What do you think?
Rate this book
The Laws of Thought: The Quest for a Mathematical Theory of the Mind
From the coauthor of
Algorithms to Live By
, an exploration of the quest to use mathematics to describe the ways we think, from its origins three hundred years ago to the ideas behind modern AI systems and the ways in which they still differ from human minds
Everyone has a basic understanding of how the physical world works. We learn about physics and chemistry in school, letting us explain the world around us in terms of concepts like force, acceleration, and gravity—the Laws of Nature. But we don’t have the same fluency with concepts needed to understand the world inside us—the Laws of Thought. While the story of how mathematics has been used to reveal the mysteries of the universe is familiar, the story of how it has been used to study the mind is not.
There is no one better to tell that story than Tom Griffiths, the head of Princeton’s AI Lab and a renowned expert in the field of cognitive science. In this groundbreaking book, he explains the three major approaches to formalizing thought—rules and symbols, neural networks, and probability and statistics—introducing each idea through the stories of the people behind it. As informed conversations about thought, language, and learning become ever more pressing in the age of AI, The Laws of Thought is an essential listen for anyone interested in the future of technology.
Everyone has a basic understanding of how the physical world works. We learn about physics and chemistry in school, letting us explain the world around us in terms of concepts like force, acceleration, and gravity—the Laws of Nature. But we don’t have the same fluency with concepts needed to understand the world inside us—the Laws of Thought. While the story of how mathematics has been used to reveal the mysteries of the universe is familiar, the story of how it has been used to study the mind is not.
There is no one better to tell that story than Tom Griffiths, the head of Princeton’s AI Lab and a renowned expert in the field of cognitive science. In this groundbreaking book, he explains the three major approaches to formalizing thought—rules and symbols, neural networks, and probability and statistics—introducing each idea through the stories of the people behind it. As informed conversations about thought, language, and learning become ever more pressing in the age of AI, The Laws of Thought is an essential listen for anyone interested in the future of technology.
400 pages, Hardcover
First published February 10, 2026
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 38 reviews
March 14, 2026
There is a book for everyone. This one was for me. A book review is a lens into a person's values/likes. I learned more about team members from their book preferences, and what they discussed, than most any other avenue.
The book is really about the evolution of thought on how the brain might work. I was familiar with most every concept, but enjoyed learning about the relationship between all the scientist/psychologist/mathematicians involved. Over the last 300 years, some ideas have remained popular, others have died, and yes people have wasted time being tested on ideas that turned out to be suspect.
I am just amazed at what we don't know. We know why planets go around the sun, why atoms attract, and can model these with enough fidelity to make accurate predictions, yet it seems with all our wisdom, the brain just might not be capable of being understood by the brain.
Many of the last chapters were on the current state of AI and how they certainly pass the Turing test and can mimic most human cognitive capabilities. But does mimicking mean neural nets are how the brain works? Jets can fly but tell us nothing about a bird.
The tension is that neural nets take so much information to train. The latest versions have pretty much read everything written. So perhaps this is not how the brain works. I have 2 theories. A) One second of vision has more data than an entire book of words. Watch mom talk for a few seconds, and you can learn much more than if you just wrote down her words. We need to train on vision, sound and touch. Lots of physics can be learned by watching a cup milk spill. Oh and human behavior. B) Neural nets start from scratch, with zero knowledge, only the ability to learn. Evolution over the last few billion years has certainly played a part in our current capabilities. So when you hear, oh this model took 2 weeks and x zillion dollars, well that compares to the last few billion year.
The best thing about the book is the books it referenced. The author has a text book, that looks amazing just a less pop science level.
Favorite quote, Geoffrey Hinton is one of the fathers of modern AI.
Young Geoffrey was sent to a British public school with a strong Christian ethos, which contrasted strongly with the ideology at home. He ultimately found this experience useful: “I think that was a very good preparation for being a scientist because I got used to the idea that at least half the people are completely wrong.”
The book is really about the evolution of thought on how the brain might work. I was familiar with most every concept, but enjoyed learning about the relationship between all the scientist/psychologist/mathematicians involved. Over the last 300 years, some ideas have remained popular, others have died, and yes people have wasted time being tested on ideas that turned out to be suspect.
I am just amazed at what we don't know. We know why planets go around the sun, why atoms attract, and can model these with enough fidelity to make accurate predictions, yet it seems with all our wisdom, the brain just might not be capable of being understood by the brain.
Many of the last chapters were on the current state of AI and how they certainly pass the Turing test and can mimic most human cognitive capabilities. But does mimicking mean neural nets are how the brain works? Jets can fly but tell us nothing about a bird.
