What do you think?
Rate this book
Grey Matters
Bridges the gap between AI and neuroscience by telling the story of how the brain came to be. 'I found this book amazing' Daniel Kahneman, Winner of the Nobel Prize in Economics and bestselling author of Thinking Fast & Slow
The entirety of the human brain’s 4-billion-year story can be summarised as the culmination of five evolutionary breakthroughs, starting from the very first brains, all the way to the modern human brains. Each breakthrough emerged from new sets of brain modifications, and equipped animals with a new suite of intellectual faculties.
These five breakthroughs are the organising map to this book, and they make up our itinerary for our adventure back in time. Each breakthrough also has fascinating corollaries to breakthroughs in AI. Indeed, there will be plenty of such surprises along the way. For the innovation that enabled AI to beat humans in the game of Go – temporal difference reinforcement learning – was an innovation discovered by our fish ancestors over 500 million years ago. The solutions to many of the current mysteries in AI – such as ‘common sense’ – can be found in the tiny brain of a mouse. Where do emotions come from? Research suggests that they may have arisen simply as a solution to navigation in ancient worm brains. Unravelling this evolutionary story will reveal the hidden features of human intelligence and with them, just how your mind came to be.
The entirety of the human brain’s 4-billion-year story can be summarised as the culmination of five evolutionary breakthroughs, starting from the very first brains, all the way to the modern human brains. Each breakthrough emerged from new sets of brain modifications, and equipped animals with a new suite of intellectual faculties.
These five breakthroughs are the organising map to this book, and they make up our itinerary for our adventure back in time. Each breakthrough also has fascinating corollaries to breakthroughs in AI. Indeed, there will be plenty of such surprises along the way. For the innovation that enabled AI to beat humans in the game of Go – temporal difference reinforcement learning – was an innovation discovered by our fish ancestors over 500 million years ago. The solutions to many of the current mysteries in AI – such as ‘common sense’ – can be found in the tiny brain of a mouse. Where do emotions come from? Research suggests that they may have arisen simply as a solution to navigation in ancient worm brains. Unravelling this evolutionary story will reveal the hidden features of human intelligence and with them, just how your mind came to be.
432 pages, Kindle Edition
Published October 24, 2023
2781 people are currently reading
22057 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 598 reviews
January 7, 2025
A Brief History of Intelligence is another "big history" book. Instead of following human history (Sapiens), Western socioeconomic success (Guns, Germs, and Steel, Why Nations Fail, The WEIRDest People in the World), political history (The End of History) or genetic history (Who We Are and How We Got Here), we are treated to the evolution of the intelligence via the human brain. And although this book has several competitors in this space, it benefits from its comprehensive scope and framing.
Max Bennett, an entrepreneur in artificial intelligence, divides his book into two parallel tracks. 1) Describe the five major breakthroughs in the evolution of human intelligences and 2) compare those breakthroughs with the current state of knowledge and history of artificial intelligence. I imagine part of the exigence of this book was for Bennett to collect and organize knowledge on human intelligence in order to leverage this knowledge to advance his work in the field of AI. Because of this, the earlier portions of the narrative are more robust. Some of this is also owed to the fact that current AI techniques don't recapitulate the higher functions of the human brain but rely on more primitive approaches. Plus, I imagine he has kept some of his more tantalizing ideas out of the book.
Bennett's prose lacks the verve and panache of many of the other sweeping works that his competes in scale with, but this isn't particularly surprising given his youth. This is also a debut work. Nonetheless, Bennett delivers an eminently digestible narrative that tightly organized. He tames what in reality is an unwieldy subject, and he does so without any glibness or embarrassing oversimplifications. Bennett's strength as a writer appears to be organization and distillation. He drills down to essential insights on the evolutionary neurobiology, draws out the AI takeaways and comparisons, recaps and then moves on. It's shoe-leather science communication. For those interested in the capacity of the brain and the origins of intelligence, Bennett has saved readers a great deal of time by assembling this book.
When it comes to the five evolutionary breakthroughs, Bennett's insights are not mind-bogglingly clever or original. Every hard-working undergraduate in evolutionary biology will have come to some of the same conclusions over their coursework and reading. However, I don't want to sell Bennett short. His framing is cleverly clear and in the context of developments in AI, very topical.
So what are these breakthroughs in neurobiology and how did intelligence result? First, the very ancient common ancestors of human, bilaterians, needed a way to navigate the environment. Thus, a nervous system capable of steering emerged. Second, these early organisms had to figure out where they could obtain resources and how they could avoid predation. So another adaption emerged, reinforcement learning. The nervous system encoded valence into stimuli from the environment (food = good, predatory = bad, etc). As the evolutionary arms war between predator and prey heated up over time, the brains of ancestral species developed a way of simulating events: If I turn this way, I might find food but if I turn the other way, I may find danger. This sort of thing. At this level, primitive brains were beginning to model the world around in order to learn from it. Once modeling the world took off, brains started to model themselves. The striking thing is there is good evidence for this recursive process occurring well before humans. Bennet likes to repeat that human brains are simply scaled up chimp brains. And then the final breakthrough and probably the most obvious one is language. The ability to symbolically label and standardize inter-individual exchange with grammar enabled the accumulation and distribution of knowledge (culture). There are still a lot of questions that both Bennett and science have left unanswered, but this broad framework is sense-making in productive ways. I think a lot of reader can benefit from the book. Even myself, as someone with graduate-level training in neuroscience, genetics and the like still found a lot of the work interesting and compelling.
The later portions of the book were less robust, and it would have behooved the work to be a bit more rigorous with citations and endnotes. It would have also helped to have a more precise definition of intelligence to return to as the narrative progressed. But overall, I'm happy to recommend this one to anyone interested in the evolution of the brain and its relationship to advancements in artificial intelligence.
Comprehensive Review at Substack
Max Bennett, an entrepreneur in artificial intelligence, divides his book into two parallel tracks. 1) Describe the five major breakthroughs in the evolution of human intelligences and 2) compare those breakthroughs with the current state of knowledge and history of artificial intelligence. I imagine part of the exigence of this book was for Bennett to collect and organize knowledge on human intelligence in order to leverage this knowledge to advance his work in the field of AI. Because of this, the earlier portions of the narrative are more robust. Some of this is also owed to the fact that current AI techniques don't recapitulate the higher functions of the human brain but rely on more primitive approaches. Plus, I imagine he has kept some of his more tantalizing ideas out of the book.
Bennett's prose lacks the verve and panache of many of the other sweeping works that his competes in scale with, but this isn't particularly surprising given his youth. This is also a debut work. Nonetheless, Bennett delivers an eminently digestible narrative that tightly organized. He tames what in reality is an unwieldy subject, and he does so without any glibness or embarrassing oversimplifications. Bennett's strength as a writer appears to be organization and distillation. He drills down to essential insights on the evolutionary neurobiology, draws out the AI takeaways and comparisons, recaps and then moves on. It's shoe-leather science communication. For those interested in the capacity of the brain and the origins of intelligence, Bennett has saved readers a great deal of time by assembling this book.
When it comes to the five evolutionary breakthroughs, Bennett's insights are not mind-bogglingly clever or original. Every hard-working undergraduate in evolutionary biology will have come to some of the same conclusions over their coursework and reading. However, I don't want to sell Bennett short. His framing is cleverly clear and in the context of developments in AI, very topical.
So what are these breakthroughs in neurobiology and how did intelligence result? First, the very ancient common ancestors of human, bilaterians, needed a way to navigate the environment. Thus, a nervous system capable of steering emerged. Second, these early organisms had to figure out where they could obtain resources and how they could avoid predation. So another adaption emerged, reinforcement learning. The nervous system encoded valence into stimuli from the environment (food = good, predatory = bad, etc). As the evolutionary arms war between predator and prey heated up over time, the brains of ancestral species developed a way of simulating events: If I turn this way, I might find food but if I turn the other way, I may find danger. This sort of thing. At this level, primitive brains were beginning to model the world around in order to learn from it. Once modeling the world took off, brains started to model themselves. The striking thing is there is good evidence for this recursive process occurring well before humans. Bennet likes to repeat that human brains are simply scaled up chimp brains. And then the final breakthrough and probably the most obvious one is language. The ability to symbolically label and standardize inter-individual exchange with grammar enabled the accumulation and distribution of knowledge (culture). There are still a lot of questions that both Bennett and science have left unanswered, but this broad framework is sense-making in productive ways. I think a lot of reader can benefit from the book. Even myself, as someone with graduate-level training in neuroscience, genetics and the like still found a lot of the work interesting and compelling.
The later portions of the book were less robust, and it would have behooved the work to be a bit more rigorous with citations and endnotes. It would have also helped to have a more precise definition of intelligence to return to as the narrative progressed. But overall, I'm happy to recommend this one to anyone interested in the evolution of the brain and its relationship to advancements in artificial intelligence.
Comprehensive Review at Substack
January 21, 2025
An extraordinary combination of explication, synthesis, connecting and extension creates a form of originality deeply grounded in neuroscience, psychology, animal behavior, human evolutionary biology, computer science and more. Among the best books I have read in many years. I confess I was initially reluctant to read it, worried that a book on non-scientific topics by a non-scientist would contain too much ungrounded speculation and big thinking. I was wrong--instead, it is like great science writing that explains but also achieves much more (and as an aside, all the figures and diagrams and text tables were incredibly helpful).
Max Solomon Bennett explains the history of human intelligence through five "breakthroughs," each based on a common ancestor of animals that share that trait today (though in some cases evolution occurred simultaneously).
1. Steering and the first bilaterians. The first animals with a brain shared the same basic bilateral body plan: one end for food intake/sensing/brain, the other for excretion. They all needed to steer. Neurons provided inputs about food or danger, but a brain organized that input into a single decision about direction and navigation. This was "associative learning," like Pavlov's dog, with affect including the valence of good and bad and varying intensity. Bennett links this form of learning to how original Roombas operated.
2. Reinforcing and the first vertebrates. A more complex challenge is recognizing patterns and responding to inputs not immediately preceding outputs. It requires more sophisticated learning. Bennett explores the benefits of curiosity (transcending the local maximums that would otherwise constrain reinforcement learning), building world models, generalizing patterns, and more. He discusses how AI has solved these challenges through reinforcement learning, temporal difference learning, backward propagation, and convoluted neural networks – while noting that the brain uses different, possibly more efficient methods.
3. Simulating and the first mammals. Mammals were small, arboreal or burrowing, nocturnal, and had first-mover advantage. This created substantial benefits to planning ahead through world simulations. They developed a neocortex, which Bennett compares to generative AI, examining examples like perception's peculiar properties (less seeing than building models that fill gaps), memories (also simulations, thus potentially faulty), etc. Mammals can flexibly choose between "habits" (System 1) or model-based simulations (System 2).
4. Mentalizing and the first primates. Perhaps more creative and speculative, this focuses on primates' ability to picture others' thoughts and feelings, building a theory of mind. This capacity is integral to their social lives, enabling both cooperation and Machiavellian scheming – allowing them to reap the benefits of group living while managing the structural challenges of multiple males coexisting.
5. Speaking and the first humans. Most human uniqueness reflects differences in degree rather than kind. Language is the exception. It led to humans' explosive takeoff and cultural learning rather than evolution. Language is more flexible and not genetic (like flight, which birds must learn), combining with our large heads, premature birth, and need for social collectivity in child-rearing. Bennett offers an interesting discussion of how language only benefits people if others use it, but group selection proves too weak in practice and kin selection insufficient, leading to mutual altruism enforced by gossip.
While the arguments occasionally seem too neat in how they fit together, suggesting some rough edges of reality might be smoothed over, the work is generally fantastic and reliable. The links between different types of learning in AI were fascinating, particularly how they mapped to different stages of human intelligence evolution – with stronger connections to the first three phases (Bilaterians, vertebrates, and mammals) than the last two (primates and humans). It combines an illuminating understanding of human intelligence – and its evolutionary selection – with our current co-evolution with AI. Could not recommend more highly.
