etc. Even some of his more sci-fi ideas in the second part of the book are not new. For instance, I just encountered his idea of communicating a code using the Sun's light passing through man-made clouds first in Cixin Liu's Remembrance of Earth's Past series.
Maybe he didn't know that other people thought of these ideas before. He professes to not like sci-fi. But in a few of the parts of the book where he does give some credit, the timeline seems fuzzy, which gave me the distinct feeling that he was trying to blur the truth. For example, in 2016, Jeff Hawkins has an excellent idea while thinking about reference frames in the cortex. He proposes that cortical columns may have grid cells similar to the entorhinal cortex. In searching for experimental data to support this hypothesis, he discovers the paper by Christian Doeller, Caswell Barry, and Neil Burgess, which found cortical grid cells using fMRI. But in the book, he doesn't mention that this paper came out a lot earlier (in 2010), or that their paper was based on even earlier ideas (circa 2006-2008) by the Mosers that there are grid representations in the cortex that might underlie cognition generally, rather than just spatial navigation. Hawkins is right about grid cells and reference frames! I have no doubt that he came up with these ideas using his own brain. But in my opinion, it isn't right to convey the impression that Hawkins "came up" with the ideas. The papers that I just referenced aren't even hard to find. They are papers in Nature and PNAS that have had a massive impact on the field. If Jeff Hawkins invented calculus and wrote a book about it, you would say, "Wow! Jeff Hawkins is so smart! But it's sort of weird that he didn't write about Leibniz and Newton. In fact, it seems really weird that he didn't read more about Leibniz and Newton through this whole process." That's how I feel about the first part of this book.
My criticism for the second part of the book is more that Hawkins doesn't do a good job of conveying the ideas of his opposition about the points of AI rights or AI danger. He seems to be completely untroubled by how the cortex generates the experience of "red," but somehow at the same time thinks it is impossible that machines could suffer or experience emotions. It's not clear to me why we should not be worried. What if just the act of programming goals is sufficient for creating an emotion? In the same way, a lot of AI safety experts are concerned with AI getting out of hand as it pursues proximal goals, inherent to understanding that it is an agent (like survival, goal protection, resource allocation), and learns to deceive us. There are a lot of really good thinkers making really good arguments about these points and trying to ameliorate the dangers. I had the sense that Hawkins hasn't read them because he doesn't address them in any detail, but is still willing to be quite confident in their wrongness.
Overall, I would say that I like the book because it summarizes ideas well and is very thought-provoking. You will want to talk about the book with others. But I think he could have done a better job about contextualizing Numenta's contributions within the framework of a large scientific movement to understand the cortex, rather than presenting Numenta like a stand-alone maverick who figured everything out.
I have separate more major criticisms for the first and second parts of the book. In the first part of the book, Hawkins discusses theory. Jeff Hawkins is very smart. He thinks and writes clearly. He has original ideas, and he is very creative. But sometimes, I feel like he has his own breakthroughs or epiphanies that make him super excited but then fails to recognize that others have had the same ideas before him. This tendency sticks out in a book that tries to reduce technical implementations of theory into general, high-level principles. He emphasizes his own thinking and "aha" moments, but the result is that it sounds like he is taking credit for older ideas. Virtually all of the big ideas presented in the book are older than Jeff Hawkins' work: the idea of reference frames in the cortex, the idea of the cortex checking predictions, the idea that cortical paths that are object-oriented or location-oriented are based on different inputs, the idea that the cortex is flexible and that columnar units across the cortex do similar things, etc. Even some of his more sci-fi ideas in the second part of the book are not new. For instance, I just encountered his idea of communicating a code using the Sun's light passing through man-made clouds first in Cixin Liu's Remembrance of Earth's Past series.
Maybe he didn't know that other people thought of these ideas before. He professes to not like sci-fi. But in a few of the parts of the book where he does give some credit, the timeline seems fuzzy, which gave me the distinct feeling that he was trying to blur the truth. For example, in 2016, Jeff Hawkins has an excellent idea while thinking about reference frames in the cortex. He proposes that cortical columns may have grid cells similar to the entorhinal cortex. In searching for experimental data to support this hypothesis, he discovers the paper by Christian Doeller, Caswell Barry, and Neil Burgess, which found cortical grid cells using fMRI. But in the book, he doesn't mention that this paper came out a lot earlier (in 2010), or that their paper was based on even earlier ideas (circa 2006-2008) by the Mosers that there are grid representations in the cortex that might underlie cognition generally, rather than just spatial navigation. Hawkins is right about grid cells and reference frames! I have no doubt that he came up with these ideas using his own brain. But in my opinion, it isn't right to convey the impression that Hawkins "came up" with the ideas. The papers that I just referenced aren't even hard to find. They are papers in Nature and PNAS that have had a massive impact on the field. If Jeff Hawkins invented calculus and wrote a book about it, you would say, "Wow! Jeff Hawkins is so smart! But it's sort of weird that he didn't write about Leibniz and Newton. In fact, it seems really weird that he didn't read more about Leibniz and Newton through this whole process." That's how I feel about the first part of this book.
My criticism for the second part of the book is more that Hawkins doesn't do a good job of conveying the ideas of his opposition about the points of AI rights or AI danger. He seems to be completely untroubled by how the cortex generates the experience of "red," but somehow at the same time thinks it is impossible that machines could suffer or experience emotions. It's not clear to me why we should not be worried. What if just the act of programming goals is sufficient for creating an emotion? In the same way, a lot of AI safety experts are concerned with AI getting out of hand as it pursues proximal goals, inherent to understanding that it is an agent (like survival, goal protection, resource allocation), and learns to deceive us. There are a lot of really good thinkers making really good arguments about these points and trying to ameliorate the dangers. I had the sense that Hawkins hasn't read them because he doesn't address them in any detail, but is still willing to be quite confident in their wrongness.
Overall, I would say that I like the book because it summarizes ideas well and is very thought-provoking. You will want to talk about the book with others. But I think he could have done a better job about contextualizing Numenta's contributions within the framework of a large scientific movement to understand the cortex, rather than presenting Numenta like a stand-alone maverick who figured everything out.
April 29, 2021
Started as a fine popular science piece on how the brain models the objects and ideas by turning the electrical signals into "reference frames".
But ended as a meagre "populist science" attempt to explain wars, social injustice and climate change as the undesired results of our evolutionary "old" brain, which is the part that evolved before the neocortex and drives our basic evolutionary instincts.
I don't know what to say. Didn't Hawkins have any friends with at least a sociology degree who would say "Erm sir, this goes deeper than you may think". After decades of discussions within the social sciences about the social classes, structures, language and culture vs nature debates, you'd expect that at least some of them would become part of the general knowledge of the physical sciences community. But it did not.
Fact 1
Your engineering or science degree and your decades of experience in neuroscience/artificial intelligence/computer programming may create the illusion that you have cracked open the mysteries of the social experience. You are probably wrong.
Fact 2
There's still a huge knowledge gap between the social scientists and engineers, computer scientists, neurologists, neuro-psychologists in the form of lack of a general idea about each other's main achievements and findings.
Here's my hypothesis: While a social science PhD is less likely to foster any illusions in its owner about solving the problem of, say, "The Use of GAN networks in Natural Language Processing" and poke her to write a book about the subject, a PhD in computer science, on the contrary, is more likely to boost the self-confidence of its owner to the extent that he can write a book about the relationship between our "old" brain and the instability it brings to our societies and ignore the mountain of social science literature on the subject.
But ended as a meagre "populist science" attempt to explain wars, social injustice and climate change as the undesired results of our evolutionary "old" brain, which is the part that evolved before the neocortex and drives our basic evolutionary instincts.
I don't know what to say. Didn't Hawkins have any friends with at least a sociology degree who would say "Erm sir, this goes deeper than you may think". After decades of discussions within the social sciences about the social classes, structures, language and culture vs nature debates, you'd expect that at least some of them would become part of the general knowledge of the physical sciences community. But it did not.
Fact 1
Your engineering or science degree and your decades of experience in neuroscience/artificial intelligence/computer programming may create the illusion that you have cracked open the mysteries of the social experience. You are probably wrong.
Fact 2
There's still a huge knowledge gap between the social scientists and engineers, computer scientists, neurologists, neuro-psychologists in the form of lack of a general idea about each other's main achievements and findings.
Here's my hypothesis: While a social science PhD is less likely to foster any illusions in its owner about solving the problem of, say, "The Use of GAN networks in Natural Language Processing" and poke her to write a book about the subject, a PhD in computer science, on the contrary, is more likely to boost the self-confidence of its owner to the extent that he can write a book about the relationship between our "old" brain and the instability it brings to our societies and ignore the mountain of social science literature on the subject.
April 14, 2021
I felt a kinship with Hawkins when he talked about finding out as a younger man that academic labs study very specific parts of the brain (or specific diseases), and not the workings of the brain as a whole. Unlike him, I did not absorb that information, invent the Palm Pilot, and then found my own whole-brain research lab. My journey involved taking a philosophy of the mind course, learning about the "hard problem", and giving up on the brain altogether. But now I'm thankfully out of the dark ages and fully enjoying a book-driven brain renaissance.
I have to rate this five stars, even despite the manifesto in the second half of the book.
This guy might have figured out how the neocortex works. I'm talking about the process of how the brain takes in sensory information, processes sensory information, and then uses that information to make conclusions and predictions. It's called the Thousand Brain Theory of Intelligence, and it has to do with maps, "movement", and voting.