The tension is that neural nets take so much information to train. The latest versions have pretty much read everything written. So perhaps this is not how the brain works. I have 2 theories. A) One second of vision has more data than an entire book of words. Watch mom talk for a few seconds, and you can learn much more than if you just wrote down her words. We need to train on vision, sound and touch. Lots of physics can be learned by watching a cup milk spill. Oh and human behavior. B) Neural nets start from scratch, with zero knowledge, only the ability to learn. Evolution over the last few billion years has certainly played a part in our current capabilities. So when you hear, oh this model took 2 weeks and x zillion dollars, well that compares to the last few billion year.
The best thing about the book is the books it referenced. The author has a text book, that looks amazing just a less pop science level.
Favorite quote, Geoffrey Hinton is one of the fathers of modern AI.
Young Geoffrey was sent to a British public school with a strong Christian ethos, which contrasted strongly with the ideology at home. He ultimately found this experience useful: “I think that was a very good preparation for being a scientist because I got used to the idea that at least half the people are completely wrong.”
February 25, 2026
Tom Griffiths did again! Wrote a book on a complicated messy topic in crystal clear captivating language so that you just can not put the book down until you're done. Bravo!
Read
April 27, 2026"The Laws of Thought: The Quest for a Mathematical Theory of the Mind" by Tom Griffiths explores one of the most ambitious intellectual pursuits in human history: understanding whether the way we think can be described using mathematical principles. Drawing from centuries of philosophy, psychology, and computer science, the book traces how thinkers have tried to uncover the hidden structure behind human reasoning, learning, and perception. It shows that while no single framework fully explains the mind, combining multiple approaches brings us closer to understanding how thought actually works.
The journey begins with early efforts to formalize reasoning through logic. Philosophers like Aristotle laid the groundwork with syllogisms - structured arguments where conclusions follow from premises. Later thinkers, including Gottfried Wilhelm Leibniz, dreamed of turning reasoning itself into a kind of calculation. This vision took a major leap forward with George Boole, who showed that logical statements could be expressed using algebra with binary variables - essentially reducing thought to operations involving zeros and ones. This became the foundation of modern formal logic and, eventually, digital computing. The appeal of this approach is its clarity: if thinking follows rules, then truth can be determined mechanically. But as powerful as formal logic is, it quickly runs into limits when faced with the uncertainty and ambiguity of real human thought.
That limitation became clearer with the rise of behaviorism in the early twentieth century. Psychologists, skeptical about studying something as intangible as the mind, focused strictly on observable behavior. However, this approach proved too restrictive. Experiments by researchers like Jerome Bruner demonstrated that internal factors - such as perception, value, and context - shape how people interpret the world. For instance, children’s estimates of coin sizes varied depending on their economic background, suggesting that the mind cannot be ignored. These findings helped spark the cognitive revolution, which reintroduced the study of mental processes as a legitimate scientific endeavor.
At the same time, developments in computing offered new ways to model thought. Early pioneers such as Charles Babbage and Alan Turing imagined machines that could follow rules to perform calculations, while Claude Shannon connected logic to electrical circuits. Building on these ideas, researchers like Herbert A. Simon and Allen Newell proposed that human thinking might operate as a system of symbols manipulated by rules. This 'rules and symbols' approach suggested that complex behavior could emerge from simple operations, much like ants creating intricate patterns by following basic instructions. Yet this model struggled to explain how humans deal with uncertainty, learn from limited data, and generalize beyond explicit rules.
Language posed another challenge. Noam Chomsky argued that humans must possess innate structures that make language learning possible, since children acquire it far too quickly for it to be based solely on experience. This introduced the idea of inductive reasoning - drawing conclusions from incomplete information - which formal logic alone cannot fully capture. Human cognition relies heavily on such inference, whether predicting the weather or recognizing objects from partial visual input.
The problem of categorization further exposed the limits of rigid logical systems. Research by Eleanor Rosch revealed that categories like 'color' or 'furniture' don’t have strict boundaries. Instead, they have central examples, or prototypes, with fuzzy edges where membership becomes ambiguous. This insight led to models that represent knowledge in terms of similarity and distance within a conceptual space, rather than strict true-or-false definitions. It’s a more flexible way of understanding how we group and interpret the world.
Another major shift came with the development of neural networks. Early models like the perceptron, created by Frank Rosenblatt, attempted to mimic the brain’s structure by connecting layers of artificial neurons. These systems could learn from experience by adjusting the strength of connections between units, allowing them to recognize patterns and make predictions. Modern neural networks have expanded this idea dramatically, powering technologies like language models and image recognition. However, they also reveal a key difference between machines and humans: while neural networks require vast amounts of data to learn, humans can often learn from just a few examples.
This gap leads to the final major framework discussed in the book: probability theory, particularly Bayesian inference. Developed by thinkers like Thomas Bayes and Pierre-Simon Laplace, this approach models thinking as a process of updating beliefs based on new evidence. Instead of seeking certainty, the mind assigns probabilities to different possibilities and adjusts them as more information becomes available. This framework elegantly explains how humans handle uncertainty and learn efficiently. It also helps clarify why people can learn quickly: they come equipped with prior assumptions - biases shaped by evolution and experience - that guide their interpretations.