Max Solomon Bennett explains the history of human intelligence through five "breakthroughs," each based on a common ancestor of animals that share that trait today (though in some cases evolution occurred simultaneously).
1. Steering and the first bilaterians. The first animals with a brain shared the same basic bilateral body plan: one end for food intake/sensing/brain, the other for excretion. They all needed to steer. Neurons provided inputs about food or danger, but a brain organized that input into a single decision about direction and navigation. This was "associative learning," like Pavlov's dog, with affect including the valence of good and bad and varying intensity. Bennett links this form of learning to how original Roombas operated.
2. Reinforcing and the first vertebrates. A more complex challenge is recognizing patterns and responding to inputs not immediately preceding outputs. It requires more sophisticated learning. Bennett explores the benefits of curiosity (transcending the local maximums that would otherwise constrain reinforcement learning), building world models, generalizing patterns, and more. He discusses how AI has solved these challenges through reinforcement learning, temporal difference learning, backward propagation, and convoluted neural networks – while noting that the brain uses different, possibly more efficient methods.
3. Simulating and the first mammals. Mammals were small, arboreal or burrowing, nocturnal, and had first-mover advantage. This created substantial benefits to planning ahead through world simulations. They developed a neocortex, which Bennett compares to generative AI, examining examples like perception's peculiar properties (less seeing than building models that fill gaps), memories (also simulations, thus potentially faulty), etc. Mammals can flexibly choose between "habits" (System 1) or model-based simulations (System 2).
4. Mentalizing and the first primates. Perhaps more creative and speculative, this focuses on primates' ability to picture others' thoughts and feelings, building a theory of mind. This capacity is integral to their social lives, enabling both cooperation and Machiavellian scheming – allowing them to reap the benefits of group living while managing the structural challenges of multiple males coexisting.
5. Speaking and the first humans. Most human uniqueness reflects differences in degree rather than kind. Language is the exception. It led to humans' explosive takeoff and cultural learning rather than evolution. Language is more flexible and not genetic (like flight, which birds must learn), combining with our large heads, premature birth, and need for social collectivity in child-rearing. Bennett offers an interesting discussion of how language only benefits people if others use it, but group selection proves too weak in practice and kin selection insufficient, leading to mutual altruism enforced by gossip.
While the arguments occasionally seem too neat in how they fit together, suggesting some rough edges of reality might be smoothed over, the work is generally fantastic and reliable. The links between different types of learning in AI were fascinating, particularly how they mapped to different stages of human intelligence evolution – with stronger connections to the first three phases (Bilaterians, vertebrates, and mammals) than the last two (primates and humans). It combines an illuminating understanding of human intelligence – and its evolutionary selection – with our current co-evolution with AI. Could not recommend more highly.
Read
February 24, 2025Max Bennett is not an academic. He is an outsider to neuroscience, evolutionary biology or psychology. This book is broadly about the evolution of a brain, its functionality and, more specifically, its mechanisms for intelligence. He is not a trained “expert” in these areas. He is an entrepreneur working in AI development. So one might be sceptical about his “expertise”. But one should not. I think this lack of “formal” training is actually beneficial as it allows him to have a fresh look on this evolving field without the preconceptions and academia’s tendency for intransigence. Obviously, in the process of working on a book, he has widely consulted a lot of the academics and other experts.
In his research of intelligence, his approach was to compare and synthesise across three broad areas: evolutionary biology, comparative psychology (of different species) and artificial intelligence. These areas amazingly “bootstrap” each other: the discovery in the field of AI occasionally might help to explain how a biological brain performs a related task. And vice versa: the architecture of the brain might direct the design of a new AI product.
First, what this book does not try to do: Max avoids definitions. There is no definition of intelligence in the book at all. This is a quite deliberate omission. I guess it helps to keep the book focused without getting bogged down into a philosophical debate. Initially i was not sure how one can write a successful book about intelligence avoiding its definition. But after having finished the book, i think it was a good idea to keep the competing criteria for this type of definition out of the scope. Famously Supreme Court Justice Potter Stewart described his test for obscenity in 1964, "I know it when I see it.". It works well with “intelligence” for the purpose of this book. Another controversial topic of consciousness and its evolvement is also not defined and practically avoided.
However for my purpose, it might be helpful to give some definition of intelligence as far as I understood it in the context of the book. Intelligence is an ability of a system (biological or artificial) either to assess trade offs and choose between them; or to learn independently by itself (self-learning) or from others; or often - both of these features. Max has developed a framework of five stages (“breakthroughs”) to show how this ability has emerged through the process of evolution. I found the framework very elegant and possessing a significant explanatory power.
But before i get to these breakthroughs, i want to mention that the book is brimming with curious factoids and interesting insights. It is the one of those long histories that are always a bit speculative and partly based on emergent ideas in natural and social sciences. Majority of such histories focus on prehistory of the humans and their societies. Those could be very speculative and often are driven by the author’s preconditioned set of ideas. Fortunately, Max is more interested what happened with intelligence well before humans, practically from the inception of life. So the book has mainly avoided sociological debates and controversies pertinent to such literature.
Before his first main protagonists about to appear, he takes time setting the scene. In the process, he tells the stories of at least two major ecological catastrophes, describes the features of neurones and initial single cell organisms. He also pays attention to the emergence of the eternal struggle: the tragically ruthless but logical war of species that has forever ended “the world of utopian peace”. Initially all existing bacteria were feeding themselves through photosynthesis. However one day fungi and breathing bacteria have appeared. This “respiratory life could survive only by stealing the energetic prize—the sugary innards—of photosynthetic life”. This has started an arms race of the predator/pray evolution, “a perpetual feedback loop: offensive innovations led to defensive innovations that required further offensive innovations”. And look: after millions years it does not get better. We are still at it, though now armed with all this accumulated sophisticated “intelligence”.
But let’s come back to the framework. From around 600 millions years in the past to our times, Max loosely maps the emergence of the brain and its individual structures to the key intelligent capabilities developed by the species’ lineage leading to the humans. His breakthroughs are as follows:
1. Steering (circa 600 millions years ago) - the first nucleus of a brain provides a creature with the ability to move straight and turn towards “good” things/ away from “bad” things.
2. Reinforcing (circa 550 millions years ago) - the brain has developed six key substructures. The main of them: basal ganglia and cortex. With their help a creature is able to learn through trial and error by rewarding itself on a right move and punishing for a bad one; the creature is able to recognise the patterns around.
3. Simulating (200 millions years ago) - neocortex appears that allows a mammal to pause the current recognition for a second and simulate “possible futures” only in their heads before making a decision which possible scenario to chose and how to act on it. In Max’s words it allows the brain to “to render a simulation that is not a current one.” That is roughly what we call “imagination” or “an episodic memory” (if the simulation is related to past as opposed to future). In machine speak, the brain is using its generative model to propagate the data top-down. In human speak, the brain produce an imagined irreducible 3d scene that the animal (human included) uses to make decisions and act on the real world.
4. Mentalizing (50 millions years ago) The primates are able to make a model of the other’s mental state. Respectively they are able to learn by imitation, to play politics, to cheat and maybe to “empathise”? Though Max does not get into this.
5. Speaking (noone knows exactly when but 50,000 years ago is probably a good estimation). This is the easiest to understand but the hardest to pin down one.
The last two “breakthroughs” are not directly related to any new physical brain structures and their far reaching “intellectual consequences” are quit difficult to define and easy to speculate about. What follows is focused on the first two “breakthroughs”. This is mainly due to my personal fascination with them. On the top of it, GR does not give me enough space in the box even to discuss the two of them. The book pays equal attention to all five of them. It also contains a short conclusion chapter on future in Max’s view. So if you find yourself intrigued, I would strongly recommend this book. But here I would stick to fish and nematodes, my new heroes.
Steering. Breakthrough #1.
I’ve simply fallen in love with the first protagonists of this drama: the bilaterians, the ancestors of nematodes, a variety of worms. Who would guess that nematodes would be so much like humans! The first creatures of this kind have evolved around 600 millions years ago. Their nervous system including a nucleus of a brain contained only 302 neurones (as opposed to 86 billion neurones that form 100 trillion connections to each other in the human brain). On a surface they could do only one practical thing: move straight or turn towards food and turn away from a danger. And yet it is amazing how many features these humble creatures have possessed already we might recognise in ourselves. At a closer look, they’ve appeared to be the dopamine-induced pleasure seekers, occasionally depression-prone but generally eager to learn. Sounds familiar? And all due to just those 302 little neurones. But let’s try to see how it works.
To be able to steer efficiently one should have a symmetrical body. The bilaterians are named after this feature. But something should give a nematode a “motivation”when to go forward and when to turn. For this even a nematode should possess a sort of primitive system of values: he needs to be able to feel what is “good” and what “bad” for him. This is of course not a moral system that the humans have developed and struggled to maintain. Though sometimes I think that in spite of our joint efforts occasionally we are not that far from nematodes in certain judgement or its lack.
Respectively feeling something “good” (a morsel of food or better temperature), a nematode would steer towards it and he would turn away if he feels something bad like a predator. It is probably not surprising that good/bad is totally subjective in relation to him and his needs. The goodness (and badness) is wired in his valence neurones. All the creatures evolved later possess these type of neurones as well even if the mechanism of their use has become much more complex*.
So how does it work in practice? Through his sensory neurones (predominantly smell) a nematode feels the proximity of food. Valence neurones tell him it is good news. They are directly connected to his motor neurones that move him in the direction of the food smell. But in many situations the creature needs to choose between the conflicting choices. For example a predator smell and a food smell from the same direction. Here the benefit of the integrated brain comes to help. Valence neuron’s signals are computed for “goodness” and “badness”. The one side wins. That would inhibit another side. So here we are: a nematode has made his first decision.
What makes nematodes even more “human” is his ability to toggle between the affective states. An affective state is something akin to what we call an emotional state. But we are not allowed to use this word not to create a precedent of mixing a very complex, “culture constructed” and society’s related” phenomena with the urges of a nematode. Though to what extent our emotional life is socially “constructed” and different from a nematode is open for debate.
Each of these affective states are broadly characterised by the two parameters: 1) valence; 2) a degree of arousal. Does it feel good or bad? And how strong this feeling is. Depending of these two parameters and their combinations there are four affective states. Apart from valence neurones mentioned before, they are regulated by another set of neurones which produce neuromodulators. These are the chemicals that when released affect a wider area, last longer than a neurone’s electrical spike and makes different group of neurones more sensitive than usual while potentially inhibiting other connections.
What neuromodulators affectively do is allowing an animal to persist in a certain state for longer, to keep it focused on the task so to speak. Without these chemicals any random fluke in the environment might rapidly change an animal’s motivation. So the nematode would jump from one task to another never achieving any physiological goals (ie eating, resting or running away successfully). It is appeared that we are well familiar with these substances:
This last state caused by a threat and releasing adrenaline we would call “a stress”. “Fight and flight” response for a nematode would be trying to physically turn away from a threat or towards the food (if starving) and run.
The last of the four nematode’s affective states is “rest and relief” driven by opioids:
Humans seem to be reacting to opioids quite like nematodes. Fortunately for the latter they did not have to tackle with some dishonest greedy players in the pharmaceutical industry causing the massive opioid epidemic. Less fortunately for nematodes they still have to deal with many consequences of human decisions in their environment.
What i found the most poignant and apt was that: “After just two minutes of no relief from this inescapable stressor, nematodes do something surprising: they give up.”. They just stop trying to escape, to fight, anything in order to actively protect themselves. They just lie there. Max comments:
It is appeared that depression is almost 600 millions years older than the first human. Also the tendency to “give up” under the prolonged stress seems to be almost a body response. At least in nematode’s case. This fact is pregnant with meaning in application to human suffering, especially the one associated with a severe coercion inflicted by other humans: slavery, Holocaust, violent exploitation and numerous bloody conflicts through the centuries and the present. “Why they didn’t resist more?” - the question is often posed referring to some of the biggest mass atrocities. It looks like the nematodes may have provided us with the one of the clues for the answer.