I was thinking this theory could work together with Graziano's theory to explain consciousness (I'm currently reading Graziano's book), and I was surprised to see Hawkins actually mention this.
The second half of the book is a lot. Maybe too much. There's a large focus on investing in preserving our knowledge (and putting it in orbit around the sun) just because future extraterrestrials may think it's cool. I agree with him, but I don't think it's a priority for most people, and I'm not even sure that it ought to be. He talks about our old brain and how a lot of the behavior that threatens our species originates from it. That makes sense.
The old brain helps genes replicate, but doesn't necessarily think in the long term. He proposes that we should strive to preserve knowledge and not genes. I think it would be hard to get people on board with this in any capacity. When it comes to preserving the human race, making it a goal to prevent future death and human suffering would probably be much more effective in propelling a response than focusing on the safeguarding of knowledge. People have more sympathy for concrete things like human beings than for more abstract things like potential aliens and knowledge. This sympathy can actually help us avoid extinction. In most cases, it's easier to work with human nature to achieve a goal than against it. I'm not sure if it would be wise to invest large amounts of resources into preserving our knowledge beyond the extinction of our species...especially since the aliens who could find that knowledge would probably already possess it, and since humans don't really have much emotional or general motivation to do that. Although I could be wrong, and I don't want to be limiting to anyone.
There are portions in which Hawkins talks about the next generation of AI, which is definitely something tech companies that want to stay relevant should look into. He doesn't think AI (intelligence itself) is an existential risk, and his argument is worth reading. What he says about uploading your brain is also very interesting and something I haven't considered at all. He mentions that he wants to stimulate discussion with this section of the book, and I do think it's an effective discussion stimulus.
The book is very logically and systematically laid out, with numbered sections and arguments that support specific conclusions. People will be annoyed by the author's certitude. His confidence is based largely on a couple of experiments and the fact that his theory doesn't conflict with a lot of research and present theories. I'd be a little more cautious if it were my book, but I found his main theory and even his speculations fascinating. Definitely worth a read, even if you stop after the first half.
I have to rate this five stars, even despite the manifesto in the second half of the book.
This guy might have figured out how the neocortex works. I'm talking about the process of how the brain takes in sensory information, processes sensory information, and then uses that information to make conclusions and predictions. It's called the Thousand Brain Theory of Intelligence, and it has to do with maps, "movement", and voting.
I was thinking this theory could work together with Graziano's theory to explain consciousness (I'm currently reading Graziano's book), and I was surprised to see Hawkins actually mention this.
The second half of the book is a lot. Maybe too much. There's a large focus on investing in preserving our knowledge (and putting it in orbit around the sun) just because future extraterrestrials may think it's cool. I agree with him, but I don't think it's a priority for most people, and I'm not even sure that it ought to be. He talks about our old brain and how a lot of the behavior that threatens our species originates from it. That makes sense.
The old brain helps genes replicate, but doesn't necessarily think in the long term. He proposes that we should strive to preserve knowledge and not genes. I think it would be hard to get people on board with this in any capacity. When it comes to preserving the human race, making it a goal to prevent future death and human suffering would probably be much more effective in propelling a response than focusing on the safeguarding of knowledge. People have more sympathy for concrete things like human beings than for more abstract things like potential aliens and knowledge. This sympathy can actually help us avoid extinction. In most cases, it's easier to work with human nature to achieve a goal than against it. I'm not sure if it would be wise to invest large amounts of resources into preserving our knowledge beyond the extinction of our species...especially since the aliens who could find that knowledge would probably already possess it, and since humans don't really have much emotional or general motivation to do that. Although I could be wrong, and I don't want to be limiting to anyone.
There are portions in which Hawkins talks about the next generation of AI, which is definitely something tech companies that want to stay relevant should look into. He doesn't think AI (intelligence itself) is an existential risk, and his argument is worth reading. What he says about uploading your brain is also very interesting and something I haven't considered at all. He mentions that he wants to stimulate discussion with this section of the book, and I do think it's an effective discussion stimulus.
The book is very logically and systematically laid out, with numbered sections and arguments that support specific conclusions. People will be annoyed by the author's certitude. His confidence is based largely on a couple of experiments and the fact that his theory doesn't conflict with a lot of research and present theories. I'd be a little more cautious if it were my book, but I found his main theory and even his speculations fascinating. Definitely worth a read, even if you stop after the first half.
July 21, 2024
This book begins by reviewing the parts of the brain and its evolution over time. The older parts of the human brain were developed to maintain life such as breathing, heart beating, fight or flight emotions, etc. The neocortex evolved later to provide the ability to perceive surroundings and the location of self in relation to those surroundings. As evolution developed the neocortex it simply repeatedly made duplicate copies of the same small unit. This unit is called a cortical column and the human neocortex contains about 150,000 of them.
The thousand brain theory postulates that each of these columns are all doing the same thing and their function is determined by the sensory organ to which they are connected. They are functioning as thousands of smaller brains each developing independent predictive models of the world and communicating that information across the other columns in a way that creates a unified experience of perception. This enables the formation of frames of reference to develop virtual models of our surroundings.
The neocortex originally developed for the purpose of determining what was around us and where we are located. But humans have used those same cognitive skills to also develop abstract concepts such as democracy and mathematics.
In the second half of the book the author proceeds to discuss such topics as intelligence and consciousness. It’s interesting to note that he doesn’t think the artificial intelligence currently available is capable of thinking (i.e. not much intelligence). Instead it is simply producing probable word order based of massive accumulations of data. True intelligence would be able to form frames of reference and adapt those models to abstract concepts in a manner similar to that of the human mind.
The thousand brain theory postulates that each of these columns are all doing the same thing and their function is determined by the sensory organ to which they are connected. They are functioning as thousands of smaller brains each developing independent predictive models of the world and communicating that information across the other columns in a way that creates a unified experience of perception. This enables the formation of frames of reference to develop virtual models of our surroundings.
The neocortex originally developed for the purpose of determining what was around us and where we are located. But humans have used those same cognitive skills to also develop abstract concepts such as democracy and mathematics.
In the second half of the book the author proceeds to discuss such topics as intelligence and consciousness. It’s interesting to note that he doesn’t think the artificial intelligence currently available is capable of thinking (i.e. not much intelligence). Instead it is simply producing probable word order based of massive accumulations of data. True intelligence would be able to form frames of reference and adapt those models to abstract concepts in a manner similar to that of the human mind.
March 27, 2021
The author sounded pompous throughout the book when speaking about himself, his observations, his background, and his understanding and beliefs. IMO he makes many assumptions and presumes facts from current theory which is tenuous at best. The author spends very little time developing his teams theory of 1000 Brains and supporting it with actual evidence. The author also wonders through many subject areas from neuro-science to artificial intelligence to the policy discussion of mankind (as if anyone today can manage the progress of our species into the future) and into projects like populating and terraforming Mars to long voyages into stellar space with genomic modifications to survive in stasis. He describes at one point how we should create an archive to back-up our clearly unique (to him) ability to comprehend our universe (assuming we are 1st or only one currently) As if this hasn't been accomplished yet (because he assumes we would have found it by now). The number of assumptions the author makes is truly amazing; he reminds me of Elon Musk who has re-introduced ever wild-ass idea in history to wonderous applause of genius (only for us to realize there are very real reasons some things haven't been done yet).
August 12, 2022
‘A Thousand Brains: a New Theory of Intelligence’ by Jeff Hawkins is a good overview of a lot of the questions many of us are asking and some theories by Hawkins to answer those questions. The questions I am talking about are: why do people believe in fake news and religions despite a lot of facts that proves their beliefs are incorrect? Why do most people, whether they believe in global warming or not, environmental concerns or not, continue to fill up the world with trash and pollution, using up resources that cannot be replaced or replenished, and continue to knowingly poison ourselves and our own home nests? Almost everyone agrees a world with more and more humans living on it is bad because we are using up too quickly what resources are available for humans to live, yet we continue to have sex without much thought in the moment to having very likely created yet another human that will need to eat, defecate, and use up resources that we know are becoming scarce - and that many people who are suddenly parents have little to none resources themselves right now to raise and educate a child properly.
Hawkin’s answers or educated guesses based on how he understands the brain to function on these questions are explored in Part Three in the book.
Part Two covers most of the questions us ordinary folk have about computer/machine intelligence, like if it is possible to create a self-aware computer, what consequences of that can be expected? Should we fear self-aware computers? Why or why not? He posits some ideas of creating smarter computers (unaware), basing their intelligence on the way the human brain is structured - neurons and electrical “spikes”. If this interests you and you want a more in-depth exploration of making and teaching a computer intelligence, I recommend The Alignment Problem: Machine Learning and Human Values.
Part One is the only part of the book to really discuss what the title infers. Does that mean the book is misleading if one is picking it up based on the title? Yes, imho.
Does that mean no one should read the book because it is being misleadingly marketed in its title? No, imho. It is an interesting brain/machine/cultural intelligence book written for the layman, or so-called little people like the proverbial blue-collar ‘mom and pop’, or the average young adult who is curious about these, probably newish to the reader, subjects.
Frankly, I suspect the book got away from the author.
Author’s biography from Wikipedia https://en.wikipedia.org/wiki/Jeff_Ha...