Taken together, these three approaches - formal logic, neural networks, and probability theory - offer complementary perspectives on the mind. Logic provides structure and clarity, neural networks capture learning from experience, and probability theory explains how we deal with uncertainty. None of them alone is sufficient, but together they form a richer picture of cognition.
The central insight of "The Laws of Thought: The Quest for a Mathematical Theory of the Mind" is that understanding the mind requires embracing this diversity of approaches. Human thought is not governed by a single set of equations or rules. Instead, it emerges from the interaction of structured reasoning, pattern recognition, and probabilistic inference. The quest to fully understand these processes is still ongoing, but each step brings us closer to unraveling one of the most profound mysteries: how we think, learn, and make sense of the world.
The journey begins with early efforts to formalize reasoning through logic. Philosophers like Aristotle laid the groundwork with syllogisms - structured arguments where conclusions follow from premises. Later thinkers, including Gottfried Wilhelm Leibniz, dreamed of turning reasoning itself into a kind of calculation. This vision took a major leap forward with George Boole, who showed that logical statements could be expressed using algebra with binary variables - essentially reducing thought to operations involving zeros and ones. This became the foundation of modern formal logic and, eventually, digital computing. The appeal of this approach is its clarity: if thinking follows rules, then truth can be determined mechanically. But as powerful as formal logic is, it quickly runs into limits when faced with the uncertainty and ambiguity of real human thought.
That limitation became clearer with the rise of behaviorism in the early twentieth century. Psychologists, skeptical about studying something as intangible as the mind, focused strictly on observable behavior. However, this approach proved too restrictive. Experiments by researchers like Jerome Bruner demonstrated that internal factors - such as perception, value, and context - shape how people interpret the world. For instance, children’s estimates of coin sizes varied depending on their economic background, suggesting that the mind cannot be ignored. These findings helped spark the cognitive revolution, which reintroduced the study of mental processes as a legitimate scientific endeavor.
At the same time, developments in computing offered new ways to model thought. Early pioneers such as Charles Babbage and Alan Turing imagined machines that could follow rules to perform calculations, while Claude Shannon connected logic to electrical circuits. Building on these ideas, researchers like Herbert A. Simon and Allen Newell proposed that human thinking might operate as a system of symbols manipulated by rules. This 'rules and symbols' approach suggested that complex behavior could emerge from simple operations, much like ants creating intricate patterns by following basic instructions. Yet this model struggled to explain how humans deal with uncertainty, learn from limited data, and generalize beyond explicit rules.
Language posed another challenge. Noam Chomsky argued that humans must possess innate structures that make language learning possible, since children acquire it far too quickly for it to be based solely on experience. This introduced the idea of inductive reasoning - drawing conclusions from incomplete information - which formal logic alone cannot fully capture. Human cognition relies heavily on such inference, whether predicting the weather or recognizing objects from partial visual input.
The problem of categorization further exposed the limits of rigid logical systems. Research by Eleanor Rosch revealed that categories like 'color' or 'furniture' don’t have strict boundaries. Instead, they have central examples, or prototypes, with fuzzy edges where membership becomes ambiguous. This insight led to models that represent knowledge in terms of similarity and distance within a conceptual space, rather than strict true-or-false definitions. It’s a more flexible way of understanding how we group and interpret the world.
Another major shift came with the development of neural networks. Early models like the perceptron, created by Frank Rosenblatt, attempted to mimic the brain’s structure by connecting layers of artificial neurons. These systems could learn from experience by adjusting the strength of connections between units, allowing them to recognize patterns and make predictions. Modern neural networks have expanded this idea dramatically, powering technologies like language models and image recognition. However, they also reveal a key difference between machines and humans: while neural networks require vast amounts of data to learn, humans can often learn from just a few examples.
This gap leads to the final major framework discussed in the book: probability theory, particularly Bayesian inference. Developed by thinkers like Thomas Bayes and Pierre-Simon Laplace, this approach models thinking as a process of updating beliefs based on new evidence. Instead of seeking certainty, the mind assigns probabilities to different possibilities and adjusts them as more information becomes available. This framework elegantly explains how humans handle uncertainty and learn efficiently. It also helps clarify why people can learn quickly: they come equipped with prior assumptions - biases shaped by evolution and experience - that guide their interpretations.
Taken together, these three approaches - formal logic, neural networks, and probability theory - offer complementary perspectives on the mind. Logic provides structure and clarity, neural networks capture learning from experience, and probability theory explains how we deal with uncertainty. None of them alone is sufficient, but together they form a richer picture of cognition.