Another amazing feature that the bilaterians can learn through associations albeit almost immediate ones in terms of time. If they regularly face a cue within maximum two seconds of a reward, they start to associate the cue with the reward (like infamous Pavlov’s dogs). This is the famous “what fire together wire together” when a connection is established between the two neurones consequently fired by the sequence of events. For example, if a smelly liquid is poured within a second of a morsel of food, after a few gos the animal start moving towards the liquid’s smell before the smell of the food appears.
Now i hope i’ve convinced you that nematodes are the most awesome creatures, almost like people. Well maybe not quite as they lack imagination for truly great things as well as proper cruel things. In fact, they do not seem to have an imagination full stop. This resurfaced later in further breakthroughs. And in terms of cruelty, according to Max, it has emerged only with language and only in humans. But now it is time to move on.
Reinforcing learning. Breakthrough #2.
50 millions years past by as a day and the vertebrates evolved. The majority of them initially were fish. If the bilaterians have just had a nucleus of neurones as a brain, these new guys have developed quite a lot of new structures within this powerful organ. Except for the cortex which is different in the mammals, the brains of fish are very similar to ours. Among other things their brain contain a structure called basal ganglia which is crucial for this breakthrough.
Apparently a fish (and everyone after her in our evolutionary lineage) can learn very effectively through trial and error. More specifically:
“The second breakthrough was reinforcement learning: the ability to learn arbitrary sequences of actions through trial and error.”
What it means in practice you might be able to find out from the first comment to this review as it did not fit into the box. Even better alternative is the book itself of course.
The main output of this breakthrough is that the vertebrates are able to learn through trial and error, recognise patterns and even have a simple spacial map in their brain. And the input from the AI science has helped the neuroscientists to understand the mechanism of these skills.
With pattern recognitions through inference, ability to be induced to act by dopamine and “memorise” some of the results, it is not that far from language, schools, states and wars. I am joking of course. But we’ve got to a good start. The following three “breakthroughs” are also very revealing and complex, but easy to understand due to the clarity and details of Max’s writing.
This is just a framework of course and might be a bit too “neat” : everything logically set out and well explained without particular contradictions or gaps that might have been there if a different optics applied. I am not sure how useful this framework would be for anyone closely involved in this research field. I guess some other author would be able to construct a different story from the same facts. I am not qualified to judge how complete the factual base of the book is and whether there are some studies contradicting the findings summarised in the book. Also this field is rapidly emerging. But as a curious general reader, I loved the overview presented and the story set out in the book, found it convincing, well sourced with evidence and also enjoyable.
_____
* I guess this might be compared to pleasure/pain dichotomy. Though it is unclear whether the nematodes actually able feel those. This might bring us to the debate whether this creatures possess consciousness. I believe they do; and they feel pain and pleasure, but i agree with Max’s strategy to avoid this debate for the sake of intelligence.
In his research of intelligence, his approach was to compare and synthesise across three broad areas: evolutionary biology, comparative psychology (of different species) and artificial intelligence. These areas amazingly “bootstrap” each other: the discovery in the field of AI occasionally might help to explain how a biological brain performs a related task. And vice versa: the architecture of the brain might direct the design of a new AI product.
First, what this book does not try to do: Max avoids definitions. There is no definition of intelligence in the book at all. This is a quite deliberate omission. I guess it helps to keep the book focused without getting bogged down into a philosophical debate. Initially i was not sure how one can write a successful book about intelligence avoiding its definition. But after having finished the book, i think it was a good idea to keep the competing criteria for this type of definition out of the scope. Famously Supreme Court Justice Potter Stewart described his test for obscenity in 1964, "I know it when I see it.". It works well with “intelligence” for the purpose of this book. Another controversial topic of consciousness and its evolvement is also not defined and practically avoided.
However for my purpose, it might be helpful to give some definition of intelligence as far as I understood it in the context of the book. Intelligence is an ability of a system (biological or artificial) either to assess trade offs and choose between them; or to learn independently by itself (self-learning) or from others; or often - both of these features. Max has developed a framework of five stages (“breakthroughs”) to show how this ability has emerged through the process of evolution. I found the framework very elegant and possessing a significant explanatory power.
But before i get to these breakthroughs, i want to mention that the book is brimming with curious factoids and interesting insights. It is the one of those long histories that are always a bit speculative and partly based on emergent ideas in natural and social sciences. Majority of such histories focus on prehistory of the humans and their societies. Those could be very speculative and often are driven by the author’s preconditioned set of ideas. Fortunately, Max is more interested what happened with intelligence well before humans, practically from the inception of life. So the book has mainly avoided sociological debates and controversies pertinent to such literature.
Before his first main protagonists about to appear, he takes time setting the scene. In the process, he tells the stories of at least two major ecological catastrophes, describes the features of neurones and initial single cell organisms. He also pays attention to the emergence of the eternal struggle: the tragically ruthless but logical war of species that has forever ended “the world of utopian peace”. Initially all existing bacteria were feeding themselves through photosynthesis. However one day fungi and breathing bacteria have appeared. This “respiratory life could survive only by stealing the energetic prize—the sugary innards—of photosynthetic life”. This has started an arms race of the predator/pray evolution, “a perpetual feedback loop: offensive innovations led to defensive innovations that required further offensive innovations”. And look: after millions years it does not get better. We are still at it, though now armed with all this accumulated sophisticated “intelligence”.
But let’s come back to the framework. From around 600 millions years in the past to our times, Max loosely maps the emergence of the brain and its individual structures to the key intelligent capabilities developed by the species’ lineage leading to the humans. His breakthroughs are as follows:
1. Steering (circa 600 millions years ago) - the first nucleus of a brain provides a creature with the ability to move straight and turn towards “good” things/ away from “bad” things.
2. Reinforcing (circa 550 millions years ago) - the brain has developed six key substructures. The main of them: basal ganglia and cortex. With their help a creature is able to learn through trial and error by rewarding itself on a right move and punishing for a bad one; the creature is able to recognise the patterns around.
3. Simulating (200 millions years ago) - neocortex appears that allows a mammal to pause the current recognition for a second and simulate “possible futures” only in their heads before making a decision which possible scenario to chose and how to act on it. In Max’s words it allows the brain to “to render a simulation that is not a current one.” That is roughly what we call “imagination” or “an episodic memory” (if the simulation is related to past as opposed to future). In machine speak, the brain is using its generative model to propagate the data top-down. In human speak, the brain produce an imagined irreducible 3d scene that the animal (human included) uses to make decisions and act on the real world.
4. Mentalizing (50 millions years ago) The primates are able to make a model of the other’s mental state. Respectively they are able to learn by imitation, to play politics, to cheat and maybe to “empathise”? Though Max does not get into this.
5. Speaking (noone knows exactly when but 50,000 years ago is probably a good estimation). This is the easiest to understand but the hardest to pin down one.
The last two “breakthroughs” are not directly related to any new physical brain structures and their far reaching “intellectual consequences” are quit difficult to define and easy to speculate about. What follows is focused on the first two “breakthroughs”. This is mainly due to my personal fascination with them. On the top of it, GR does not give me enough space in the box even to discuss the two of them. The book pays equal attention to all five of them. It also contains a short conclusion chapter on future in Max’s view. So if you find yourself intrigued, I would strongly recommend this book. But here I would stick to fish and nematodes, my new heroes.
Steering. Breakthrough #1.
I’ve simply fallen in love with the first protagonists of this drama: the bilaterians, the ancestors of nematodes, a variety of worms. Who would guess that nematodes would be so much like humans! The first creatures of this kind have evolved around 600 millions years ago. Their nervous system including a nucleus of a brain contained only 302 neurones (as opposed to 86 billion neurones that form 100 trillion connections to each other in the human brain). On a surface they could do only one practical thing: move straight or turn towards food and turn away from a danger. And yet it is amazing how many features these humble creatures have possessed already we might recognise in ourselves. At a closer look, they’ve appeared to be the dopamine-induced pleasure seekers, occasionally depression-prone but generally eager to learn. Sounds familiar? And all due to just those 302 little neurones. But let’s try to see how it works.
To be able to steer efficiently one should have a symmetrical body. The bilaterians are named after this feature. But something should give a nematode a “motivation”when to go forward and when to turn. For this even a nematode should possess a sort of primitive system of values: he needs to be able to feel what is “good” and what “bad” for him. This is of course not a moral system that the humans have developed and struggled to maintain. Though sometimes I think that in spite of our joint efforts occasionally we are not that far from nematodes in certain judgement or its lack.
Respectively feeling something “good” (a morsel of food or better temperature), a nematode would steer towards it and he would turn away if he feels something bad like a predator. It is probably not surprising that good/bad is totally subjective in relation to him and his needs. The goodness (and badness) is wired in his valence neurones. All the creatures evolved later possess these type of neurones as well even if the mechanism of their use has become much more complex*.
So how does it work in practice? Through his sensory neurones (predominantly smell) a nematode feels the proximity of food. Valence neurones tell him it is good news. They are directly connected to his motor neurones that move him in the direction of the food smell. But in many situations the creature needs to choose between the conflicting choices. For example a predator smell and a food smell from the same direction. Here the benefit of the integrated brain comes to help. Valence neuron’s signals are computed for “goodness” and “badness”. The one side wins. That would inhibit another side. So here we are: a nematode has made his first decision.
What makes nematodes even more “human” is his ability to toggle between the affective states. An affective state is something akin to what we call an emotional state. But we are not allowed to use this word not to create a precedent of mixing a very complex, “culture constructed” and society’s related” phenomena with the urges of a nematode. Though to what extent our emotional life is socially “constructed” and different from a nematode is open for debate.
Each of these affective states are broadly characterised by the two parameters: 1) valence; 2) a degree of arousal. Does it feel good or bad? And how strong this feeling is. Depending of these two parameters and their combinations there are four affective states. Apart from valence neurones mentioned before, they are regulated by another set of neurones which produce neuromodulators. These are the chemicals that when released affect a wider area, last longer than a neurone’s electrical spike and makes different group of neurones more sensitive than usual while potentially inhibiting other connections.
What neuromodulators affectively do is allowing an animal to persist in a certain state for longer, to keep it focused on the task so to speak. Without these chemicals any random fluke in the environment might rapidly change an animal’s motivation. So the nematode would jump from one task to another never achieving any physiological goals (ie eating, resting or running away successfully). It is appeared that we are well familiar with these substances:
“ dopamine is released by nearby rewards and triggers the affective state of arousal and pursuit (exploitation); and serotonin is released by the consumption of rewards and triggers a state of low arousal, inhibiting the pursuit of rewards (satiation)... class of neuromodulators: norepinephrine, octopamine, and epinephrine (also called adrenaline) are released by negative-valanced stimuli and trigger the well-known fight-or-flight response: increasing heart rate, constricting blood vessels, dilating pupils and suppressing various luxury activities, such as sleep, reproduction, and digestion.”
This last state caused by a threat and releasing adrenaline we would call “a stress”. “Fight and flight” response for a nematode would be trying to physically turn away from a threat or towards the food (if starving) and run.
The last of the four nematode’s affective states is “rest and relief” driven by opioids:
“When the stressor goes away and adrenaline levels drop, nematodes do not go back to their baseline state. Instead, the leftover anti-stress chemicals initiate a suite of recovery-related processes—immune responses, appetite, and digestion are turned back on. These relief-and-recover chemicals like opioids do this in part by enhancing serotonin and dopamine signals (both of which are inhibited by acute stressors). Opioids also inhibit negative-valence neurons, which helps an animal recover and rest despite any injuries. This, of course, is why opioids are such potent painkillers across all bilaterians. Opioids also keep certain luxury functions, such as reproductive activities, turned off until the relief-and-recover process is done; this is why opioids decrease sex drive. It is no surprise, then, that nematodes, other invertebrates, and humans all have similar responses to opioids—prolonged bouts of feeding, inhibited pain responses, and inhibited reproductive behaviour.
Opioids make everything better; they increase liking reactions and decrease disliking reactions; increasing pleasure and inhibiting pain.”