”Jeffrey Hawkins is a founder of Palm Computing and Handspring where he was one of the creators of the PalmPilot and Treo, respectively. He has since turned to work on neuroscience, founding the Redwood Center for Theoretical Neuroscience (formerly the Redwood Neuroscience Institute) in 2002 and Numenta in 2005, where he leads a team in efforts to reverse-engineer the neocortex and enable machine intelligence technology based on brain theory. Hawkins is the author of On Intelligence which explains his memory-prediction framework theory of the brain. In March 2021, he released his second book, A Thousand Brains: A New Theory of Intelligence, which details the discoveries he and the Numenta team made that led to the Thousand Brains Theory of Intelligence.
In 2003, Hawkins was elected as a member of the National Academy of Engineering "for the creation of the hand-held computing paradigm and the creation of the first commercially successful example of a hand-held computing device." He also serves on the Advisory Board of the Secular Coalition for America where he has advised on the acceptance and inclusion of nontheism in American life.
He attended Cornell University, where he received a bachelor's degree in electrical engineering in 1979.”
I copied the book blurb below, which is misleading in the certainty of fact and proofs that Hawkins supposedly has in his theories, which he does not, at least as he actually says in his book:
”An author, neuroscientist, and computer engineer unveils a theory of intelligence, of understanding 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.”
Part One is the section that explores the way the brain functions at the neuron level - what we do truly know about it - and also some theories of some brains’ scientists, especially by Vernon Mountcastle, a very respected scientist https://en.wikipedia.org/wiki/Vernon_.... Hawkins has been moving forward and developing hypothesis and theories based on Mountcastle’s work.
Hawkins primarily discusses the reason he believes in the existence of cortical columns, which create and hold models of the ‘real-world’ exterior of our brains (you know our bodies and brains filter reality, right?), and how they probably work with the sensory information of our eyes, nose, ears, skin, etc., and how columns work with the reasoning bits like remembering memories, thinking, predictions of what is going to happen next (voting - majority rules!). He explains how electrical signals transmit sensory information to our neurons, at least what we know about it which isn’t much, and how our neurons use these inputs, which again we don’t really know about in stone-cold fact. He discusses the “old brain” - the parts of our brain responsible for emotions, instincts, and automatic responses. These parts very likely evolved first. Then he discusses in general terms the neocortex, where our intelligence, memories, and reasoning abilities are, most likely. He explains how the neocortex works with the “old brain”, and explains how often it is the “old brain” bits that are very likely responsible for Humankind’s stupidities and ignorance. Stupidities often trump all over reason. Pun intentional.
For the laymen, the first part will be slightly difficult, and maybe boring, and of course, speculative (Hawkins doesn’t think it is THAT speculative, though, just at early stages). But part two will probably be more interesting to many; however, I think it is speculative. Part three will probably be interesting to everyone, even if only to argue over with an evangelical or Trump supporter or a conservative. One thing I definitely agree with Hawkins is that science is better for reasoning and decisions than beliefs are.
There are Suggested Readings, Acknowledgments and Discover More sections in the back of the book.
Hawkin’s answers or educated guesses based on how he understands the brain to function on these questions are explored in Part Three in the book.
Part Two covers most of the questions us ordinary folk have about computer/machine intelligence, like if it is possible to create a self-aware computer, what consequences of that can be expected? Should we fear self-aware computers? Why or why not? He posits some ideas of creating smarter computers (unaware), basing their intelligence on the way the human brain is structured - neurons and electrical “spikes”. If this interests you and you want a more in-depth exploration of making and teaching a computer intelligence, I recommend The Alignment Problem: Machine Learning and Human Values.
Part One is the only part of the book to really discuss what the title infers. Does that mean the book is misleading if one is picking it up based on the title? Yes, imho.
Does that mean no one should read the book because it is being misleadingly marketed in its title? No, imho. It is an interesting brain/machine/cultural intelligence book written for the layman, or so-called little people like the proverbial blue-collar ‘mom and pop’, or the average young adult who is curious about these, probably newish to the reader, subjects.
Frankly, I suspect the book got away from the author.
Author’s biography from Wikipedia https://en.wikipedia.org/wiki/Jeff_Ha...
”Jeffrey Hawkins is a founder of Palm Computing and Handspring where he was one of the creators of the PalmPilot and Treo, respectively. He has since turned to work on neuroscience, founding the Redwood Center for Theoretical Neuroscience (formerly the Redwood Neuroscience Institute) in 2002 and Numenta in 2005, where he leads a team in efforts to reverse-engineer the neocortex and enable machine intelligence technology based on brain theory. Hawkins is the author of On Intelligence which explains his memory-prediction framework theory of the brain. In March 2021, he released his second book, A Thousand Brains: A New Theory of Intelligence, which details the discoveries he and the Numenta team made that led to the Thousand Brains Theory of Intelligence.
In 2003, Hawkins was elected as a member of the National Academy of Engineering "for the creation of the hand-held computing paradigm and the creation of the first commercially successful example of a hand-held computing device." He also serves on the Advisory Board of the Secular Coalition for America where he has advised on the acceptance and inclusion of nontheism in American life.
He attended Cornell University, where he received a bachelor's degree in electrical engineering in 1979.”
I copied the book blurb below, which is misleading in the certainty of fact and proofs that Hawkins supposedly has in his theories, which he does not, at least as he actually says in his book:
”An author, neuroscientist, and computer engineer unveils a theory of intelligence, of understanding 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.”
Part One is the section that explores the way the brain functions at the neuron level - what we do truly know about it - and also some theories of some brains’ scientists, especially by Vernon Mountcastle, a very respected scientist https://en.wikipedia.org/wiki/Vernon_.... Hawkins has been moving forward and developing hypothesis and theories based on Mountcastle’s work.
Hawkins primarily discusses the reason he believes in the existence of cortical columns, which create and hold models of the ‘real-world’ exterior of our brains (you know our bodies and brains filter reality, right?), and how they probably work with the sensory information of our eyes, nose, ears, skin, etc., and how columns work with the reasoning bits like remembering memories, thinking, predictions of what is going to happen next (voting - majority rules!). He explains how electrical signals transmit sensory information to our neurons, at least what we know about it which isn’t much, and how our neurons use these inputs, which again we don’t really know about in stone-cold fact. He discusses the “old brain” - the parts of our brain responsible for emotions, instincts, and automatic responses. These parts very likely evolved first. Then he discusses in general terms the neocortex, where our intelligence, memories, and reasoning abilities are, most likely. He explains how the neocortex works with the “old brain”, and explains how often it is the “old brain” bits that are very likely responsible for Humankind’s stupidities and ignorance. Stupidities often trump all over reason. Pun intentional.
For the laymen, the first part will be slightly difficult, and maybe boring, and of course, speculative (Hawkins doesn’t think it is THAT speculative, though, just at early stages). But part two will probably be more interesting to many; however, I think it is speculative. Part three will probably be interesting to everyone, even if only to argue over with an evangelical or Trump supporter or a conservative. One thing I definitely agree with Hawkins is that science is better for reasoning and decisions than beliefs are.
There are Suggested Readings, Acknowledgments and Discover More sections in the back of the book.
April 19, 2021
Big disappointment. Hawkins is a bit of a hero of mine, but this rambling pop-sci book does nothing to advance his reputation. While he and his team have made substantial contributions to theoretical neuroscience and the potential next wave of a more generalized AI, this book provides few details other than a steady stream of self-congratulatory remarks. He lauds Vernon Mountcastle, who pioneered cortical columns as the source of intelligent computation, but provides little description of his work. He devotes almost a whole paragraph to Geoffrey Hinton, who ignited the current advancements of Deep Learning through his work on back-propagation. Hawkins says that Hinton is now disillusioned with AIs success, and has recently introduced the 'capsule' as a mechanism to overcome the limitations of Deep Learning, which Hawkins somehow finds similar to his own research, which it is not. The last third of the book is a bunch of proclamations about communication with extraterrestrial life and the dangers of climate change. What was he thinking?
March 3, 2021
Jeff Hawkins has likely solved the "easy" problem of consciousness. That is, he has likely figured out the algorithm by which the mammalian neocortex operates - the root of intelligence. By some accounts, this also solves the "hard" problem of consciousness - what consciousness itself is.
Let me say this again: Jeff Hawkins has likely solved humanity's longest and greatest philosophical and scientific question. Thus, if you like books, and I imagine you do, this is arguably the most important book written since Darwin's "On the Origin of Species".
The Thousand Brains theory of intelligence - that the neocortex is composed of many thousands of columns of connected neural structures based on older grid and place cells, which create and link together models of the world, and also vote together to construct hierarchically more nested models - is brilliant. The number of empirical constraints it solves is incredibly satisfying. Let me cite just three examples.
1) What and Where Pathways. Without getting into details (Hawkins provides some of these in the book and points you to the appropriate papers for detailed background information), what/where pathways in the brain are explained if "reference frames" attach to objects in the what pathway, but attach to our bodies in the where pathway. The simplicity is breathtaking and powerful.
2) The Binding Problem. Hawkins has solved the perceptual binding problem. That is, when cortical columns agree on an object via consensus voting, they will naturally have different sensory inputs, thus they contribute diverse sensory models to a complete perception.
3) The neocortex is built atop non-intelligent behavioral primitives. If the cortex is so powerful, could it one day theoretically operate the human body without the older, non-intelligent parts of the brain? No. The Thousand Brains theory shows that the voting process underpinning the power and flexibility of our intelligence comes at the cost of being able to be fooled. This is a feature not a bug - there is no way to model your environment in realtime without this cost. The Thousand Brains theory thus makes sense of mysterious empirical studies such as "Size-contrast illusions deceive the eye but not the hand" by Aglioti et al.