The central insight of "The Laws of Thought: The Quest for a Mathematical Theory of the Mind" is that understanding the mind requires embracing this diversity of approaches. Human thought is not governed by a single set of equations or rules. Instead, it emerges from the interaction of structured reasoning, pattern recognition, and probabilistic inference. The quest to fully understand these processes is still ongoing, but each step brings us closer to unraveling one of the most profound mysteries: how we think, learn, and make sense of the world.
April 13, 2026
Griffiths presents a tidy evolution of cognitive science. I appreciate how accessible the book is to the layman, as well as how Griffiths fluidly builds upon each research discovery until we reach the understanding of how artificial intelligence models work in modern day. I also enjoyed how Griffiths contextualized each advancement with the individual behind it; so many had wonky personalities or indecisive academic histories.
For reference, Griffiths walks through the following:
- Boolean logic
- Deduction (inferring for a situation based on a set of facts)
- Induction (generalizing a rule from many instances)
- Similarity measurement via color theory and Manhattan distance
- Implicit syntax as proposed by Chomsky
- Neural networks (activation, failures/difference as a factor in how well information is retained) -- the feathers of the brain
- Probability (and Bayesian probability) as the aerodynamics of the brain
I left with a strong impression around the labor of research. Many ideas we now take for granted were originally contrarian or were discounted for several years before being brought back to life; novel research methodologies came from creative experiments or exotic ventures (one example was a scientist visiting a tropical indigenous colony which had a particular categorization of colors different from that of the English language). There's also something rather existential and beautiful about how mathematical formalism defines our world. My favorite bit from the book was Blaise Pascal renouncing his career in mathematics and devoting himself to religion, only to resolve his quandary on whether God exists by returning to math.
In his last chapter, Griffiths presents an overview of the continued work in cognitive science. I'm curious about how our brains build memory or how they interpret the world beyond language. Much of our internal world still lies undiscovered, but this book formalizes the lens by which we can understand it.
For reference, Griffiths walks through the following:
- Boolean logic
- Deduction (inferring for a situation based on a set of facts)
- Induction (generalizing a rule from many instances)
- Similarity measurement via color theory and Manhattan distance
- Implicit syntax as proposed by Chomsky
- Neural networks (activation, failures/difference as a factor in how well information is retained) -- the feathers of the brain
- Probability (and Bayesian probability) as the aerodynamics of the brain
I left with a strong impression around the labor of research. Many ideas we now take for granted were originally contrarian or were discounted for several years before being brought back to life; novel research methodologies came from creative experiments or exotic ventures (one example was a scientist visiting a tropical indigenous colony which had a particular categorization of colors different from that of the English language). There's also something rather existential and beautiful about how mathematical formalism defines our world. My favorite bit from the book was Blaise Pascal renouncing his career in mathematics and devoting himself to religion, only to resolve his quandary on whether God exists by returning to math.
If God exists, there are potentially infinite gains to be had in the form of an infinite afterlife. If he doesn't there is nothing to gain. Belief that he exists has only finite cost--the cost of living just a single observant life. So it is rational to believe, as long as the probability that God exists is anything greater than zero.
In his last chapter, Griffiths presents an overview of the continued work in cognitive science. I'm curious about how our brains build memory or how they interpret the world beyond language. Much of our internal world still lies undiscovered, but this book formalizes the lens by which we can understand it.
March 8, 2026
Far too basic, and even then not eminently readable like his first book on algorithms. It seemed somewhat of a "curve fitting" to refer those laws in the framework trifecta between logic, probability and neural networks. For the uninitiated, it gives a somewhat superficial walkthrough of the key events, or what the author believes to be key events, like Chomsky's work.
April 18, 2026
You'll need some background in mathematics and artificial intelligence to appreciate this book. It's largely a recounting of artificial intelligence research conducted since the nineteen-fifties. I found the book informative although much of the discussion required some prior knowledge of the subject which I have very little.
For me the best part of the book were the handful of examples used to illustrate the various theories discussed such as the concept "Kahneman and Tversky showed that people make judgements that are inconsistent with the basic idea that probabilities are assigned to different possible worlds." The author then follows with a simple example of bank tellers to make this concept come alive.
Or take this, "In his dissertation Horning showed that probabilistic languages as complex as those generated by probabilistic context-free grammar could be learned just from randomely generated examples of sentences in the language via Bayesian inference." What the heck does this mean? Fortunately, the author provides a beautiful example using two dice, one six-sided and the other twelve sided, to nicely illustrate the concept of Bayesian inference. The book would have been far more instructive, and interesting, had the author used more of these simple types of examples.
Recommended for experts or, otherwise, patient readers.
For me the best part of the book were the handful of examples used to illustrate the various theories discussed such as the concept "Kahneman and Tversky showed that people make judgements that are inconsistent with the basic idea that probabilities are assigned to different possible worlds." The author then follows with a simple example of bank tellers to make this concept come alive.