Humans seem to be reacting to opioids quite like nematodes. Fortunately for the latter they did not have to tackle with some dishonest greedy players in the pharmaceutical industry causing the massive opioid epidemic. Less fortunately for nematodes they still have to deal with many consequences of human decisions in their environment.
What i found the most poignant and apt was that: “After just two minutes of no relief from this inescapable stressor, nematodes do something surprising: they give up.”. They just stop trying to escape, to fight, anything in order to actively protect themselves. They just lie there. Max comments:
“chronic stress differs from acute stress in at least one important way: it turns off arousal and motivation...This is, perhaps, the most primitive form of depression.”
It is appeared that depression is almost 600 millions years older than the first human. Also the tendency to “give up” under the prolonged stress seems to be almost a body response. At least in nematode’s case. This fact is pregnant with meaning in application to human suffering, especially the one associated with a severe coercion inflicted by other humans: slavery, Holocaust, violent exploitation and numerous bloody conflicts through the centuries and the present. “Why they didn’t resist more?” - the question is often posed referring to some of the biggest mass atrocities. It looks like the nematodes may have provided us with the one of the clues for the answer.
Another amazing feature that the bilaterians can learn through associations albeit almost immediate ones in terms of time. If they regularly face a cue within maximum two seconds of a reward, they start to associate the cue with the reward (like infamous Pavlov’s dogs). This is the famous “what fire together wire together” when a connection is established between the two neurones consequently fired by the sequence of events. For example, if a smelly liquid is poured within a second of a morsel of food, after a few gos the animal start moving towards the liquid’s smell before the smell of the food appears.
Now i hope i’ve convinced you that nematodes are the most awesome creatures, almost like people. Well maybe not quite as they lack imagination for truly great things as well as proper cruel things. In fact, they do not seem to have an imagination full stop. This resurfaced later in further breakthroughs. And in terms of cruelty, according to Max, it has emerged only with language and only in humans. But now it is time to move on.
Reinforcing learning. Breakthrough #2.
50 millions years past by as a day and the vertebrates evolved. The majority of them initially were fish. If the bilaterians have just had a nucleus of neurones as a brain, these new guys have developed quite a lot of new structures within this powerful organ. Except for the cortex which is different in the mammals, the brains of fish are very similar to ours. Among other things their brain contain a structure called basal ganglia which is crucial for this breakthrough.
Apparently a fish (and everyone after her in our evolutionary lineage) can learn very effectively through trial and error. More specifically:
“The second breakthrough was reinforcement learning: the ability to learn arbitrary sequences of actions through trial and error.”
What it means in practice you might be able to find out from the first comment to this review as it did not fit into the box. Even better alternative is the book itself of course.
The main output of this breakthrough is that the vertebrates are able to learn through trial and error, recognise patterns and even have a simple spacial map in their brain. And the input from the AI science has helped the neuroscientists to understand the mechanism of these skills.
With pattern recognitions through inference, ability to be induced to act by dopamine and “memorise” some of the results, it is not that far from language, schools, states and wars. I am joking of course. But we’ve got to a good start. The following three “breakthroughs” are also very revealing and complex, but easy to understand due to the clarity and details of Max’s writing.
This is just a framework of course and might be a bit too “neat” : everything logically set out and well explained without particular contradictions or gaps that might have been there if a different optics applied. I am not sure how useful this framework would be for anyone closely involved in this research field. I guess some other author would be able to construct a different story from the same facts. I am not qualified to judge how complete the factual base of the book is and whether there are some studies contradicting the findings summarised in the book. Also this field is rapidly emerging. But as a curious general reader, I loved the overview presented and the story set out in the book, found it convincing, well sourced with evidence and also enjoyable.
_____
* I guess this might be compared to pleasure/pain dichotomy. Though it is unclear whether the nematodes actually able feel those. This might bring us to the debate whether this creatures possess consciousness. I believe they do; and they feel pain and pleasure, but i agree with Max’s strategy to avoid this debate for the sake of intelligence.
May 11, 2025
A Brief History of Intelligence is not brief as it is looking into four billion years of evolutionary neurobiology and then comparative psychology to map five breakthroughs in intelligence (neurobehavioral transformations) to the concepts of AI with a perspective that “The better we are at understanding our own minds, the better equipped we are to create artificial minds in our image”. Bennett's synthesis is interesting indeed, but I need to reread it, and several of the references to say anything intelligent about any of it. My mind is still wandering on the part on language. Great read, both mind blowing and mind boggling.
December 8, 2024
How would you figure out how to make intelligent machines? The historical precedent for building intelligence sounds like a good start. What if that precedent is a disordered mess of papers and insights across several fields and decades that may require a couple of textbooks to start to understand? In that case, you organize the mess into a neat evolutionary story, filling gaps with proposed hypotheses along the way. Max Bennet did a fantastic job tidying up his story.
Bennett's book is a delightful tour of the evolution of intelligence through the specific path from the first animals to homo sapiens. Other forms of animal intelligence are mentioned but not covered in the book. The framework depends on identifying 5 breakthroughs in the evolutionary history of human intelligence, and pondering which one will be the 6th. Any excuses for metacognition are welcome, but this book is a particularly fun one.
More than fun, it's an impressive endeavor. As far as I know, there's no similarly accessible synthesis on intelligence out there. It successfully integrates work in reinforcement learning, linguistics, biology, and neuroscience. The proposed 5-breakthrough path is the result of a lot of reading and deep familiarity with the research discussed.
This book is likely my favorite of 2024. It doesn't feel as outdated as it could be given the pace of robotics advances this year and accounts for the capabilities LLMs had as of July 2023. I hope Bennett decides to publish more books, for this is a great start.
Bennett's book is a delightful tour of the evolution of intelligence through the specific path from the first animals to homo sapiens. Other forms of animal intelligence are mentioned but not covered in the book. The framework depends on identifying 5 breakthroughs in the evolutionary history of human intelligence, and pondering which one will be the 6th. Any excuses for metacognition are welcome, but this book is a particularly fun one.
More than fun, it's an impressive endeavor. As far as I know, there's no similarly accessible synthesis on intelligence out there. It successfully integrates work in reinforcement learning, linguistics, biology, and neuroscience. The proposed 5-breakthrough path is the result of a lot of reading and deep familiarity with the research discussed.
This book is likely my favorite of 2024. It doesn't feel as outdated as it could be given the pace of robotics advances this year and accounts for the capabilities LLMs had as of July 2023. I hope Bennett decides to publish more books, for this is a great start.
February 17, 2025
This is a book of evolutionary neuroscience, starting with the primordial soup and proceeding through time to the modern human brain. The author focuses on the five big breakthroughs in the development of human intelligence and correlates them to what is currently happening in the development of artificial intelligence. The five breakthroughs include: directional movement, learning through positive and negative reinforcement, imagining various outcomes before selecting a course of action, social learning (including mimicking and anticipating future needs), and the development of communication through language.
The author outlines these breakthroughs in an accessible manner. The book is well organized and logically explained. It requires the reader’s keen interest in neuroscience. It is not a book to pick up on a whim. Even though the title mentions “a brief history,” be prepared to go into detail about each breakthrough. I particularly enjoyed the explanations of how artificial intelligence currently performs in each of the five areas. Perhaps I should not be surprised at how much behavioral psychology is included, and I was familiar with many of the experiments mentioned. If you are interested in a journey through the evolution of the human brain, this book will fill the bill. It will be interesting to see how closely AI will be able to mimic the real thing. So far, it is not even close.
The author outlines these breakthroughs in an accessible manner. The book is well organized and logically explained. It requires the reader’s keen interest in neuroscience. It is not a book to pick up on a whim. Even though the title mentions “a brief history,” be prepared to go into detail about each breakthrough. I particularly enjoyed the explanations of how artificial intelligence currently performs in each of the five areas. Perhaps I should not be surprised at how much behavioral psychology is included, and I was familiar with many of the experiments mentioned. If you are interested in a journey through the evolution of the human brain, this book will fill the bill. It will be interesting to see how closely AI will be able to mimic the real thing. So far, it is not even close.
October 1, 2025
Though AI may be on the lips of many, there is the realm of non-fungible skills that are unto a lifesaver in a raging ocean on which many may drown. For example…
“The more we understand about our own minds, the better equipped we are to create artificial minds in our image. The more we understand about the process by which our minds came to be, the better equipped we are to choose which features of intelligence we want to discard, which we want to preserve, and which we want to improve upon.”
—Max Bennett
“The more we understand about our own minds, the better equipped we are to create artificial minds in our image. The more we understand about the process by which our minds came to be, the better equipped we are to choose which features of intelligence we want to discard, which we want to preserve, and which we want to improve upon.”
—Max Bennett
January 27, 2024
Honestly a tough read for me cuz I'm intellectually challenged. All I remembered from the book was how AI was a thing in the 1950s..(?) Way before the 2000s, and how it started from this concept called reinforcement learning based on actual rewards (this was v1 that's why AI didn't take off because it took too long to get to the actual result) VS being credited for expected reward. Anyway, its a long read - if anyone wants to know about intelligence on Earth in these 4 billion years, this may be worth a shot.
December 28, 2024
It's about intelligence but not from the human intelligence research perspective. Rather it's a history of how intelligence evolved on earth and how mimicking it led to AI.
February 9, 2025
‘A Brief History of Intelligence: Evolution, AI, and the Five Breakthroughs That Made Our Brains’ (2023) by Max Bennett is a work that takes us on a fascinating journey through the evolution of human intelligence, exploring how biological advances have shaped our cognitive capacity and how this vision has come to inform the development of artificial intelligence (AI). Bennett identifies five crucial transformations in the evolution of the biological brain that would support advances in human intelligence, each building on the previous one to culminate in the form of intelligence that characterises human beings today.
Loved it.
For the complete review in Portuguese, with excerpts, check my blog: https://narrativax.blogspot.com/2025/...
Loved it.
For the complete review in Portuguese, with excerpts, check my blog: https://narrativax.blogspot.com/2025/...
July 6, 2024
This book is an excellent read if you want to know the current state of AI and human brain comparisons. It is not much for forecasting the future, which is why only four stars.
October 17, 2024
‘A Brief History of Intelligence: Evolution, AI, and the Five Breakthroughs That Made Our Brains’ by Max Bennett is a wonderful science book about the evolution of the brain, and what is currently known about how it functions. It is simply the best book on the brain I’ve ever read!
I have copied the book blurb:
”“I found this book amazing. I read it through quickly because it was so interesting, then turned around and read much of it again.”—Daniel Kahneman, winner of the Nobel Prize in Economics and bestselling author of Thinking Fast & Slow
“I've been recommending A Brief History of Intelligence to everyone I know. A truly novel, beautifully crafted thesis on what intelligence is and how it has developed since the dawn of life itself."—Angela Duckworth, bestselling author of Grit
Equal parts Sapiens, Behave, and Superintelligence, but wholly original in scope, A Brief History of Intelligence offers a paradigm shift for how we understand neuroscience and AI. Artificial intelligence entrepreneur Max Bennett chronicles the five “breakthroughs” in the evolution of human intelligence and reveals what brains of the past can tell us about the AI of tomorrow.
In the last decade, capabilities of artificial intelligence that had long been the realm of science fiction have, for the first time, become our reality. AI is now able to produce original art, identify tumors in pictures, and even steer our cars. And yet, large gaps remain in what modern AI systems can achieve—indeed, human brains still easily perform intellectual feats that we can’t replicate in AI systems. How is it possible that AI can beat a grandmaster at chess but can’t effectively load a dishwasher? As AI entrepreneur Max Bennett compellingly argues, finding the answer requires diving into the billion-year history of how the human brain evolved; a history filled with countless half-starts, calamities, and clever innovations. Not only do our brains have a story to tell—the future of AI may depend on it.
Now, in A Brief History of Intelligence, Bennett bridges the gap between neuroscience and AI to tell the brain’s evolutionary story, revealing how understanding that story can help shape the next generation of AI breakthroughs. Deploying a fresh perspective and working with the support of many top minds in neuroscience, Bennett consolidates this immense history into an approachable new framework, identifying the “Five Breakthroughs” that mark the brain’s most important evolutionary leaps forward. Each breakthrough brings new insight into the biggest mysteries of human intelligence. Containing fascinating corollaries to developments in AI, A Brief History of Intelligence shows where current AI systems have matched or surpassed our brains, as well as where AI systems still fall short. Simply put, until AI systems successfully replicate each part of our brain’s long journey, AI systems will fail to exhibit human-like intelligence.