I hope you're starting to see why this book is so important, and why the Thousand Brains theory is so powerful in its explanatory capacity. These outstanding puzzle pieces just seem to fall into place.
So much for what David Chalmers calls the easy problem of consciousness - how the brain operates.
As for Chalmers' hard problem - e.g. even if we know how the brain works, what even is "redness"? etc. - I will only note, Hawkins is a thorough physicalist when it comes to consciousness, such that this and his previous book insist on using the word "intelligence" when one might want to see the word "consciousness". Unlike his previous book, Hawkins confronts this notion and dedicates a chapter to it here, in which the hard problem is recognized. As a physicalist who acknowledges the hard problem myself, I nonetheless find Hawkins' commentary here unsatisfying. For example, is the qualia of green a cortical map of green experiences with dimensions corresponding to green surface orientations? I'm not convinced. However, as Hawkins notes, qualia indeed seem "out there". In other words, qualia have the quale of location, an aspect nicely satisfied by reference frames at the core of the Thousand Brains theory. A theory which, as testified to in the beginning of this review, is brilliant and satisfying in its own right.
Even if you consider the hard problem to remain standing after reading A Thousand Brains, it's hard to argue that Hawkins hasn't largely solved the easy problem. In the end, that may indeed be the harder problem to solve anyway.
Read this book. Your grandchildren will be learning it as brain gospel.
Let me say this again: Jeff Hawkins has likely solved humanity's longest and greatest philosophical and scientific question. Thus, if you like books, and I imagine you do, this is arguably the most important book written since Darwin's "On the Origin of Species".
The Thousand Brains theory of intelligence - that the neocortex is composed of many thousands of columns of connected neural structures based on older grid and place cells, which create and link together models of the world, and also vote together to construct hierarchically more nested models - is brilliant. The number of empirical constraints it solves is incredibly satisfying. Let me cite just three examples.
1) What and Where Pathways. Without getting into details (Hawkins provides some of these in the book and points you to the appropriate papers for detailed background information), what/where pathways in the brain are explained if "reference frames" attach to objects in the what pathway, but attach to our bodies in the where pathway. The simplicity is breathtaking and powerful.
2) The Binding Problem. Hawkins has solved the perceptual binding problem. That is, when cortical columns agree on an object via consensus voting, they will naturally have different sensory inputs, thus they contribute diverse sensory models to a complete perception.
3) The neocortex is built atop non-intelligent behavioral primitives. If the cortex is so powerful, could it one day theoretically operate the human body without the older, non-intelligent parts of the brain? No. The Thousand Brains theory shows that the voting process underpinning the power and flexibility of our intelligence comes at the cost of being able to be fooled. This is a feature not a bug - there is no way to model your environment in realtime without this cost. The Thousand Brains theory thus makes sense of mysterious empirical studies such as "Size-contrast illusions deceive the eye but not the hand" by Aglioti et al.
I hope you're starting to see why this book is so important, and why the Thousand Brains theory is so powerful in its explanatory capacity. These outstanding puzzle pieces just seem to fall into place.
So much for what David Chalmers calls the easy problem of consciousness - how the brain operates.
As for Chalmers' hard problem - e.g. even if we know how the brain works, what even is "redness"? etc. - I will only note, Hawkins is a thorough physicalist when it comes to consciousness, such that this and his previous book insist on using the word "intelligence" when one might want to see the word "consciousness". Unlike his previous book, Hawkins confronts this notion and dedicates a chapter to it here, in which the hard problem is recognized. As a physicalist who acknowledges the hard problem myself, I nonetheless find Hawkins' commentary here unsatisfying. For example, is the qualia of green a cortical map of green experiences with dimensions corresponding to green surface orientations? I'm not convinced. However, as Hawkins notes, qualia indeed seem "out there". In other words, qualia have the quale of location, an aspect nicely satisfied by reference frames at the core of the Thousand Brains theory. A theory which, as testified to in the beginning of this review, is brilliant and satisfying in its own right.
Even if you consider the hard problem to remain standing after reading A Thousand Brains, it's hard to argue that Hawkins hasn't largely solved the easy problem. In the end, that may indeed be the harder problem to solve anyway.
Read this book. Your grandchildren will be learning it as brain gospel.
August 24, 2022
Non-Fiction>Science, neuroscience and AI
The first chapter was a little slow for me but maybe I just wasn't in the right mood when I read it because the whole book quickly got more interesting and easier to read for me.
Hawkins is very clear about his outline and intentions...he pretty much states there will be 3 parts, what each part is, and what he's trying to tell us. Then at the end of every chapter and at the end of every part he ends with, "To summarize what we've learned so far...". He has definitely done a great job of making and extremely accessible book for public consumption. He states at one point that he won't go into to much detail about brain activity in the book and that the particulars of the current theories are better suited for academic journals. I am very much in the habit of noting every 'new word' in my reading and I only had one from the whole book: qualia (which is a technical term for how sensory inputs are received/perceived and he explains what they are not just using the term and moving on as if you should know it).
Part I was what I was most interested in when I picked up the book and finding out it was only 1/3 of the book almost made me give up on it immediately: pop sci-fi-come-to-life is not my cup of tea. But I did end up enjoying all of the parts and the author's thoughts on humanity, consciousness, and brain theory.
The entire premise/perspective of the book is the Thousand Brains Theory which the author explains in Part I but I will summarize as: the brain is not a flow chart where nerves fire 1>2>3>4 to get from visual input to an image in your head. The brain works with reference frames in columns of nerves--there are redundancies and safeguards in place to make sure we hold on to knowledge gained. It is complicated and isn't quite what we've thought for the last 70 years but new info has lead the author to believe our idea of how the brain works does need to be tweaked.
First I love how enthused he is about how the human (and other beings) brains work. He also feels very strongly that the info we DO know about how brains work should be common knowledge and all people should understand how their brains work just like they understand how to drive a car. I think that's a great way to think about it. We all have brains so we should pay attention to what that means and learn about it a little bit.
The second 2 parts address upcoming or currently possible technologies for programming machines to merge with our brains or to enhance our existence (or extend our existence) with technology. This isn't like the ramblings of a Trekkie...there is a lot of valid discussion and insight here. Just a little warning for potential readers--Hawkins definitely dismisses all religious views of the origin of humanity and the universe as bygone mythologies. He doesn't dwell on it but does address this as it to clarify the angle he's coming from to address the future of humanity. He alse treats climate change deniers with similar dismissal after he explains his stance on that. Just fyi for ya.
I highly recommend this even if you (like me) don't have much interest in tech or futuristic AI because it really is all tied to how we humans can use tech to sustain our race. It's pretty interesting.
One downside for me (downside might not be the right word bc without this perspective he probably wouldn't be bold enough to write a book) was that Hawkins seems to think that since he had an idea for handheld computers in the 90s before this was possible and he turned out to be right that he must be right about his brain theory. He never states the argument quite this clearly, but it is certainly a logical fallacy and he believes it wholeheartedly. Of course he has a lot of other reasons for his theory but it really seems that he's determined to make this analogy that since he was right and everyone was short-sighted before that that must be the way of the world and naysayers are just uninlightened and don't understand YET. Another drawback for me was the incessent example of a coffee cup. I mean, c'mon, he could have come up with so many other examples to draw on instead of always going back to a coffee cup (with his company logo of course)!
"Democracy in the brain? Consensus, and even dispute? What an amazing idea."
"Our ability to do this tells us that knowledge of concepts must also be stored in reference frames. But these reference frames may not be easily equated to the reference frames we use for coffee cups and other physical objects. For example, it is possible that the reference frames that are most useful for certain concepts have more than three dimensions."
"After surgery, these patients act as if they have two brains. Experiments clearly show that the two sides of the brain have different thoughts and reach different conclusions."
"The Thousand Brains Theory is a framework; it is like finishing the puzzle’s border and knowing what the overall picture looks like."
"We are intelligent not because we can do one thing particularly well, but because we can learn to do practically anything."
"If you touch a hot stove, your arm will retract in pain before your neocortex knows what is happening. Therefore, pain can’t be understood in the same way as the color green, which I am proposing is learned in the neocortex."
"Finding evidence that supports our beliefs is helpful, but not definitive. Finding contrary evidence, however, is proof that the model in our head is not right and needs to be modified. Actively looking for evidence to disprove our beliefs is the scientific method."
"It is human nature—aka old brain—to suspect everyone wants to steal your idea, where the reality is that you are lucky if anyone cares about your idea at all."
For Aug 2022 this was a buddy read with the Non-Fiction Book Club here on GR.
The first chapter was a little slow for me but maybe I just wasn't in the right mood when I read it because the whole book quickly got more interesting and easier to read for me.
Hawkins is very clear about his outline and intentions...he pretty much states there will be 3 parts, what each part is, and what he's trying to tell us. Then at the end of every chapter and at the end of every part he ends with, "To summarize what we've learned so far...". He has definitely done a great job of making and extremely accessible book for public consumption. He states at one point that he won't go into to much detail about brain activity in the book and that the particulars of the current theories are better suited for academic journals. I am very much in the habit of noting every 'new word' in my reading and I only had one from the whole book: qualia (which is a technical term for how sensory inputs are received/perceived and he explains what they are not just using the term and moving on as if you should know it).