Or take this, "In his dissertation Horning showed that probabilistic languages as complex as those generated by probabilistic context-free grammar could be learned just from randomely generated examples of sentences in the language via Bayesian inference." What the heck does this mean? Fortunately, the author provides a beautiful example using two dice, one six-sided and the other twelve sided, to nicely illustrate the concept of Bayesian inference. The book would have been far more instructive, and interesting, had the author used more of these simple types of examples.
Recommended for experts or, otherwise, patient readers.
March 28, 2026
Weirdly enough, I totally loved this. It's a mix of history & cognitive science 101. It certainly required my whole attention--probably the first time I've listened to something on 1x speed in years (listening to someone talk about math requires incredible amounts of brainpower from me).
As someone who (generally speaking) has never understood how math can be used to represent anything other than literal quantities, this felt like an introduction to a new world. For the very first time, I felt like I understood why logic and probability theory matter, how math can be used to represent both, how computers work, and how this all is essential in understanding the nuts & bolts of what is really going on inside of a neural network. It does have some math, yes, but overall I found the book quite accessible (and funny), while still engaging with the mechanics on a deeper level than I've been able to understand in the past.
It also led to a lot of really fun conversations with my partner (who studied cognitive science, thinks about just about everything in terms of probabilities, and often tries to explain his thoughts by asking me to "imagine them as embeddings within a multidimensional space"). He was extremely excited to talk to me about the various concepts within the book, and also a convenient source of explanations & further drawing out of concepts I struggled with (namely, what is actually happening during deep learning from a basic math perspective without getting into calculus). I finishing the book feeling like I'd learned a lot--and a lot that was practically important for making wise decisions about the use of AI in my own life.
I appreciated how it felt like a window into a way of thinking about & approaching the world which feels really foreign to me (but very familiar to my partner). Which was, frankly, really cool (despite feeling like quite the stretching experience for me). There are so many different ways of being in the world & experiencing life, it was really different to think about questions that frankly have never interested me before, like "what is thought?" I think in addition to helping me just understand our increasingly AI-dependent world, this book also helped me to better understand the people who are creating this technology.
Overall, I highly recommend, especially for those of us entirely alien to the AI world (and to computer science generally) and who keep asking, "Okay, but what is AI really" and never receive a satisfactory answer.
As someone who (generally speaking) has never understood how math can be used to represent anything other than literal quantities, this felt like an introduction to a new world. For the very first time, I felt like I understood why logic and probability theory matter, how math can be used to represent both, how computers work, and how this all is essential in understanding the nuts & bolts of what is really going on inside of a neural network. It does have some math, yes, but overall I found the book quite accessible (and funny), while still engaging with the mechanics on a deeper level than I've been able to understand in the past.
It also led to a lot of really fun conversations with my partner (who studied cognitive science, thinks about just about everything in terms of probabilities, and often tries to explain his thoughts by asking me to "imagine them as embeddings within a multidimensional space"). He was extremely excited to talk to me about the various concepts within the book, and also a convenient source of explanations & further drawing out of concepts I struggled with (namely, what is actually happening during deep learning from a basic math perspective without getting into calculus). I finishing the book feeling like I'd learned a lot--and a lot that was practically important for making wise decisions about the use of AI in my own life.
I appreciated how it felt like a window into a way of thinking about & approaching the world which feels really foreign to me (but very familiar to my partner). Which was, frankly, really cool (despite feeling like quite the stretching experience for me). There are so many different ways of being in the world & experiencing life, it was really different to think about questions that frankly have never interested me before, like "what is thought?" I think in addition to helping me just understand our increasingly AI-dependent world, this book also helped me to better understand the people who are creating this technology.
Overall, I highly recommend, especially for those of us entirely alien to the AI world (and to computer science generally) and who keep asking, "Okay, but what is AI really" and never receive a satisfactory answer.
February 28, 2026
I enjoyed this book.
We go way back in ancient history and forward to modern day in order to build upon different ideas in order to describe and mathematically solve how the human brain works. From coming up with simple logic puzzles, to and/or statements, to if-then statements, to trying to organize speech, to large language models, the quest continues.
#GoodReadsGiveAwayWinner
We go way back in ancient history and forward to modern day in order to build upon different ideas in order to describe and mathematically solve how the human brain works. From coming up with simple logic puzzles, to and/or statements, to if-then statements, to trying to organize speech, to large language models, the quest continues.
#GoodReadsGiveAwayWinner
March 29, 2026
I received a free copy of this book through Goodreads Giveaways. Thank you!