Endorsed and lauded by many of the top neuroscientists in the field today, Bennett’s work synthesizes the most relevant scientific knowledge and cutting-edge research into an easy-to-understand and riveting evolutionary story. With sweeping scope and stunning insights, A Brief History of Intelligence proves that understanding the arc of our brain’s history can unlock the tools for successfully navigating our technological future.”
There is a lot of information and artwork in this factually dense, but very readable, science book. I learned more about the brain and its evolution then I have done in reading any other book about this area of science. If your time is limited in reading about the recent discoveries and science about the brain, this is the one book I recommend to get. It pulls together all of the bits of information I have read here and there in other science and AI books. He explains with an in-depth synthesis for the general reader. The author focuses primarily on evolution and function of the brain, less on how computer scientists are transferring the skills of the brain to AI algorithms on the computer. For that angle, I recommend The Alignment Problem: Machine Learning and Human Values by Brian Christian.
Included are a Glossary, extensive Notes, Bibliography and Index sections, plus Art, Photo, and Figure credits in the back of the book.
I have copied the book blurb:
”“I found this book amazing. I read it through quickly because it was so interesting, then turned around and read much of it again.”—Daniel Kahneman, winner of the Nobel Prize in Economics and bestselling author of Thinking Fast & Slow
“I've been recommending A Brief History of Intelligence to everyone I know. A truly novel, beautifully crafted thesis on what intelligence is and how it has developed since the dawn of life itself."—Angela Duckworth, bestselling author of Grit
Equal parts Sapiens, Behave, and Superintelligence, but wholly original in scope, A Brief History of Intelligence offers a paradigm shift for how we understand neuroscience and AI. Artificial intelligence entrepreneur Max Bennett chronicles the five “breakthroughs” in the evolution of human intelligence and reveals what brains of the past can tell us about the AI of tomorrow.
In the last decade, capabilities of artificial intelligence that had long been the realm of science fiction have, for the first time, become our reality. AI is now able to produce original art, identify tumors in pictures, and even steer our cars. And yet, large gaps remain in what modern AI systems can achieve—indeed, human brains still easily perform intellectual feats that we can’t replicate in AI systems. How is it possible that AI can beat a grandmaster at chess but can’t effectively load a dishwasher? As AI entrepreneur Max Bennett compellingly argues, finding the answer requires diving into the billion-year history of how the human brain evolved; a history filled with countless half-starts, calamities, and clever innovations. Not only do our brains have a story to tell—the future of AI may depend on it.
Now, in A Brief History of Intelligence, Bennett bridges the gap between neuroscience and AI to tell the brain’s evolutionary story, revealing how understanding that story can help shape the next generation of AI breakthroughs. Deploying a fresh perspective and working with the support of many top minds in neuroscience, Bennett consolidates this immense history into an approachable new framework, identifying the “Five Breakthroughs” that mark the brain’s most important evolutionary leaps forward. Each breakthrough brings new insight into the biggest mysteries of human intelligence. Containing fascinating corollaries to developments in AI, A Brief History of Intelligence shows where current AI systems have matched or surpassed our brains, as well as where AI systems still fall short. Simply put, until AI systems successfully replicate each part of our brain’s long journey, AI systems will fail to exhibit human-like intelligence.
Endorsed and lauded by many of the top neuroscientists in the field today, Bennett’s work synthesizes the most relevant scientific knowledge and cutting-edge research into an easy-to-understand and riveting evolutionary story. With sweeping scope and stunning insights, A Brief History of Intelligence proves that understanding the arc of our brain’s history can unlock the tools for successfully navigating our technological future.”
There is a lot of information and artwork in this factually dense, but very readable, science book. I learned more about the brain and its evolution then I have done in reading any other book about this area of science. If your time is limited in reading about the recent discoveries and science about the brain, this is the one book I recommend to get. It pulls together all of the bits of information I have read here and there in other science and AI books. He explains with an in-depth synthesis for the general reader. The author focuses primarily on evolution and function of the brain, less on how computer scientists are transferring the skills of the brain to AI algorithms on the computer. For that angle, I recommend The Alignment Problem: Machine Learning and Human Values by Brian Christian.
Included are a Glossary, extensive Notes, Bibliography and Index sections, plus Art, Photo, and Figure credits in the back of the book.
March 4, 2025
This paragraph hit pretty hard:
Herein lies both the tragedy and beauty of humanity. We are indeed some of the most altruistic animals, but we may have paid the price for this altruism with our darker side: our instinct to punish those who we deem to be moral violators; our reflexive delineation of people into good and evil; our desperation to conform to our in-group and the ease with which we demonize those in the out-group. And with these new traits, empowered by our newly enlarged brains and cumulative language, the human instinct for politics—derived from our ancestral primates—was no longer a little trick for climbing social hierarchies but a cudgel of coordinated conquest. All this is the inevitable result of a survival niche requiring high levels of altruism between unrelated individuals.
March 4, 2025
Y una vez más gracias a la plataforma #masacritica de Babelio por darme la oportunidad de leer una obra de lo más interesante. Yo diría que en este caso, más que nunca. Y a la editorial tendencias por el envío de este ejemplar.
En mi casa tenemos un robot Roomba para que vaya por todo el piso aspirando la suciedad del suelo. La llamamos cariñosamente Paquita (por lo de pa’quitar la mierda) y sé que es un dato Random que de fijo que os interesa muchísimo y que gracias a él he protagonizado el inicio de reseña más estúpido de toda la historia. La gracia está en que este pequeño aparato es hoy por hoy lo más parecido a un robot funcional que hay en el mercado, y estamos hablando de una industria que lleva trabajando años y años por conseguir hacer avanzar la robótica hasta crear entes informáticos lo más parecido posibles a los humanos. Y solo una pequeña muestra de hasta qué punto la IA está lenta implacablemente integrándose en nuestra vida cotidiana. Lleva muchos años ya presente en ella, y no solo en las roombas, también en aplicaciones tan usadas como el Google Maps o el TomTom, los correos electrónicos o las diferentes aplicaciones que nos dan recomendaciones de música, series y películas que disfrutar en nuestro tiempo libre.
“Una Historia de la Inteligencia” es una mezcla de biología, historia de la evolución, neuro ciencia, psicología y tecnología que nos enseña hasta qué punto los avances tecnológicos beben de el camino que la raza humana ha estado siguiendo a lo largo de los milenios que ha supuesto la evolución. Max S. Benett nos habla de cerebro humano, esa maquinaria fascinante y llena aún de misterios que todos tenemos, como hilo conductor para meditar sobre los avances que se han vivido, se vivirán y se están buscando en el campo de la IA, sobre un futuro y un presente que ya va a ser parte, para bien o para mal, del devenir humano. El autor nos propone un recorrido de milenios que va desde el origen de la vida hasta el presente, explorando los cinco eventos clave que han dado forma al cerebro. Enfatizando en cada uno de ellos y analizándolos, nos va enseñando cómo ha ido evolucionando con el paso del tiempo y la forma en que la IA debe buscar por lo menos equipararse a él para conseguir crear robot y herramientas avanzados. Lo cual permite que el autor nos explique cuáles son los retos a los que la tecnología y la informática se está enfrentando para poder conseguir estos avances.
El tema es interesantísimo y fascinante. Pero reconozco que a mí habido muchos momentos en los que la lectura me ha costado mucho. Te lo venden como una manera muy amena de introducirte en campos como la neurociencia, la biología o la tecnología, como si fuera una lectura ideal para personas totalmente profanas respecto a todos los temas que toca. Pero ha habido muchos capítulos y momentos (especialmente al principio del libro) que me han parecido muy difíciles de leer y de entender, a los que he tenido que dedicar mucho tiempo para poder comprenderlos mínimamente. Pienso que manejan muchos conceptos para los que creo que si que se necesitan un mínimo de conocimientos previos. Desde luego no me parece el libro tan accesible que nos venden en la contraportada. Pero aún así, si se le dedica el tiempo necesario y cada cada uno lo lee a su ritmo, al final la lectura te atrapa totalmente y aprendes mucho con ella. Una de las razones por las que la elegí para #masacritica de Babelio fue porque me pareció un tema fascinante del que apenas se nada, una lectura de esas que me ayudarían a salir de mi zona de confort y adquirir conocimientos que siendo de letras puras puras nunca he estudiado o lo he hecho muy por encima en el colegio y algunas asignaturas de la carrera. Y desde luego que lo he conseguido totalmente. Además, es cierto que hay muchos conceptos difíciles, pero el autor maneja todo esto de una manera bastante amena y didáctica que facilita mucho las cosas. Y que ilustra hasta qué punto. La evolución humana es un camino a dos bandas. El cerebro es el modelo y el profesor a la hora de crear inteligencia artificial, pero a la vez la IA puede enseñarnos mucho de cómo funciona, cuáles son los mecanismos que lo rigen y hasta qué punto está conectado con todos los aspectos de nuestra vida.
Últimamente leo por redes que se habla de la IA como un monstruo que pretende terminar con nuestro mundo, controlar a las masas y crear una sociedad fría y estandarizada, donde no haya lugar para la imaginación, la creatividad o la libertad. Personalmente yo creo que este avance Acaba de llegar, como quien dice, a otras vidas. Y que la humanidad aún está intentando aclimatarse a él, mientras esté busca su lugar en la cultura. Pero creo que es inevitable que la IA vaya formando, cada vez más, parte de nuestras existencias. El momento histórico que estamos viviendo puede compararse con lo que supuso la revolución industrial en el siglo XVIII. Se crearon nuevos puestos de trabajo a la luz de las nuevas tecnologías industriales, mientras que otros quedaron desfasados por no tener cabida en la nueva sociedad económica y cultural, que estaba apareciendo. Muchas personas perdieron su trabajo, y lo único que les quedó fue dejar que el paso de la evolución les llevara por delante, o unirse a este nuevo tren, y esto creo que es lo que ilustra el libro con bastante eficacia, además de la relación entre tecnología y cerebro. Max S. Bennett es ceo de una empresa que proporciona tecnologías IA a las mayores empresas del mundo. En ningún momento busca hablarnos de decisiones morales, y de todo lo que conlleva la IA a nivel filosófico. Su intención es solo relacionar el cerebro con la tecnología, mostrar como el uno y la otra han ido evolucionando y como la segunda debe tratar de emular a la primera. Lo suyo es darnos datos y más datos. Y bucear en los muchos datos y misterios que tenemos sobre qué es y cómo funciona la inteligencia a secas.
La máquina humana más compleja, difícil de emular y fascinante, no la ha creado ninguna persona. Porque es el ser humano en si mismo. Con todas sus complejidades, misterios y avances que ha conseguido a lo largo de la historia. Es con eso con lo que más me quedo de esta lectura.
En mi casa tenemos un robot Roomba para que vaya por todo el piso aspirando la suciedad del suelo. La llamamos cariñosamente Paquita (por lo de pa’quitar la mierda) y sé que es un dato Random que de fijo que os interesa muchísimo y que gracias a él he protagonizado el inicio de reseña más estúpido de toda la historia. La gracia está en que este pequeño aparato es hoy por hoy lo más parecido a un robot funcional que hay en el mercado, y estamos hablando de una industria que lleva trabajando años y años por conseguir hacer avanzar la robótica hasta crear entes informáticos lo más parecido posibles a los humanos. Y solo una pequeña muestra de hasta qué punto la IA está lenta implacablemente integrándose en nuestra vida cotidiana. Lleva muchos años ya presente en ella, y no solo en las roombas, también en aplicaciones tan usadas como el Google Maps o el TomTom, los correos electrónicos o las diferentes aplicaciones que nos dan recomendaciones de música, series y películas que disfrutar en nuestro tiempo libre.