Part I was what I was most interested in when I picked up the book and finding out it was only 1/3 of the book almost made me give up on it immediately: pop sci-fi-come-to-life is not my cup of tea. But I did end up enjoying all of the parts and the author's thoughts on humanity, consciousness, and brain theory.
The entire premise/perspective of the book is the Thousand Brains Theory which the author explains in Part I but I will summarize as: the brain is not a flow chart where nerves fire 1>2>3>4 to get from visual input to an image in your head. The brain works with reference frames in columns of nerves--there are redundancies and safeguards in place to make sure we hold on to knowledge gained. It is complicated and isn't quite what we've thought for the last 70 years but new info has lead the author to believe our idea of how the brain works does need to be tweaked.
First I love how enthused he is about how the human (and other beings) brains work. He also feels very strongly that the info we DO know about how brains work should be common knowledge and all people should understand how their brains work just like they understand how to drive a car. I think that's a great way to think about it. We all have brains so we should pay attention to what that means and learn about it a little bit.
The second 2 parts address upcoming or currently possible technologies for programming machines to merge with our brains or to enhance our existence (or extend our existence) with technology. This isn't like the ramblings of a Trekkie...there is a lot of valid discussion and insight here. Just a little warning for potential readers--Hawkins definitely dismisses all religious views of the origin of humanity and the universe as bygone mythologies. He doesn't dwell on it but does address this as it to clarify the angle he's coming from to address the future of humanity. He alse treats climate change deniers with similar dismissal after he explains his stance on that. Just fyi for ya.
I highly recommend this even if you (like me) don't have much interest in tech or futuristic AI because it really is all tied to how we humans can use tech to sustain our race. It's pretty interesting.
One downside for me (downside might not be the right word bc without this perspective he probably wouldn't be bold enough to write a book) was that Hawkins seems to think that since he had an idea for handheld computers in the 90s before this was possible and he turned out to be right that he must be right about his brain theory. He never states the argument quite this clearly, but it is certainly a logical fallacy and he believes it wholeheartedly. Of course he has a lot of other reasons for his theory but it really seems that he's determined to make this analogy that since he was right and everyone was short-sighted before that that must be the way of the world and naysayers are just uninlightened and don't understand YET. Another drawback for me was the incessent example of a coffee cup. I mean, c'mon, he could have come up with so many other examples to draw on instead of always going back to a coffee cup (with his company logo of course)!
"Democracy in the brain? Consensus, and even dispute? What an amazing idea."
"Our ability to do this tells us that knowledge of concepts must also be stored in reference frames. But these reference frames may not be easily equated to the reference frames we use for coffee cups and other physical objects. For example, it is possible that the reference frames that are most useful for certain concepts have more than three dimensions."
"After surgery, these patients act as if they have two brains. Experiments clearly show that the two sides of the brain have different thoughts and reach different conclusions."
"The Thousand Brains Theory is a framework; it is like finishing the puzzle’s border and knowing what the overall picture looks like."
"We are intelligent not because we can do one thing particularly well, but because we can learn to do practically anything."
"If you touch a hot stove, your arm will retract in pain before your neocortex knows what is happening. Therefore, pain can’t be understood in the same way as the color green, which I am proposing is learned in the neocortex."
"Finding evidence that supports our beliefs is helpful, but not definitive. Finding contrary evidence, however, is proof that the model in our head is not right and needs to be modified. Actively looking for evidence to disprove our beliefs is the scientific method."
"It is human nature—aka old brain—to suspect everyone wants to steal your idea, where the reality is that you are lucky if anyone cares about your idea at all."
For Aug 2022 this was a buddy read with the Non-Fiction Book Club here on GR.
November 30, 2023
RIP 12.1.23
“The list of things everyone should know is short. I would include how the brain is composed of the new part and the older parts. I would include how the neocortex learns a model of the world, whereas the older parts of the brain generate our emotions and more primitive behaviors. I would include how the old brain can take control, causing us to act in ways we know we shouldn’t.”
—Jeff Hawkins
“The list of things everyone should know is short. I would include how the brain is composed of the new part and the older parts. I would include how the neocortex learns a model of the world, whereas the older parts of the brain generate our emotions and more primitive behaviors. I would include how the old brain can take control, causing us to act in ways we know we shouldn’t.”
—Jeff Hawkins
This entire review has been hidden because of spoilers.
February 11, 2024
Teoria celor o mie de creieri vine să înlocuiască sau să completeze, încă nu-mi este clar, teoria ierarhiei proprietăților.
În teoria ierhiei proprietăților elementele sunt prelucrate TREPTAT. Spre exemplu, în cazul văzului, mai întâi creierul proiectează liniile și muchiile, apoi formele, pentru ca în final să creeze imaginea completă. Pe de altă parte, teoria celor o mie de creieri împarte creierul în 150 000 de coloane corticale, fiecare coloană fiind capabilă să creeze imaginea COMPLETĂ.
Prima parte este dificilă și trebuie citită cu atenție. Autorul recunoaște că dacă ar intra mai mult în detalii ar da în articole științifice.
Teoria celor o mie de creieri are la bază un studiu publicat în anii '70 de Vernon Mountcastle.
Recomand
În teoria ierhiei proprietăților elementele sunt prelucrate TREPTAT. Spre exemplu, în cazul văzului, mai întâi creierul proiectează liniile și muchiile, apoi formele, pentru ca în final să creeze imaginea completă. Pe de altă parte, teoria celor o mie de creieri împarte creierul în 150 000 de coloane corticale, fiecare coloană fiind capabilă să creeze imaginea COMPLETĂ.
Prima parte este dificilă și trebuie citită cu atenție. Autorul recunoaște că dacă ar intra mai mult în detalii ar da în articole științifice.
Teoria celor o mie de creieri are la bază un studiu publicat în anii '70 de Vernon Mountcastle.
Recomand
August 9, 2021
The book comes in two main parts, the first part describes Jeff Hawkins's theory of intelligence and the latter part discusses philosophical implications, popular retorts from the AI safety community, and some suggestions for how humanity might preserve intelligence going forward.
The main theory of the book is that the neocortex is arranged in cortical columns and that the purpose of these neocortical columns is to create & predict spatial reference frames of the world around us and how we can interact with those reference frames through the movement of our muscles. This seems straightforward and not that new, considering many neuroscientists in the past have concluded that the brain is for motion. For example, neuroscientists Daniel Wolpert and Gerald Edleman have both observed "neuromuscular" and "sensorimotor" are portmanteaus for a reason. One cannot exist without the other, as Edelman stated: “Brains are embodied and bodies are embedded in an environment.” Furthermore, 'the world' does not have to mean Earth, any abstract environment with relationships works, which Hawkins would argue encompasses math & linguistics. Hawkins cites some evidence that grid cells & place cells exist not just in the hippocampus but also exist in cortical columns. There is no slam-dunk evidence here that cortical columns exist for the sole purpose of creating reference frames, rather it's more of an appeal to Stafford Beer's notion of "the purpose of a system is what it does."
The latter part of the book is not as interesting, it seems more like filler content that could be left unread. For example, there is a discussion about how humans might signal to extraterrestrial life that we exist, but these issues are worthy of books on their own and it does not do any of the topics justice. It would have been better to cite the work of others to really start a discussion. There is also a very out-of-place 'new atheist' take that the notion of the afterlife is wrong prima facie and this human meme is a bug, not a feature. Spelled out, it reads: "I assume materialism, and haven't read anything else." Without me relitigating the umteen debates on science vs. religion, this struck me as a dumb can of worms to open in a book that purports to be about neuroscience. This alone was odd, but there is also a perplexing section of the book where Hawkins argues that increasing the human population is wrongheaded and that it is prudent to try and stop population growth. The discussion of this topic was disturbingly ignorant like the author had never heard the name "Malthus." Again, another landmine that sours the entire book.
Overall, from the second half of the book, I got the sense that this book was not a serious evaluation of the current state of the field, rather a browbeating of Hawkins's current theory of intelligence that doesn't seem all that novel. I can't solidly recommend this book, so in lieu, I would recommend listening to Sam Harris's podcast with Jeff which discusses the main highlights of the book (and incidentally also contains some browbeating but not as egregious as the latter half of this book).
The main theory of the book is that the neocortex is arranged in cortical columns and that the purpose of these neocortical columns is to create & predict spatial reference frames of the world around us and how we can interact with those reference frames through the movement of our muscles. This seems straightforward and not that new, considering many neuroscientists in the past have concluded that the brain is for motion. For example, neuroscientists Daniel Wolpert and Gerald Edleman have both observed "neuromuscular" and "sensorimotor" are portmanteaus for a reason. One cannot exist without the other, as Edelman stated: “Brains are embodied and bodies are embedded in an environment.” Furthermore, 'the world' does not have to mean Earth, any abstract environment with relationships works, which Hawkins would argue encompasses math & linguistics. Hawkins cites some evidence that grid cells & place cells exist not just in the hippocampus but also exist in cortical columns. There is no slam-dunk evidence here that cortical columns exist for the sole purpose of creating reference frames, rather it's more of an appeal to Stafford Beer's notion of "the purpose of a system is what it does."