I really enjoyed reading this book! Griffiths discusses three areas of exploration and development that all work together to form the current mathematical models of our brain that are also used in current AI systems. For each line of inquiry, he first talks about the history of and teaches the mathematics behind that idea. Then he talks about its shortcomings and how another idea had to come along to improve on it. Finally, he connects all this to where we currently are with understanding the mind and applying this understanding to AI. Griffiths does a good job of breaking down complex topics to make them easy to understand and interesting to read about. My math and science background helped here, but I think any reader could work through it, too. I loved the approach of talking about the history of all these developments and how they spawned new ideas or contrasting ideas that strengthened our understanding. It really forms a compelling picture of how science works, where you have new ideas, test them out, find their flaws, and try to improve each time.
I really enjoyed reading this book! Griffiths discusses three areas of exploration and development that all work together to form the current mathematical models of our brain that are also used in current AI systems. For each line of inquiry, he first talks about the history of and teaches the mathematics behind that idea. Then he talks about its shortcomings and how another idea had to come along to improve on it. Finally, he connects all this to where we currently are with understanding the mind and applying this understanding to AI. Griffiths does a good job of breaking down complex topics to make them easy to understand and interesting to read about. My math and science background helped here, but I think any reader could work through it, too. I loved the approach of talking about the history of all these developments and how they spawned new ideas or contrasting ideas that strengthened our understanding. It really forms a compelling picture of how science works, where you have new ideas, test them out, find their flaws, and try to improve each time.
March 8, 2026
4.5 stars. Really fascinating book about the different computational approaches to understand the mind over the years.
Griffiths covers symbolic systems, neural networks, and Bayesian approaches and the academic debates that shaped them. One part that stood out was the notion that human brains can learn from so few examples, compared with LLMs, is because evolution gives us strong priors about how the world works. In that sense, LLM training, which starts with truly random weights, has to account both for a person’s “learning” before they were born as well as in their childhood.
I’m excited to see how the field continues to evolve!
Griffiths covers symbolic systems, neural networks, and Bayesian approaches and the academic debates that shaped them. One part that stood out was the notion that human brains can learn from so few examples, compared with LLMs, is because evolution gives us strong priors about how the world works. In that sense, LLM training, which starts with truly random weights, has to account both for a person’s “learning” before they were born as well as in their childhood.
I’m excited to see how the field continues to evolve!
Review of advance copy received from Netgalley
Welcome to an enlightening and quirky adventure through the history of logic using mathematics, linguistics, and pattern evolution to understand the human mind. This book is not just a dry exploration of complex theories, but a captivating narrative filled with personal biographies and insights born of the minds of some brilliant psychological heroes.
Griffiths masterfully ties together the heavy logic of foundational concepts like rules and symbols, neural networks, and probability and statistics, with the human stories behind their discoveries. The result is a tale that is as engaging as it is informative. It's like having a cup of coffee with a brilliant professor, who not only shares their latest research but also regales us with tales of the eccentric geniuses who came before them.
For those of us who may have struggled to grasp the intricacies of cognitive science and AI, "The Laws of Thought" is a breath of fresh air. Griffiths' ability to make these complex concepts accessible and relatable is truly commendable; however, I would truly recommend accessing the pdf or viewing some of Griffiths’ other written work as some of the topics are hard to grasp from the audio alone.
As we navigate the ever-evolving landscape of technology and artificial intelligence, it's essential to have a solid understanding of the foundational principles that underpin these fields. "The Laws of Thought" is an indispensable resource for anyone interested in the future of technology and the human mind. So, grab a copy, settle in, and prepare to be both entertained and enlightened!
Personally, I was surprised to find that the names of the mid-20th century psychologist-- that I’ve studied in classes for years-- belonged to men who held conferences together, influencing not only each other’s work, but also the field of AI. Weaving these tales together has shed new light on the entire field. If that isn’t the sign of a good book and time well spent, I’m not sure what is.
Thank you to NetGalley, Macmillan Audio, and Tom Griffiths for an ALC of this book.
Griffiths masterfully ties together the heavy logic of foundational concepts like rules and symbols, neural networks, and probability and statistics, with the human stories behind their discoveries. The result is a tale that is as engaging as it is informative. It's like having a cup of coffee with a brilliant professor, who not only shares their latest research but also regales us with tales of the eccentric geniuses who came before them.
For those of us who may have struggled to grasp the intricacies of cognitive science and AI, "The Laws of Thought" is a breath of fresh air. Griffiths' ability to make these complex concepts accessible and relatable is truly commendable; however, I would truly recommend accessing the pdf or viewing some of Griffiths’ other written work as some of the topics are hard to grasp from the audio alone.
As we navigate the ever-evolving landscape of technology and artificial intelligence, it's essential to have a solid understanding of the foundational principles that underpin these fields. "The Laws of Thought" is an indispensable resource for anyone interested in the future of technology and the human mind. So, grab a copy, settle in, and prepare to be both entertained and enlightened!