“Una Historia de la Inteligencia” es una mezcla de biología, historia de la evolución, neuro ciencia, psicología y tecnología que nos enseña hasta qué punto los avances tecnológicos beben de el camino que la raza humana ha estado siguiendo a lo largo de los milenios que ha supuesto la evolución. Max S. Benett nos habla de cerebro humano, esa maquinaria fascinante y llena aún de misterios que todos tenemos, como hilo conductor para meditar sobre los avances que se han vivido, se vivirán y se están buscando en el campo de la IA, sobre un futuro y un presente que ya va a ser parte, para bien o para mal, del devenir humano. El autor nos propone un recorrido de milenios que va desde el origen de la vida hasta el presente, explorando los cinco eventos clave que han dado forma al cerebro. Enfatizando en cada uno de ellos y analizándolos, nos va enseñando cómo ha ido evolucionando con el paso del tiempo y la forma en que la IA debe buscar por lo menos equipararse a él para conseguir crear robot y herramientas avanzados. Lo cual permite que el autor nos explique cuáles son los retos a los que la tecnología y la informática se está enfrentando para poder conseguir estos avances.
El tema es interesantísimo y fascinante. Pero reconozco que a mí habido muchos momentos en los que la lectura me ha costado mucho. Te lo venden como una manera muy amena de introducirte en campos como la neurociencia, la biología o la tecnología, como si fuera una lectura ideal para personas totalmente profanas respecto a todos los temas que toca. Pero ha habido muchos capítulos y momentos (especialmente al principio del libro) que me han parecido muy difíciles de leer y de entender, a los que he tenido que dedicar mucho tiempo para poder comprenderlos mínimamente. Pienso que manejan muchos conceptos para los que creo que si que se necesitan un mínimo de conocimientos previos. Desde luego no me parece el libro tan accesible que nos venden en la contraportada. Pero aún así, si se le dedica el tiempo necesario y cada cada uno lo lee a su ritmo, al final la lectura te atrapa totalmente y aprendes mucho con ella. Una de las razones por las que la elegí para #masacritica de Babelio fue porque me pareció un tema fascinante del que apenas se nada, una lectura de esas que me ayudarían a salir de mi zona de confort y adquirir conocimientos que siendo de letras puras puras nunca he estudiado o lo he hecho muy por encima en el colegio y algunas asignaturas de la carrera. Y desde luego que lo he conseguido totalmente. Además, es cierto que hay muchos conceptos difíciles, pero el autor maneja todo esto de una manera bastante amena y didáctica que facilita mucho las cosas. Y que ilustra hasta qué punto. La evolución humana es un camino a dos bandas. El cerebro es el modelo y el profesor a la hora de crear inteligencia artificial, pero a la vez la IA puede enseñarnos mucho de cómo funciona, cuáles son los mecanismos que lo rigen y hasta qué punto está conectado con todos los aspectos de nuestra vida.
Últimamente leo por redes que se habla de la IA como un monstruo que pretende terminar con nuestro mundo, controlar a las masas y crear una sociedad fría y estandarizada, donde no haya lugar para la imaginación, la creatividad o la libertad. Personalmente yo creo que este avance Acaba de llegar, como quien dice, a otras vidas. Y que la humanidad aún está intentando aclimatarse a él, mientras esté busca su lugar en la cultura. Pero creo que es inevitable que la IA vaya formando, cada vez más, parte de nuestras existencias. El momento histórico que estamos viviendo puede compararse con lo que supuso la revolución industrial en el siglo XVIII. Se crearon nuevos puestos de trabajo a la luz de las nuevas tecnologías industriales, mientras que otros quedaron desfasados por no tener cabida en la nueva sociedad económica y cultural, que estaba apareciendo. Muchas personas perdieron su trabajo, y lo único que les quedó fue dejar que el paso de la evolución les llevara por delante, o unirse a este nuevo tren, y esto creo que es lo que ilustra el libro con bastante eficacia, además de la relación entre tecnología y cerebro. Max S. Bennett es ceo de una empresa que proporciona tecnologías IA a las mayores empresas del mundo. En ningún momento busca hablarnos de decisiones morales, y de todo lo que conlleva la IA a nivel filosófico. Su intención es solo relacionar el cerebro con la tecnología, mostrar como el uno y la otra han ido evolucionando y como la segunda debe tratar de emular a la primera. Lo suyo es darnos datos y más datos. Y bucear en los muchos datos y misterios que tenemos sobre qué es y cómo funciona la inteligencia a secas.
La máquina humana más compleja, difícil de emular y fascinante, no la ha creado ninguna persona. Porque es el ser humano en si mismo. Con todas sus complejidades, misterios y avances que ha conseguido a lo largo de la historia. Es con eso con lo que más me quedo de esta lectura.
August 15, 2025
absolutely fantastic
Extremely well written and perfectly structured, this book is one of the top experiences of 2025 for me - a must read if you want to understand where human intelligence came from and how it relates to the AI
Extremely well written and perfectly structured, this book is one of the top experiences of 2025 for me - a must read if you want to understand where human intelligence came from and how it relates to the AI
Did not finish
October 3, 2024DNF. Unfortunately, I couldn't get into this one. The writing was just a little too technical to hold my interest.
December 13, 2025
This book is so good that I lack the intelligence to describe it well. I also do not want to use AI to describe it. The book describes both.
July 14, 2025
This book was brilliant. Bennett aims to create a kind of grand narrative of how human intelligence formed throughout evolution, starting from lifeless material, to DNA, to single and multi-celled organisms, all the way through the bilaterians, vertebrates, mammals, primates, and finally, to the sapiens lineage. With each breakthrough of intelligence described, parallels are drawn to the corresponding machine learning models that mirror that feature of animal intelligence. The result is absurdly illuminating. I've read a fair amount on both neuroscience and artificial intelligence before, but the red lines drawn between them here through the lens of evolution are so, so cool. I learned so much, and loved pretty much every page.
In addition to learning more and creating cross-disciplinary connections I hadn't before, reading such an overview of the evolutionary history of intelligence also doubled down as a wonderful meditation on my anthropocentric views of the world, of humans, and of myself. Failing to find the point in the story at which inanimate matter recognizably becomes life - becomes *me* - is quite unsettling, while also eye-opening in a refreshing way. I feel much closer to nature after reading this book.
Another review mentioned how Bennett curiously does not define intelligence at any point of the book, and I would agree with this reviewer that this choice is to the benefit of the narrative told. Paradoxically, intelligence isn't really...anything. The book describes the unfolding of an unthinking process: whether us humans choose to call it intelligence, and at which point it qualifies as such, becomes quite arbitrary.
Bennett points this out in the introduction as well, by making sure that the reader knows this book is a purely descriptive story of specifically human intelligence, and not a claim on the *superiority* of that specific type of intelligence. Beautiful quote from there: "Species fall into different survival niches, each of which optimizes for different things. Many niches - in fact, most niches - are better served by smaller and simpler brains (or no brains at all). Big-brained apes are the result of a different survival strategy than that of worms, bacteria, or butterflies. But none are "better." In the eyes of evolution, the hierarchy has only two rungs: on one, there are those who survived, and on the other, those that did not."
In addition, somehow Bennet is able to dodge a major hurdle faced by "grand narrative" books: overly simplifying in the name of preserving the narrative. The narrative here seems to be quite self-aware of its strengths and weaknesses, and makes sure to steelman some opposing views whenever possible. Of course, specifically *evolutionary* neuroscience is not something I'd consider myself an expert in, and so might not be the best judge of how balanced the arguments in the book were: but, at least most times I raised an eyebrow at something presented, thinking "well, it's probably more complex than that," Bennett would address my worry quite promptly after. I loved that!
Please read this book so we can yap about it: I'll happily borrow my copy. :)
In addition to learning more and creating cross-disciplinary connections I hadn't before, reading such an overview of the evolutionary history of intelligence also doubled down as a wonderful meditation on my anthropocentric views of the world, of humans, and of myself. Failing to find the point in the story at which inanimate matter recognizably becomes life - becomes *me* - is quite unsettling, while also eye-opening in a refreshing way. I feel much closer to nature after reading this book.
Another review mentioned how Bennett curiously does not define intelligence at any point of the book, and I would agree with this reviewer that this choice is to the benefit of the narrative told. Paradoxically, intelligence isn't really...anything. The book describes the unfolding of an unthinking process: whether us humans choose to call it intelligence, and at which point it qualifies as such, becomes quite arbitrary.
Bennett points this out in the introduction as well, by making sure that the reader knows this book is a purely descriptive story of specifically human intelligence, and not a claim on the *superiority* of that specific type of intelligence. Beautiful quote from there: "Species fall into different survival niches, each of which optimizes for different things. Many niches - in fact, most niches - are better served by smaller and simpler brains (or no brains at all). Big-brained apes are the result of a different survival strategy than that of worms, bacteria, or butterflies. But none are "better." In the eyes of evolution, the hierarchy has only two rungs: on one, there are those who survived, and on the other, those that did not."
In addition, somehow Bennet is able to dodge a major hurdle faced by "grand narrative" books: overly simplifying in the name of preserving the narrative. The narrative here seems to be quite self-aware of its strengths and weaknesses, and makes sure to steelman some opposing views whenever possible. Of course, specifically *evolutionary* neuroscience is not something I'd consider myself an expert in, and so might not be the best judge of how balanced the arguments in the book were: but, at least most times I raised an eyebrow at something presented, thinking "well, it's probably more complex than that," Bennett would address my worry quite promptly after. I loved that!
Please read this book so we can yap about it: I'll happily borrow my copy. :)
September 6, 2025
this was absolutely brilliant!
i know already that i'll be re-reading this many times in the coming years.
this book tackles the evolution of intelligence from simple neurons in worms to the complex brains of modern humans.
5 breakthoughs - steering, reinforceing, simulating, mentalizing, and speaking.
also the developments regarding ai are observed with the context of brain evolution.
so i heard about Moravec's paradox before. i thought it was interesting but didn't really grok it at that time.
it's basically this
> Moravec's paradox is the phenomenon where tasks that are easy for humans, such as perception and mobility, are difficult for machines to replicate, while tasks that are difficult for humans, like complex calculations and data analysis, are relatively easy for AI
after reading this book, now i understand the why in a better way. it also made me understand and appreciate the differences between various brains.
also there were some really interesting things i didn't know before
- curiosity evolved to solve the explore-exploit dilemma
- selection pressures from scary dinosaurs forced our mammal ancestors to develop the ability to simulate the world in their neocortex
- sutton's actor-critic system is present in the brain (cool cross over from neuroscience to computing)
- rats can't do grocery shopping but monkeys can
- ants don't have spacial intelligence but fish do
...
finally, this book also inspired me to pick up and resume Sapolsky's magnum opus Behave.
spoiler alert: i finished it!!
i know already that i'll be re-reading this many times in the coming years.
this book tackles the evolution of intelligence from simple neurons in worms to the complex brains of modern humans.
5 breakthoughs - steering, reinforceing, simulating, mentalizing, and speaking.
also the developments regarding ai are observed with the context of brain evolution.
so i heard about Moravec's paradox before. i thought it was interesting but didn't really grok it at that time.
it's basically this
> Moravec's paradox is the phenomenon where tasks that are easy for humans, such as perception and mobility, are difficult for machines to replicate, while tasks that are difficult for humans, like complex calculations and data analysis, are relatively easy for AI
after reading this book, now i understand the why in a better way. it also made me understand and appreciate the differences between various brains.
also there were some really interesting things i didn't know before
- curiosity evolved to solve the explore-exploit dilemma
- selection pressures from scary dinosaurs forced our mammal ancestors to develop the ability to simulate the world in their neocortex
- sutton's actor-critic system is present in the brain (cool cross over from neuroscience to computing)
- rats can't do grocery shopping but monkeys can
- ants don't have spacial intelligence but fish do
...
finally, this book also inspired me to pick up and resume Sapolsky's magnum opus Behave.
spoiler alert: i finished it!!
December 25, 2023
A great book to read following Fei Fei Li’s memoir and The Alignment Problem.