The latter part of the book is not as interesting, it seems more like filler content that could be left unread. For example, there is a discussion about how humans might signal to extraterrestrial life that we exist, but these issues are worthy of books on their own and it does not do any of the topics justice. It would have been better to cite the work of others to really start a discussion. There is also a very out-of-place 'new atheist' take that the notion of the afterlife is wrong prima facie and this human meme is a bug, not a feature. Spelled out, it reads: "I assume materialism, and haven't read anything else." Without me relitigating the umteen debates on science vs. religion, this struck me as a dumb can of worms to open in a book that purports to be about neuroscience. This alone was odd, but there is also a perplexing section of the book where Hawkins argues that increasing the human population is wrongheaded and that it is prudent to try and stop population growth. The discussion of this topic was disturbingly ignorant like the author had never heard the name "Malthus." Again, another landmine that sours the entire book.
Overall, from the second half of the book, I got the sense that this book was not a serious evaluation of the current state of the field, rather a browbeating of Hawkins's current theory of intelligence that doesn't seem all that novel. I can't solidly recommend this book, so in lieu, I would recommend listening to Sam Harris's podcast with Jeff which discusses the main highlights of the book (and incidentally also contains some browbeating but not as egregious as the latter half of this book).
April 8, 2021
I have been involved with AI/Machine Learning since late 1970s. Even though my professional involvement stopped a while ago, I never stopped reading about AI, Cognitive Science, Neuroscience, Consciousness, and many other related fields. Over the years as my exposure to this area has increased, so did my list of open questions for which I could not get a satisfactory answer. Looking back, all these annoying gaps and open questions ultimately boil down to one fact – we still don’t have a broad theory of how our brain can produce a mind. There is a mountain of experimental data, and there are many successful theories of how each small piece work, but there is nothing yet that gives us a broad framework to fit all these pieces together.
The situation is similar to where biology was before the introduction of Darwin’s theory of evolution. We knew a lot about specific animals, their properties and their behavior, how different organs work etc. Yet, we did not understand the whole system of life, where each of these isolated facts can be tied together.
In this ambitious book that’s exactly what the author is trying to do. He created a basic framework that may explain the underlying architecture of any brain, especially mammalian and more specifically human brain. It is too early to say if this is the theory we have been waiting for. However, judging by how many of my accumulated list of questions it could potentially answer gives me tremendous hope that this could the basic seed. Since there is so much experimental data about the brain already available, it could be easier to confirm or reject this theory without too many new experiments.
This theory could not only help us better understand the architecture of the brain but could also allow us to build better AI. There are many areas of Artificial General Intelligence (AGI) where we have no solution path. This theory shines a tentative light towards such solutions.
Irrespective of whether the theory proves to be correct or not, it is a remarkably thought-provoking book, written in a lucid style that almost anyone should be able to understand. If you are interested in our brain, and I cannot imagine why anyone would not be, then this is a must read.
The final few chapters of the book deal with more speculative areas such as the future of humanity, ultimate preservation of knowledge, whether AI could be an existential threat, and many other similar thoughts. While each of these chapters are very interesting and thought-provoking on their own, I am not sure if they had a room in this book. It seemed like a distraction to me, except for the discussion around the question -- what is more important to preserve, our genes or our knowledge?
The situation is similar to where biology was before the introduction of Darwin’s theory of evolution. We knew a lot about specific animals, their properties and their behavior, how different organs work etc. Yet, we did not understand the whole system of life, where each of these isolated facts can be tied together.
In this ambitious book that’s exactly what the author is trying to do. He created a basic framework that may explain the underlying architecture of any brain, especially mammalian and more specifically human brain. It is too early to say if this is the theory we have been waiting for. However, judging by how many of my accumulated list of questions it could potentially answer gives me tremendous hope that this could the basic seed. Since there is so much experimental data about the brain already available, it could be easier to confirm or reject this theory without too many new experiments.
This theory could not only help us better understand the architecture of the brain but could also allow us to build better AI. There are many areas of Artificial General Intelligence (AGI) where we have no solution path. This theory shines a tentative light towards such solutions.
Irrespective of whether the theory proves to be correct or not, it is a remarkably thought-provoking book, written in a lucid style that almost anyone should be able to understand. If you are interested in our brain, and I cannot imagine why anyone would not be, then this is a must read.
The final few chapters of the book deal with more speculative areas such as the future of humanity, ultimate preservation of knowledge, whether AI could be an existential threat, and many other similar thoughts. While each of these chapters are very interesting and thought-provoking on their own, I am not sure if they had a room in this book. It seemed like a distraction to me, except for the discussion around the question -- what is more important to preserve, our genes or our knowledge?
May 1, 2021
I've decided to read this book mainly because I was looking for something on mapping modern artificial intelligence on the learnings of neuro-morphology. I can't say I've found what that, but the book was interesting enough.
There are three distinctive parts here.
1. the first one is about the anatomy of a human brain - it wasn't fascinating, a lot of information is already covered in many sources (e.g. the role of the neocortex, "old brain", etc.), but the final theory (about the reference frames and how the same model works for all the senses) is actually quite interesting; however, it's just a theory - in fact the author doesn't have much to back it up (despite the fact that it sounds reasonable ...)
2. the second one is about the artificial intelligence we face these days and how does it refer to human intelligence - this part is far less detailed than e.g. in "Army of None", but it's not a big deal - all that was important was covered quite nicely. What I was missing was some research about actual attempts of re-applying our knowledge about the human brain in building AI.
3. the third one felt most chaotic because, in fact, it has covered few very different considerations: cloning the human brain, looking for contact w/ extraterrestrial civilizations (srsly!), the role of genes in the development of (artificial) intelligence.
In the end, it was a solid book - a nice primer for someone who'd like to understand the nature of what we call 'artificial intelligence' these days. As long as you'll be able to survive the whole first part which is a bit lengthy :)
Solid 3.7 stars.
There are three distinctive parts here.
1. the first one is about the anatomy of a human brain - it wasn't fascinating, a lot of information is already covered in many sources (e.g. the role of the neocortex, "old brain", etc.), but the final theory (about the reference frames and how the same model works for all the senses) is actually quite interesting; however, it's just a theory - in fact the author doesn't have much to back it up (despite the fact that it sounds reasonable ...)
2. the second one is about the artificial intelligence we face these days and how does it refer to human intelligence - this part is far less detailed than e.g. in "Army of None", but it's not a big deal - all that was important was covered quite nicely. What I was missing was some research about actual attempts of re-applying our knowledge about the human brain in building AI.
3. the third one felt most chaotic because, in fact, it has covered few very different considerations: cloning the human brain, looking for contact w/ extraterrestrial civilizations (srsly!), the role of genes in the development of (artificial) intelligence.
In the end, it was a solid book - a nice primer for someone who'd like to understand the nature of what we call 'artificial intelligence' these days. As long as you'll be able to survive the whole first part which is a bit lengthy :)
Solid 3.7 stars.
January 16, 2022
Wanted to learn more about the introduced brain theory. After the first third, instead of going deeper, the author goes on a journey over a long list of other topics. they have, in my opinion, little to do with what I reasonably expected from the book.
April 7, 2021
Old ideas about neural networks + grating self-agrandizement + endless repetition + unrelated musings about culture + unrelated musings about aliens = new book about neural networks?
November 14, 2021
A Thousand Brains is made up of an inconsistent mishmash of ideas, varying in quality and profundity. The parts of the book in which Hawkins talks about neuroscience are insightful, but everything else is parochial with a hint of arrogance.
Hawkins offers a physiological grounding to theories that explain cognition as a process of modelling/inference. It does this by presenting well-reasoned evidence proposing that each of the ~150k cortical columns in our brain contains a model of reality that together form our overall experience.
I thought that this was also very insightful:
The first half is worth five stars, but the second half is worth one star, resulting in a three stars average.
Hawkins offers a physiological grounding to theories that explain cognition as a process of modelling/inference. It does this by presenting well-reasoned evidence proposing that each of the ~150k cortical columns in our brain contains a model of reality that together form our overall experience.
"The brain creates a predictive model. This just means that the brain continuously predicts what its inputs will be. Prediction isn’t something that the brain does every now and then; it is an intrinsic property that never stops, and it serves an essential role in learning. When the brain’s predictions are verified, that means the brain’s model of the world is accurate. A misprediction causes you to attend to the error and update the model."
I thought that this was also very insightful:
"How can I speak confidently about a theory if it hasn’t been tested experimentally? I just described one of these situations. I had an insight that the neocortex is infused with reference frames, and I immediately started talking about it with certainty. [...] Here is why.
As we work on a problem, we uncover what I call constraints. Constraints are things that the solution to the problem must address. [...] The longer you work on a problem, the more constraints you discover and the harder it becomes to imagine a solution. The aha moments I described in this chapter were about problems that we worked on for years. Therefore, we understood these problems deeply and our list of constraints was long. The likelihood that a solution is correct increases exponentially with the number of constraints it satisfies. It is like solving a crossword puzzle: There are often several words that match an individual clue. If you pick one of those words, it could be wrong. If you find two intersecting words that work, then it is much more likely they are both correct. If you find ten intersecting words, the chance that they are all wrong is minuscule."
The first half is worth five stars, but the second half is worth one star, resulting in a three stars average.
September 29, 2024
This book proposes a new theory of the brain – how it learns and stores knowledge. He starts with the basics, walking readers through the operations of the neocortex (the “new” part of the brain associated with intelligence as opposed to the “old” brain associated with primal functions and emotions. We learn about the composition of the neocortex, which he asserts is comprised of “cortical columns” and operates by creating “reference frames” as a person moves through the world. This is all explained in a manner that is easily understood by a non-neuroscientist (though it helps to have an interest in science). The title reflects the author’s research that indicates thousands of cortical columns operate in parallel, each one predicting the next sensory input, thus functioning as its own separate learning machine.