Personally, I was surprised to find that the names of the mid-20th century psychologist-- that I’ve studied in classes for years-- belonged to men who held conferences together, influencing not only each other’s work, but also the field of AI. Weaving these tales together has shed new light on the entire field. If that isn’t the sign of a good book and time well spent, I’m not sure what is.
Thank you to NetGalley, Macmillan Audio, and Tom Griffiths for an ALC of this book.
February 19, 2026
Thank you to NetGalley and Macmillan Audio for the advanced listening copy.
I listened to the audiobook and really enjoyed it, though it is definitely dense, intellectually packed, and layered.
My former economics brain appreciated how rigorously the author connected math, logic, and human decision-making. The historical threads from early logic to modern AI were fascinating, and I loved how ambitious the scope is. It feels like a serious intellectual workout in the best way.
That said, this is not a casual listen. I had to rewind a few sections to fully grasp the arguments, especially in audio format. It is rewarding, but you need to stay engaged. I think I would have appreciated it a bit more had I read a physical copy.
If you enjoy big ideas about how humans and machines think and do not mind something substantial, this book is worth picking up.
I listened to the audiobook and really enjoyed it, though it is definitely dense, intellectually packed, and layered.
My former economics brain appreciated how rigorously the author connected math, logic, and human decision-making. The historical threads from early logic to modern AI were fascinating, and I loved how ambitious the scope is. It feels like a serious intellectual workout in the best way.
That said, this is not a casual listen. I had to rewind a few sections to fully grasp the arguments, especially in audio format. It is rewarding, but you need to stay engaged. I think I would have appreciated it a bit more had I read a physical copy.
If you enjoy big ideas about how humans and machines think and do not mind something substantial, this book is worth picking up.
May 10, 2026
The Laws of Thought by Tom Griffiths
Most people acquire, almost by cultural osmosis, a rudimentary understanding of how the physical world operates.
Apples fall; planets orbit; if I jump off a building, I’ll accelerate at 9.8 meters per second squared until air resistance slows me; Newton has us covered.
Science in its grandiosity, has corralled the physical world with imperial efficiency. It has mapped the genome, weighed subatomic particles, and photographed black holes
The physical world once inscrutable, now appears understandable.
Yet, the internal world, by contrast, our mind’s mechanisms of reasoning, inference, categorization, and logical computation, remains largely terra incognita, its principles known to few and formalized by even fewer.
The Laws of Nature are familiar; the Laws of Thought remain largely inscrutable.
For a species capable of using mathematics to predict the oscillations of quantum fields, we remain perplexingly ignorant to our own cognition. We possess no canonical framework for curiosity, no prescriptive law for envy, no generally sanctioned account of how a child extrapolates grammar from the detritus of everyday speech.
Cognitive science, inaugurated in the mid-20th century with the ambition of rendering thought empirically tractable, has yielded formalism and devised experiments. What it has conspicuously failed to yield is consensus. Physicists may quarrel over interpretations, yet they cohere around the architecture of the universe…kind of. Cognitive scientists, by contrast, are still disputing the blueprint of the mind.
Is the mind a system of explicit rules manipulating symbols with bureaucratic precision? Is it a distributed network of units adjusting weights in response to data? Is it a probabilistic inference engine, perpetually updating beliefs under uncertainty?
Could mathematics, which charts the physical world with precision, also chart the landscape of our minds?
In The Laws of Thought: The Quest for a Mathematical Theory of the Mind, Tom Griffiths takes up this interesting quest. The book traces the long and uneven effort to apply mathematics—the tool that so successfully tamed the physical world—to cognition itself. If the Scientific Revolution delivered the Laws of Nature, Griffiths asks whether we are finally approaching a comparable account of the principles governing thought.
The story begins with the Scientific Revolution, when observation, experiment and mathematical formalism coalesced into a method powerful enough to explain phenomena from falling apples to celestial mechanics. Some thinkers, notably Hobbes, Descartes and Leibniz, wondered whether reasoning might be treated similarly—as a form of calculation. The ambition was clear: to render thought measurable, perhaps even computable. For centuries, that ambition outran available tools.
A breakthrough arrived in the 19th century with George Boole’s algebra of logic, which showed that reasoning could be rendered in symbolic form. From this flowed formal logic and, eventually, the conceptual foundations of computing. Yet only in the mid-20th century did scholars begin rigorously testing mathematical theories of thought against empirical data. The Cognitive Revolution transformed the mind from a philosophical curiosity into a scientific object.
Early models cast the mind as a rule-following system: cognition as symbol manipulation. Formal grammars, as developed in linguistics, showed how finite rules could generate infinite sentences. The approach was elegant…and incomplete. Human reasoning is often graded rather than binary; categories blur; learning proceeds with a speed and flexibility that rigid rule systems struggle to explain.
Two further frameworks emerged. Artificial neural networks replaced explicit rules with continuous representations and statistical learning. Given sufficient data, such systems can approximate remarkably complex functions. Modern artificial intelligence rests on this architecture. Yet its appetite for data is like hungry hungry hippopotamus.