- “The better we are at understanding our own minds, the better equipped we are to create artificial minds in our image”
- Language is the last breakthrough in terms of human intelligence, built on top of steering, reinforcing, simulating and mentalizing. LLMs today are language models built without simulating and mentalizing, hence it still lacks a true world model and it fails (sometimes) to produce answers that align with us.
- The book actually glosses over technical details (would recommend The Alignment Problem for that)
- “The better we are at understanding our own minds, the better equipped we are to create artificial minds in our image”
- Language is the last breakthrough in terms of human intelligence, built on top of steering, reinforcing, simulating and mentalizing. LLMs today are language models built without simulating and mentalizing, hence it still lacks a true world model and it fails (sometimes) to produce answers that align with us.
- The book actually glosses over technical details (would recommend The Alignment Problem for that)
November 25, 2024
The best book of science I have read this year. I read it bank to back as there was so much to learn here that one reading was not sufficient. This book is on evolution of brain in last 4 Billion years and at each stage what capacities it added fundamentally and they at each share where AI stands. This book gave me the same level of excitement that I get spending a lot of money to buy a material item that I aspire for some amount of time and the explained that too , curiosity releases Dopomine. Without the review of other members I would had not picked this and I am thankful for their reviews. I guess this book is the outcome of authors 30 plus years of trying to understand brain and the knowledge is communicated seamlessly. This book will remain the most gifted book to friends for the near term by me.
December 2, 2025
I feel like these big history books, a là Sapiens, always have the same structure. They start strong, with solid evidence and reasoning, and then slowly decay in quality to conjectures, opinions, and misleading arguments.
I really liked the bilaterian part and how the body plan gave rise to centralized brains. But then it gets to humans and, as always when talking about recent history, it gets sketchy.
He also brings up AI all the time, which I deeply disliked. Please keep AI away from my books, thanks. Especially since there is nothing original about it and doesn’t contribute much to the main story.
I really liked the bilaterian part and how the body plan gave rise to centralized brains. But then it gets to humans and, as always when talking about recent history, it gets sketchy.
He also brings up AI all the time, which I deeply disliked. Please keep AI away from my books, thanks. Especially since there is nothing original about it and doesn’t contribute much to the main story.
February 15, 2025
I appreciate the author’s ability to collect the work of so many in such an approachable way. But personally I just found the early evolutionary pieces quite boring and towards the end I didn’t feel like there were any “wow” moments especially in relation to AI.
I think you’d learn more about evolution and nature from something like “An Immense World” by Ed Young and you’d probably learn more about AI from asking AI.
I think you’d learn more about evolution and nature from something like “An Immense World” by Ed Young and you’d probably learn more about AI from asking AI.
January 30, 2025
author should have taken out the word brief. Went into far too much detail not just in the science but the extent to which he explained every single thing. Book could have been half the length and more articulate in getting across the main points. other than that, interesting book and i liked learning about how the brain developed
August 18, 2024
The author is an artificial Intelligence (AI) entrepreneur. His book, 'A Brief History of Intelligence' published in 2023, is a well written consolidation of the ongoing neuroscience information avalanche. This book is a fantastic resource for the non specialist even though new information and understanding of how the brain functions becomes available on a daily basis.
Large gaps remain in what modern AI can achieve. How is it possible that AI can beat a grandmaster at chess, but it can't effectively load a dishwasher? This book proves that understanding the arc of our brain's history can unlock the tools for successfully navigating our technological future.
All aspects of our brain's history and its functionality presented by the author, such as the various dopamine responses, the Cambrian explosion, how the neuro functions of animals like invertebrates and fish differentiate between what's good and bad for the individual and respond differently to these two information inputs, the benefits of being curious capable, differences between the reptile and mammal brains, and much more are exceedingly informative. However, my curiosity was especially appeased by the sections focused on the mammals' advanced neocortex part of the brain.
It's worth knowing that birds seem to have evolved their own neocortex-like structure. But it doesn't exist for other animals like invertebrates, fish, amphibians, and reptiles.
Our neocortex is 2 to 4 millimeters thick and is significantly folded to fit inside the cranium. For an adult human, the neocortex unfolded is about 3 square feet or the size of a small desk. Names for regions of the neocortex, like the visual cortex located in the back of the neocortex, drop the neo prefix.
Surprisingly, under a microscope the neocortex looks identical everywhere and it looks relatively the same from one mammal species to another. The cortical sheet is a bunch of neocortical columns packed densely together which are repeating and duplicated micro circuits. The human neocortex is made up of over 10 billion neurons and trillions of connections. Rather than the entire structure's outrageous complexity, it's possible in the future, we need only to focus learning and understanding on the function of this micro circuit that is repeated a million, or so, times.
Each neocortical column does exactly the same thing. The only difference between regions of the neocortex is the input they receive and where they send the output. As you know, for example, the visual cortex of congenitally blind people does not become a functionally useless region. Instead, it has becomes repurposed and responsive to a multitude of other sensory input like hearing and touch. Areas of the neocortex seem interchangeable.
Due to our neocortex, we have peculiar properties of perception. For example: 1) Our neocortex fills in to substitute for missing or noisy visual or auditory input data, 2) Our minds like to have an interpretation that explains sensory input. Therefore, we are very sensitive to suggestions because we can't unsee, 3) At a noisy cocktail party, we can only perceive one conversation at a time. Though, we are capable of switching our attention to a different conversation, 4) We cannot imagine things and recognize things simultaneously. For example, we cannot read a book and imagine having breakfast at the same time, 5) Dreaming and imagination are the same general neuroprocess. When people are imagining things, their pupils dilate as the brain stops processing actual visual data and temporarily become pseudo blind.
The neocortex, in generation mode, can imagine things they are not currently experiencing. The imagined thing is a simulated reality in our neocortex.
Some neuroscientists refer to perception, even when it's functioning properly, as a constrained hallucination. Without sensory input, this hallucination becomes unconstrained. Our neocortex continuously strives to match our inner hallucination of reality to the sensory information we are seeing. It wants to predict things before they happen by continuously comparing the actual sensory data with the data predicted by its simulation.
As an example, our brain can generalize, impressively. If we look at two different, hand-written pictures of sevens, our neocortex learns without supervision by capturing the essential features of the different pictures of sevens. Human perception optimizes input data to learn without being told the right answer.
This book opens the door to new ways of understanding and thinking.
Large gaps remain in what modern AI can achieve. How is it possible that AI can beat a grandmaster at chess, but it can't effectively load a dishwasher? This book proves that understanding the arc of our brain's history can unlock the tools for successfully navigating our technological future.
All aspects of our brain's history and its functionality presented by the author, such as the various dopamine responses, the Cambrian explosion, how the neuro functions of animals like invertebrates and fish differentiate between what's good and bad for the individual and respond differently to these two information inputs, the benefits of being curious capable, differences between the reptile and mammal brains, and much more are exceedingly informative. However, my curiosity was especially appeased by the sections focused on the mammals' advanced neocortex part of the brain.
It's worth knowing that birds seem to have evolved their own neocortex-like structure. But it doesn't exist for other animals like invertebrates, fish, amphibians, and reptiles.
Our neocortex is 2 to 4 millimeters thick and is significantly folded to fit inside the cranium. For an adult human, the neocortex unfolded is about 3 square feet or the size of a small desk. Names for regions of the neocortex, like the visual cortex located in the back of the neocortex, drop the neo prefix.
Surprisingly, under a microscope the neocortex looks identical everywhere and it looks relatively the same from one mammal species to another. The cortical sheet is a bunch of neocortical columns packed densely together which are repeating and duplicated micro circuits. The human neocortex is made up of over 10 billion neurons and trillions of connections. Rather than the entire structure's outrageous complexity, it's possible in the future, we need only to focus learning and understanding on the function of this micro circuit that is repeated a million, or so, times.
Each neocortical column does exactly the same thing. The only difference between regions of the neocortex is the input they receive and where they send the output. As you know, for example, the visual cortex of congenitally blind people does not become a functionally useless region. Instead, it has becomes repurposed and responsive to a multitude of other sensory input like hearing and touch. Areas of the neocortex seem interchangeable.
Due to our neocortex, we have peculiar properties of perception. For example: 1) Our neocortex fills in to substitute for missing or noisy visual or auditory input data, 2) Our minds like to have an interpretation that explains sensory input. Therefore, we are very sensitive to suggestions because we can't unsee, 3) At a noisy cocktail party, we can only perceive one conversation at a time. Though, we are capable of switching our attention to a different conversation, 4) We cannot imagine things and recognize things simultaneously. For example, we cannot read a book and imagine having breakfast at the same time, 5) Dreaming and imagination are the same general neuroprocess. When people are imagining things, their pupils dilate as the brain stops processing actual visual data and temporarily become pseudo blind.
The neocortex, in generation mode, can imagine things they are not currently experiencing. The imagined thing is a simulated reality in our neocortex.
Some neuroscientists refer to perception, even when it's functioning properly, as a constrained hallucination. Without sensory input, this hallucination becomes unconstrained. Our neocortex continuously strives to match our inner hallucination of reality to the sensory information we are seeing. It wants to predict things before they happen by continuously comparing the actual sensory data with the data predicted by its simulation.
As an example, our brain can generalize, impressively. If we look at two different, hand-written pictures of sevens, our neocortex learns without supervision by capturing the essential features of the different pictures of sevens. Human perception optimizes input data to learn without being told the right answer.
This book opens the door to new ways of understanding and thinking.
July 6, 2025
2025/7 Update:
Upon a second reading, I better understand why I find this book so enjoyable. Beyond the reasons I mentioned in my previous review, the "five breakthroughs" framework provides readers with a satisfying perspective for understanding both the brain's inner structure and the long evolutionary history from single-cell life to humans.
I can also pinpoint where my perspective diverges from the author’s. Instead of language, I see the ability to accumulate knowledge across generations as the fifth and most recent breakthrough in intelligence. This view is influenced by my earlier reading of The Secret of Our Success by Joseph Henrich. I noticed that Max Bennett rarely cites work by cultural evolutionists like Henrich, which may account for our differing viewpoints.
To elaborate on my position: while language vastly enhances the ability to accumulate knowledge, knowledge accumulation across generations appears to have begun before language emerged. As early as 2.5 million years ago, human ancestors were using Oldowan stone tools, which I believe were more sophisticated than tools used by any other animals. However, consistent evidence of symbolic behavior—which could indicate language—only dates back about 100,000 years. Even if early Homo erectus had already developed language, a 500,000-year gap remains between sophisticated tool use and language use.
Furthermore, Max Bennett mentions that human children over-imitate compared to chimpanzees, which he argues helps knowledge accumulation. But this over-imitation instinct doesn't appear to depend on language.
I also wish the book drew more parallels between human intelligence and the latest AI developments, particularly the large language models. Max Bennett notes GPT-4's improved performance on common-sense and theory-of-mind questions but argues this improvement doesn't result from breakthroughs in simulation and mentalizing. While I agree with his position, I would like to see more analysis supporting this conclusion.
------
This book is absolutely fascinating—it offers a compelling synthesis of natural and artificial intelligence by drawing insightful comparisons between the two. For instance, by learning about the role of dopamine in animals' reinforcement learning, I was finally able to grasp how AlphaGo works on an intuitive level—and that was only the second of the five breakthroughs the book sets out to explain!
Of course, there isn’t always a one-to-one correspondence between the brain and a machine. For example, we still don’t know how to make machines develop a theory of mind. But whenever Max Bennett draws parallels between the two worlds, the connections are consistently thought-provoking. The first striking example is the importance of steering in the development of both natural and artificial intelligence—just as steering proves to be the breakthrough that has brought about all the following development in natural intelligence, the first commercial robot, the Roomba, achieved its success by equipping itself with the ability to steer.