The main reason I put this book on my list to read is that it also covers the field of artificial intelligence, including the current state of machine learning and what Hawkins envisions will happen with AI in the future. He believes we need to adopt a different track than the computing specializations currently being pursued. Currently, we find computers who can beat a human at games such as Chess or Go, but the computer does not know it is playing a game, knows nothing of its history, and is not intelligent in that it cannot do multiple things or learn about its environment. He also discusses newer applications of AI, which can approximate language and communication, and answer questions based on being fed a huge database of information (such as the entire content of Wikipedia). It basically imitates human communication forms but is not doing so in any creative manner and does not represent intelligence.
At the heart of this book is Hawkins’s view that the only way to achieve artificial intelligence is to imitate the workings of the neocortex. This is the path Hawkins recommends, as he believes it is the only way we will truly achieve artificial general intelligence (AGI). The last parts of the book explain the logical conclusions, namely, that it is highly improbable that intelligent machines will replace or dominate humans. He explains the roots of false thinking (such as believing the earth is flat or the space program never happened), and how these incorrect notions are spread. He also discusses the existential threats and the possibilities of using AGI for use in space exploration. I found it fascinating and thought-provoking, and plan to read more about this field.
The main reason I put this book on my list to read is that it also covers the field of artificial intelligence, including the current state of machine learning and what Hawkins envisions will happen with AI in the future. He believes we need to adopt a different track than the computing specializations currently being pursued. Currently, we find computers who can beat a human at games such as Chess or Go, but the computer does not know it is playing a game, knows nothing of its history, and is not intelligent in that it cannot do multiple things or learn about its environment. He also discusses newer applications of AI, which can approximate language and communication, and answer questions based on being fed a huge database of information (such as the entire content of Wikipedia). It basically imitates human communication forms but is not doing so in any creative manner and does not represent intelligence.
At the heart of this book is Hawkins’s view that the only way to achieve artificial intelligence is to imitate the workings of the neocortex. This is the path Hawkins recommends, as he believes it is the only way we will truly achieve artificial general intelligence (AGI). The last parts of the book explain the logical conclusions, namely, that it is highly improbable that intelligent machines will replace or dominate humans. He explains the roots of false thinking (such as believing the earth is flat or the space program never happened), and how these incorrect notions are spread. He also discusses the existential threats and the possibilities of using AGI for use in space exploration. I found it fascinating and thought-provoking, and plan to read more about this field.
July 11, 2023
Best book I read on how the brain works.
August 19, 2021
4/5
The first part of the book is cool. As usual, Jeff explains ideas clearly and precisely. The unified theory of neocortex operations is intriguing.
- Neocortex consists of replicated similar structures (cortical columns) 2 sq. mm each.
- Every cortical column works quite independently and can construct complex models of reality
- Neocortex creates frames of reference for all surrounding objects and concepts and navigates them (most likely via place and grid neurons).
- In a nutshell, the neocortex navigates ideas in a similar way that it navigates real space around us. It means there is a unified mechanism of thoughts.
- Everything (almost) in the neocortex is a set of connected models that we can explore, re-wire and expand.
The second part of the book is a separate book in fact. It has very weak relations to the first and I was surprised to some degree to see it here. It has few new thoughts, just some philosophical discourse around AI, intelligence, human race survival, etc. There are some good enough ideas about knowledge evolution vs. genes evolution and a list of properties that define intelligent systems.
Overall, the first book was denser and more coherent, than this one. So you can safely read about a half and then skip the rest. The first half is 5/5, the second is 3/5.
The first part of the book is cool. As usual, Jeff explains ideas clearly and precisely. The unified theory of neocortex operations is intriguing.
- Neocortex consists of replicated similar structures (cortical columns) 2 sq. mm each.
- Every cortical column works quite independently and can construct complex models of reality
- Neocortex creates frames of reference for all surrounding objects and concepts and navigates them (most likely via place and grid neurons).
- In a nutshell, the neocortex navigates ideas in a similar way that it navigates real space around us. It means there is a unified mechanism of thoughts.
- Everything (almost) in the neocortex is a set of connected models that we can explore, re-wire and expand.
The second part of the book is a separate book in fact. It has very weak relations to the first and I was surprised to some degree to see it here. It has few new thoughts, just some philosophical discourse around AI, intelligence, human race survival, etc. There are some good enough ideas about knowledge evolution vs. genes evolution and a list of properties that define intelligent systems.
Overall, the first book was denser and more coherent, than this one. So you can safely read about a half and then skip the rest. The first half is 5/5, the second is 3/5.
March 2, 2021
The premise is that neuroscience has always interested me, but I am not an expert. That said, the book is divided into three parts, and the first two (respectively the new theory of a Thousand Brains and its application to AI) seemed to me extremely interesting, scientifically founded and mandatory to continue to deepen in the appropriate places. Unfortunately in the third part the author has turned into a kind of prophet of doom with some statements that I found unclear and above all unfounded, but it is probably a problem due to my global ignorance on the subject.
La premessa d'obbligo é che le neuroscienze mi hanno sempre interessato, ma non sono un'esperta. Detto questo il libro é diviso in tre parti, e le prime due (rispettivamente la nuova teoria dei 1000 cervelli e la sua applicazione all'AI) mi sono sembrate estremamente interessanti, scientificamente fondate e obbligatoriamente da continuare ad approfondire nei luoghi appositi. Purtroppo nella terza parte l'autore si é trasformato in una specie di profeta di sventura con delle affermazioni che io ho trovato poco chiare e soprattutto poco fondate, ma appunto, é probabilmente un problema dovuto alla mia ignoranza globale sull'argomento.
THANKS NETGALLEY FOR THE PREVIEW!
La premessa d'obbligo é che le neuroscienze mi hanno sempre interessato, ma non sono un'esperta. Detto questo il libro é diviso in tre parti, e le prime due (rispettivamente la nuova teoria dei 1000 cervelli e la sua applicazione all'AI) mi sono sembrate estremamente interessanti, scientificamente fondate e obbligatoriamente da continuare ad approfondire nei luoghi appositi. Purtroppo nella terza parte l'autore si é trasformato in una specie di profeta di sventura con delle affermazioni che io ho trovato poco chiare e soprattutto poco fondate, ma appunto, é probabilmente un problema dovuto alla mia ignoranza globale sull'argomento.
THANKS NETGALLEY FOR THE PREVIEW!
April 5, 2021
Jeff Hawkins and his lab have some really valuable ideas and approaches so I was excited about his new book, but I can’t recommend it, even to readers unfamiliar with Hakwins. The information available here can be better gained by reading his first book On Intelligence, listening to his interviews with Ginger Campbell, and reading some of his recent papers. Those sources contain more detailed explanations and a bit less of his ego and condescension.
The first half recaps his bio and On Intelligence and adds a short explanation of some new findings, the “thousand brains theory of intelligence” which basically says the neocortex contains large numbers of models working together to resolve inputs. He does have more specific ideas, but I feel Hawkins fails to convey them in this book.
Hawkins spends the second half musing from a soap box on theoretical topics in religion, memetics, the future of AI, and our extraterrestrial legacy. You’ll feel spoken down to even if you agree with nearly all of his points. Read Yuval Noah Harari’s Sapiens and Homo Deus instead for more thoughtful discussions of these topics on the big picture past and future of humanity.
The first half recaps his bio and On Intelligence and adds a short explanation of some new findings, the “thousand brains theory of intelligence” which basically says the neocortex contains large numbers of models working together to resolve inputs. He does have more specific ideas, but I feel Hawkins fails to convey them in this book.
Hawkins spends the second half musing from a soap box on theoretical topics in religion, memetics, the future of AI, and our extraterrestrial legacy. You’ll feel spoken down to even if you agree with nearly all of his points. Read Yuval Noah Harari’s Sapiens and Homo Deus instead for more thoughtful discussions of these topics on the big picture past and future of humanity.
September 29, 2021
"Fake history book" is one of the best euphemisms I've ever heard. Chapter about false beliefs is absolutely amazing.
March 6, 2023
I kept going back and forth here between, "This guy is the equivalent of those crackpots who think they've cracked perpetual motion or P=NP, except he got rich first so he can fund his research" and "well, I would like to do independent research if I were wealthy, and I'd sure hate it if people dismissed me as a crackpot just because I wasn't playing the academia game". In the end, I think he ends up falling somewhere between "crackpot" and "well-meaning but mostly amateurish".
A couple of general problems with this book:
- It is purportedly about his "Thousand Brains" theory, but he doesn't spend nearly enough time explaining what that is, or how it differs from the standard neuroscience model. He makes a lot out of the existence of cortical columns, and it seems like he's basically saying that the brain does distributed computing with a consensus model. That sounds a lot like the standard story to me, or at least the most compelling working theory I've heard so far. Is he saying he was dramatically influential here, or is there some subtle difference I'm missing? The book came out in 2021 and it says it's a "New Theory of Intelligence", but the distributed computing with consensus idea seems like it was popular way before then.
- Despite the fact that he devotes a surprisingly small amount of space to his actual theory, this book is padded out like crazy. He has chapters about consciousness, religion-as-a-meme, and his thoughts about how we should leave a beacon for future civilizations after we die. What‽ It feels like he had 2 chapters about the actual topic, then threw together a grab bag of other ideas and called it a book.