The most beautiful connection, in my opinion, is how animals and AI solve the temporal credit assignment in the same way. In simple trial and error learning, an agent (either an animal or an AI) reinforces an action if it leads to a positive outcome, and inhibits it vice versa. In more complex problems that involves many steps, however, agents face the temporal credit assignment problem. If it simply reinforces all the prior moves before winning, it will take agent forever to learn the correct steps: multiple-steps problems have way too many variations. It turns out animal and AlphaGo solve the problem in the same way, by reinforcing actions not by its final reward, but by predicted rewards. Moreover, computer scientist and neuroscientist inspired each other on the discovery of this mechanism. Computer scientists invented a RL algorithm that wins backgammon against human by hypothesizing how animals' brains work, and the breakthrough inspired neuroscientists to find out the actual way animal solve problems.
I also admire the book for its structure. By presenting intelligence through five key breakthroughs, the author provides readers with a clear, hierarchical framework for understanding it. When we talk about intelligence, we often only think of logic and reasoning, math and art. However, as Yann LeCun has famously noted, to truly grasp human-level intelligence, we must first study simpler forms of intelligence.
The book illustrates this by exploring fundamental mental mechanisms, such as excitement and satiation, which enable animals—including humans—to solve a wide range of problems and adapt to their environments. To illustrate this, the author draws attention to the nematode. Despite having only 302 neurons—an astonishing contrast to the 85 billion neurons in the human brain—this tiny organism exhibits remarkably sophisticated behavior. I was surprised to learn that nematodes employ multiple strategies to form associations, and even more astounded to discover they can experience mental states akin to human stress and depression.
For an animal to function, all levels of intelligence need to work. The author introduces patient L, who retains the ability to understand and speak yet lost all intentions to move, speak, and care about anything, due to damage in the anterior prefrontal cortex (aPFC), an ancient area evolved in the first mammals. Her condition highlights the importance of intelligence mechanisms more primitive than reasoning and language.
Furthermore, these simpler mental mechanisms lay the foundation for more complex cognitive processes. For example, reinforcement learning builds upon basic valence signals of good and bad, while simulation is only possible due to pre-existing learning within the basal ganglia. Furthermore, intention stems from the ability to simulate one's own actions. Max Bennett identifies this elegant bootstrapping. By directing readers’ attention to it, he deepens their appreciation for the brilliance of evolution and clarifies the intricate relationships between different forms of intelligence.
One of the most unexpected aspects of this book is the sheer range of topics it covers. The extensive list includes emotion, stress, depression, curiosity, gambling, hallucination, dreams, sleep, intention, habits, imitation—and possibly even consciousness. Why does a book about intelligence delve into these subjects?
In retrospect, my surprise stemmed from a human-centric understanding of the world. This perspective tends to focus on how mental states such as emotions and stress are experienced and interpreted within individual lives and cultures. However, these states are not exclusively human phenomena; they are evolutionary products that emerged long before humans. Naturally, they belong in a discussion about intelligence, since intelligence itself is fundamentally about adaptation—and these mental states help animals adapt to their environments.
Because the human-centric perspective is so deeply ingrained, Frans de Waal had to devote an entire book, Mama’s Last Hug, to convincing readers that animals experience emotions. After reading this book, I gained a level of clarity I had never reached before. Of course animals have emotions! Even nematodes, despite their tiny neural networks, possess dopamine and serotonin and exhibit emotional behaviors.
This is a dense book, and I prepare to re-read it, such that I can better understand all the ideas presented. I look forward to the re-reading experience and updating this review later!
Side Note:
• The original arts by Rebecca Gelernter are informative and beautiful.
• Daniel Dennett introduced a comparable yet distinct hierarchy of intelligence in his book Kinds of Minds, but that framework didn’t enhance my understanding as much as Max Bennett’s five breakthroughs.
(edited with AI assistance)
Upon a second reading, I better understand why I find this book so enjoyable. Beyond the reasons I mentioned in my previous review, the "five breakthroughs" framework provides readers with a satisfying perspective for understanding both the brain's inner structure and the long evolutionary history from single-cell life to humans.
I can also pinpoint where my perspective diverges from the author’s. Instead of language, I see the ability to accumulate knowledge across generations as the fifth and most recent breakthrough in intelligence. This view is influenced by my earlier reading of The Secret of Our Success by Joseph Henrich. I noticed that Max Bennett rarely cites work by cultural evolutionists like Henrich, which may account for our differing viewpoints.
To elaborate on my position: while language vastly enhances the ability to accumulate knowledge, knowledge accumulation across generations appears to have begun before language emerged. As early as 2.5 million years ago, human ancestors were using Oldowan stone tools, which I believe were more sophisticated than tools used by any other animals. However, consistent evidence of symbolic behavior—which could indicate language—only dates back about 100,000 years. Even if early Homo erectus had already developed language, a 500,000-year gap remains between sophisticated tool use and language use.
Furthermore, Max Bennett mentions that human children over-imitate compared to chimpanzees, which he argues helps knowledge accumulation. But this over-imitation instinct doesn't appear to depend on language.
I also wish the book drew more parallels between human intelligence and the latest AI developments, particularly the large language models. Max Bennett notes GPT-4's improved performance on common-sense and theory-of-mind questions but argues this improvement doesn't result from breakthroughs in simulation and mentalizing. While I agree with his position, I would like to see more analysis supporting this conclusion.
------
This book is absolutely fascinating—it offers a compelling synthesis of natural and artificial intelligence by drawing insightful comparisons between the two. For instance, by learning about the role of dopamine in animals' reinforcement learning, I was finally able to grasp how AlphaGo works on an intuitive level—and that was only the second of the five breakthroughs the book sets out to explain!
Of course, there isn’t always a one-to-one correspondence between the brain and a machine. For example, we still don’t know how to make machines develop a theory of mind. But whenever Max Bennett draws parallels between the two worlds, the connections are consistently thought-provoking. The first striking example is the importance of steering in the development of both natural and artificial intelligence—just as steering proves to be the breakthrough that has brought about all the following development in natural intelligence, the first commercial robot, the Roomba, achieved its success by equipping itself with the ability to steer.
The most beautiful connection, in my opinion, is how animals and AI solve the temporal credit assignment in the same way. In simple trial and error learning, an agent (either an animal or an AI) reinforces an action if it leads to a positive outcome, and inhibits it vice versa. In more complex problems that involves many steps, however, agents face the temporal credit assignment problem. If it simply reinforces all the prior moves before winning, it will take agent forever to learn the correct steps: multiple-steps problems have way too many variations. It turns out animal and AlphaGo solve the problem in the same way, by reinforcing actions not by its final reward, but by predicted rewards. Moreover, computer scientist and neuroscientist inspired each other on the discovery of this mechanism. Computer scientists invented a RL algorithm that wins backgammon against human by hypothesizing how animals' brains work, and the breakthrough inspired neuroscientists to find out the actual way animal solve problems.
I also admire the book for its structure. By presenting intelligence through five key breakthroughs, the author provides readers with a clear, hierarchical framework for understanding it. When we talk about intelligence, we often only think of logic and reasoning, math and art. However, as Yann LeCun has famously noted, to truly grasp human-level intelligence, we must first study simpler forms of intelligence.
The book illustrates this by exploring fundamental mental mechanisms, such as excitement and satiation, which enable animals—including humans—to solve a wide range of problems and adapt to their environments. To illustrate this, the author draws attention to the nematode. Despite having only 302 neurons—an astonishing contrast to the 85 billion neurons in the human brain—this tiny organism exhibits remarkably sophisticated behavior. I was surprised to learn that nematodes employ multiple strategies to form associations, and even more astounded to discover they can experience mental states akin to human stress and depression.
For an animal to function, all levels of intelligence need to work. The author introduces patient L, who retains the ability to understand and speak yet lost all intentions to move, speak, and care about anything, due to damage in the anterior prefrontal cortex (aPFC), an ancient area evolved in the first mammals. Her condition highlights the importance of intelligence mechanisms more primitive than reasoning and language.
Furthermore, these simpler mental mechanisms lay the foundation for more complex cognitive processes. For example, reinforcement learning builds upon basic valence signals of good and bad, while simulation is only possible due to pre-existing learning within the basal ganglia. Furthermore, intention stems from the ability to simulate one's own actions. Max Bennett identifies this elegant bootstrapping. By directing readers’ attention to it, he deepens their appreciation for the brilliance of evolution and clarifies the intricate relationships between different forms of intelligence.
One of the most unexpected aspects of this book is the sheer range of topics it covers. The extensive list includes emotion, stress, depression, curiosity, gambling, hallucination, dreams, sleep, intention, habits, imitation—and possibly even consciousness. Why does a book about intelligence delve into these subjects?
In retrospect, my surprise stemmed from a human-centric understanding of the world. This perspective tends to focus on how mental states such as emotions and stress are experienced and interpreted within individual lives and cultures. However, these states are not exclusively human phenomena; they are evolutionary products that emerged long before humans. Naturally, they belong in a discussion about intelligence, since intelligence itself is fundamentally about adaptation—and these mental states help animals adapt to their environments.
Because the human-centric perspective is so deeply ingrained, Frans de Waal had to devote an entire book, Mama’s Last Hug, to convincing readers that animals experience emotions. After reading this book, I gained a level of clarity I had never reached before. Of course animals have emotions! Even nematodes, despite their tiny neural networks, possess dopamine and serotonin and exhibit emotional behaviors.
This is a dense book, and I prepare to re-read it, such that I can better understand all the ideas presented. I look forward to the re-reading experience and updating this review later!
Side Note:
• The original arts by Rebecca Gelernter are informative and beautiful.
• Daniel Dennett introduced a comparable yet distinct hierarchy of intelligence in his book Kinds of Minds, but that framework didn’t enhance my understanding as much as Max Bennett’s five breakthroughs.
(edited with AI assistance)
This entire review has been hidden because of spoilers.
October 9, 2024
This was a well researched book that at some point I felt that I needed to see the author's work product and go through what did not make it into the book. The author, Solomon Bennett, even said he set out to write a book he himself would have wanted to read on intelligence and the breakthroughs that made our brains. There are a lot of novel concepts here and even better built from first principles like a seemingly simple decision on how single celled organisms decide on where to move towards or away from and the decision process. This is the book to read if you want to know how our brains evolved.
For all it's good, I had thought the book will evenly tackle evolution and AI in fact I had picked it mostly for the AI part. Instead the book leaned heavily on brain evolution and AI concepts only got passing mentions. I also had qualms on how most of the concepts here were stated in the definitive. A lot of evolution concepts are speculative and educated guesses and that fact should always be stated, even brain research itself has speculative aspects and some concepts have diverging sometimes contradicting interpretations and this fact was not clearly brought out. Still, highly recommended especially for anyone curious on how the brain evolved.
For all it's good, I had thought the book will evenly tackle evolution and AI in fact I had picked it mostly for the AI part. Instead the book leaned heavily on brain evolution and AI concepts only got passing mentions. I also had qualms on how most of the concepts here were stated in the definitive. A lot of evolution concepts are speculative and educated guesses and that fact should always be stated, even brain research itself has speculative aspects and some concepts have diverging sometimes contradicting interpretations and this fact was not clearly brought out. Still, highly recommended especially for anyone curious on how the brain evolved.
February 28, 2026
I really enjoy A Brief History of Intelligence — it weaves evolution and neuroscience into the hottest topic of the day: artificial intelligence. It's made me more discerning when using LLMs, and more curious about observing how they respond to various questions in different contexts, and how my own "intelligence" works in contrast.
Even though today's AI can do and replace many jobs, there are fundamental differences between the human mind and machine. I still love the human mind more. As much as I find LLMs impressive at solving practical problems and eloquent in answering even difficult questions, The Brothers Karamazov and Beethoven's 7th were created by human beings who saw the abyss, lived through it, and made something profound and beautiful out of it. I'm not sure that's something a machine can replicate — or that it would mean the same thing if it did.
Even though today's AI can do and replace many jobs, there are fundamental differences between the human mind and machine. I still love the human mind more. As much as I find LLMs impressive at solving practical problems and eloquent in answering even difficult questions, The Brothers Karamazov and Beethoven's 7th were created by human beings who saw the abyss, lived through it, and made something profound and beautiful out of it. I'm not sure that's something a machine can replicate — or that it would mean the same thing if it did.
Displaying 1 - 30 of 598 reviews