- There are several places where Hawkins makes bold proclamations, but his reasoning shows that he doesn't understand the problems he's purporting to solve, for example:
1. Hawkins seems to think that perceptual illusions like the "blue/gold" dress illusion tells us something about the nature of qualia, which is a fundamental misunderstanding of the "hard problem" of consciousness. The "hard problem" of consciousness is the mapping between brain activity and subjective experience. By nature you cannot describe qualia except in terms of other qualia, and so getting any information about where qualia are generated and how they are experienced in other people is extraordinarily difficult, maybe impossible. It's not enough to say, "The same neuronal spike happens in both brains, so they are having the same internal experience" because that's the easy problem of consciousness. The rest of his reasoning follows from this flawed premise.
2. Despite obviously being exposed to it, he doesn't understand the nature of the AI alignment threat. He again is pattern matching to the easy problem of "why would an AI spontaneously come up with a desire to hurt us", which is easily dismissed as basically superstition or anthropomorphism. The real alignment problem is that AIs will be in some ways fundamentally alien, possibly superhuman in many ways, and will have a goal set determined by the way we design it. If you take as a premise that AIs could have superhuman intelligence (likely), even without them being able to directly command resources, if they have any interaction with humans (and why would we build them if they didn't), they will likely be able to manipulate the humans into doing what they want. Consider the fact that humans of normal intelligence or even below-normal intelligence can pull off phone scams and get other humans (even intelligent ones) to act against their own best interests, easily. A superhuman AI is basically a genie and you are going to give it your wish and then unleash it on the world. Alignment researchers are consistently showing that we're really bad at constructing "genie wishes" that can't go horribly wrong and lead to a plausible interpretation that involves horrible outcomes.
I'm not particularly pessimistic about AI, but I at least understand that the problem is not so easily dismissed as Hawkins seems to think.
- In general, Hawkins seems to be sloppy with his reasoning and quick to be over-confident. This works out really well for TED speakers and, basically, cons, but it's not great epistemology. Some of the strongest evidence for this sloppiness is this very telling paragraph:
There's a germ of truth here: when you do have a bunch of problems to solve and you find one nice clean theory that solves them all, it's way more likely to be true than one that solves only some of the problems. That said, this is way too confident, and feels like the hallmark of someone who hasn't actually seen how messy the real world is. In the real world, some of your "constraints" won't actually be real constraints. Some of your evidence is fake or p-hacked or overblown. Sometimes your brilliant solution doesn't actually solve everything you think it solves. That's why you design experiments that increase your confidence rather than boldly proclaiming your solution a fact.
- In addition to the other problems, Hawkins also seems a bit stuck in the past. He seems to be fighting the last culture war with a bunch of talk about how religion is an evil mind virus. I think in general memetics is an interesting topic, but he doesn't get into it in enough detail, and the whole chapter about that feels a lot like him signaling his tribal membership. His examples of bad viral memes are all basically the standard skeptic stuff, "vaccines are evil", "mass shootings are faked", "climate change isn't real". He has an example of a possibly good viral meme that he believes, that every child should get a good education, but no self awareness of how this makes him look like the guy who says, "I'm too hard working" when asked for his biggest flaw, and minimal awareness that maybe his "education" meme isn't as benign as he thinks. If I were writing this chapter, I'd be trying to use it to show how viral memes of all sorts are parasitic in some ways and to challenge people about their patterns of thought. But as usual, Hawkins takes the easy path and basically says, "Hey I'm affiliated with this tribe and if you're affiliated with it, too, you are a good brain and don't need to worry about mind viruses, hooray!"
1 of 5 stars
A couple of general problems with this book:
- It is purportedly about his "Thousand Brains" theory, but he doesn't spend nearly enough time explaining what that is, or how it differs from the standard neuroscience model. He makes a lot out of the existence of cortical columns, and it seems like he's basically saying that the brain does distributed computing with a consensus model. That sounds a lot like the standard story to me, or at least the most compelling working theory I've heard so far. Is he saying he was dramatically influential here, or is there some subtle difference I'm missing? The book came out in 2021 and it says it's a "New Theory of Intelligence", but the distributed computing with consensus idea seems like it was popular way before then.
- Despite the fact that he devotes a surprisingly small amount of space to his actual theory, this book is padded out like crazy. He has chapters about consciousness, religion-as-a-meme, and his thoughts about how we should leave a beacon for future civilizations after we die. What‽ It feels like he had 2 chapters about the actual topic, then threw together a grab bag of other ideas and called it a book.
- There are several places where Hawkins makes bold proclamations, but his reasoning shows that he doesn't understand the problems he's purporting to solve, for example:
1. Hawkins seems to think that perceptual illusions like the "blue/gold" dress illusion tells us something about the nature of qualia, which is a fundamental misunderstanding of the "hard problem" of consciousness. The "hard problem" of consciousness is the mapping between brain activity and subjective experience. By nature you cannot describe qualia except in terms of other qualia, and so getting any information about where qualia are generated and how they are experienced in other people is extraordinarily difficult, maybe impossible. It's not enough to say, "The same neuronal spike happens in both brains, so they are having the same internal experience" because that's the easy problem of consciousness. The rest of his reasoning follows from this flawed premise.
2. Despite obviously being exposed to it, he doesn't understand the nature of the AI alignment threat. He again is pattern matching to the easy problem of "why would an AI spontaneously come up with a desire to hurt us", which is easily dismissed as basically superstition or anthropomorphism. The real alignment problem is that AIs will be in some ways fundamentally alien, possibly superhuman in many ways, and will have a goal set determined by the way we design it. If you take as a premise that AIs could have superhuman intelligence (likely), even without them being able to directly command resources, if they have any interaction with humans (and why would we build them if they didn't), they will likely be able to manipulate the humans into doing what they want. Consider the fact that humans of normal intelligence or even below-normal intelligence can pull off phone scams and get other humans (even intelligent ones) to act against their own best interests, easily. A superhuman AI is basically a genie and you are going to give it your wish and then unleash it on the world. Alignment researchers are consistently showing that we're really bad at constructing "genie wishes" that can't go horribly wrong and lead to a plausible interpretation that involves horrible outcomes.
I'm not particularly pessimistic about AI, but I at least understand that the problem is not so easily dismissed as Hawkins seems to think.
- In general, Hawkins seems to be sloppy with his reasoning and quick to be over-confident. This works out really well for TED speakers and, basically, cons, but it's not great epistemology. Some of the strongest evidence for this sloppiness is this very telling paragraph:
This is a good point to address a question I am often asked: How can I speak confidently about a theory if it hasn’t been tested experimentally? I just described one of these situations. I had an insight that the neocortex is infused with reference frames, and I immediately started talking about it with certainty. As I write this book, there is growing evidence to support this new idea, but it still has not been thoroughly tested. And yet, I have no hesitation describing this idea as a fact. Here is why.
As we work on a problem, we uncover what I call constraints. Constraints are things that the solution to the problem must address. I gave a few examples of constraints when describing sequence memory, for example, the Name That Tune requirement. The anatomy and physiology of the brain are also constraints. Brain theory must ultimately explain all the details of the brain, and a correct theory cannot violate any of those details.
The longer you work on a problem, the more constraints you discover and the harder it becomes to imagine a solution. The aha moments I described in this chapter were about problems that we worked on for years. Therefore, we understood these problems deeply and our list of constraints was long. The likelihood that a solution is correct increases exponentially with the number of constraints it satisfies. It is like solving a crossword puzzle: There are often several words that match an individual clue. If you pick one of those words, it could be wrong. If you find two intersecting words that work, then it is much more likely they are both correct. If you find ten intersecting words, the chance that they are all wrong is miniscule. You can write the answer in ink without any worries.
Aha moments occur when a new idea satisfies multiple constraints. The longer you have worked on a problem—and, consequently, the more constraints the solution resolves—the bigger the aha feeling and the more confident you are in trusting the answer. The idea that the neocortex is infused with reference frames solved so many constraints that I immediately knew it was correct.
There's a germ of truth here: when you do have a bunch of problems to solve and you find one nice clean theory that solves them all, it's way more likely to be true than one that solves only some of the problems. That said, this is way too confident, and feels like the hallmark of someone who hasn't actually seen how messy the real world is. In the real world, some of your "constraints" won't actually be real constraints. Some of your evidence is fake or p-hacked or overblown. Sometimes your brilliant solution doesn't actually solve everything you think it solves. That's why you design experiments that increase your confidence rather than boldly proclaiming your solution a fact.
- In addition to the other problems, Hawkins also seems a bit stuck in the past. He seems to be fighting the last culture war with a bunch of talk about how religion is an evil mind virus. I think in general memetics is an interesting topic, but he doesn't get into it in enough detail, and the whole chapter about that feels a lot like him signaling his tribal membership. His examples of bad viral memes are all basically the standard skeptic stuff, "vaccines are evil", "mass shootings are faked", "climate change isn't real". He has an example of a possibly good viral meme that he believes, that every child should get a good education, but no self awareness of how this makes him look like the guy who says, "I'm too hard working" when asked for his biggest flaw, and minimal awareness that maybe his "education" meme isn't as benign as he thinks. If I were writing this chapter, I'd be trying to use it to show how viral memes of all sorts are parasitic in some ways and to challenge people about their patterns of thought. But as usual, Hawkins takes the easy path and basically says, "Hey I'm affiliated with this tribe and if you're affiliated with it, too, you are a good brain and don't need to worry about mind viruses, hooray!"
1 of 5 stars
Displaying 1 - 30 of 817 reviews