Bayesian models offer a different perspective. Drawing on probability theory, they describe learning as rational belief-updating, shaped by prior assumptions. Humans, on this view, are efficient learners not because they passively absorb information, but because they approach the world with structured expectations. The central question shifts from “What rules govern thought?” to “What priors does the mind bring to experience?”
Griffiths presents these three traditions—rules and symbols, neural networks and Bayesian inference—not as mutually exclusive dogmas but as complementary approaches to a stubborn problem. Cognitive science has not achieved the theoretical consolidation of physics. Disagreement persists. Yet he suggests that recent work hints at convergence: after centuries of fragmentation, a more coherent account of the Laws of Thought may be emerging.
The book’s contemporary relevance is hard to miss. Artificial intelligence systems now perform tasks once deemed uniquely human, from strategic gameplay to fluent conversation. Their successes are impressive, yet they typically require prodigious quantities of data. Humans do not. Understanding why may illuminate both the promise of automation and its boundaries.
In the age of AI, this book is more than an intellectual history. It is a sober assessment of how far we have come in mathematizing the mind—and how far we have yet to go.
Most people acquire, almost by cultural osmosis, a rudimentary understanding of how the physical world operates.
Apples fall; planets orbit; if I jump off a building, I’ll accelerate at 9.8 meters per second squared until air resistance slows me; Newton has us covered.
Science in its grandiosity, has corralled the physical world with imperial efficiency. It has mapped the genome, weighed subatomic particles, and photographed black holes
The physical world once inscrutable, now appears understandable.
Yet, the internal world, by contrast, our mind’s mechanisms of reasoning, inference, categorization, and logical computation, remains largely terra incognita, its principles known to few and formalized by even fewer.
The Laws of Nature are familiar; the Laws of Thought remain largely inscrutable.
For a species capable of using mathematics to predict the oscillations of quantum fields, we remain perplexingly ignorant to our own cognition. We possess no canonical framework for curiosity, no prescriptive law for envy, no generally sanctioned account of how a child extrapolates grammar from the detritus of everyday speech.
Cognitive science, inaugurated in the mid-20th century with the ambition of rendering thought empirically tractable, has yielded formalism and devised experiments. What it has conspicuously failed to yield is consensus. Physicists may quarrel over interpretations, yet they cohere around the architecture of the universe…kind of. Cognitive scientists, by contrast, are still disputing the blueprint of the mind.
Is the mind a system of explicit rules manipulating symbols with bureaucratic precision? Is it a distributed network of units adjusting weights in response to data? Is it a probabilistic inference engine, perpetually updating beliefs under uncertainty?
Could mathematics, which charts the physical world with precision, also chart the landscape of our minds?
In The Laws of Thought: The Quest for a Mathematical Theory of the Mind, Tom Griffiths takes up this interesting quest. The book traces the long and uneven effort to apply mathematics—the tool that so successfully tamed the physical world—to cognition itself. If the Scientific Revolution delivered the Laws of Nature, Griffiths asks whether we are finally approaching a comparable account of the principles governing thought.
The story begins with the Scientific Revolution, when observation, experiment and mathematical formalism coalesced into a method powerful enough to explain phenomena from falling apples to celestial mechanics. Some thinkers, notably Hobbes, Descartes and Leibniz, wondered whether reasoning might be treated similarly—as a form of calculation. The ambition was clear: to render thought measurable, perhaps even computable. For centuries, that ambition outran available tools.
A breakthrough arrived in the 19th century with George Boole’s algebra of logic, which showed that reasoning could be rendered in symbolic form. From this flowed formal logic and, eventually, the conceptual foundations of computing. Yet only in the mid-20th century did scholars begin rigorously testing mathematical theories of thought against empirical data. The Cognitive Revolution transformed the mind from a philosophical curiosity into a scientific object.
Early models cast the mind as a rule-following system: cognition as symbol manipulation. Formal grammars, as developed in linguistics, showed how finite rules could generate infinite sentences. The approach was elegant…and incomplete. Human reasoning is often graded rather than binary; categories blur; learning proceeds with a speed and flexibility that rigid rule systems struggle to explain.
Two further frameworks emerged. Artificial neural networks replaced explicit rules with continuous representations and statistical learning. Given sufficient data, such systems can approximate remarkably complex functions. Modern artificial intelligence rests on this architecture. Yet its appetite for data is like hungry hungry hippopotamus.
Bayesian models offer a different perspective. Drawing on probability theory, they describe learning as rational belief-updating, shaped by prior assumptions. Humans, on this view, are efficient learners not because they passively absorb information, but because they approach the world with structured expectations. The central question shifts from “What rules govern thought?” to “What priors does the mind bring to experience?”
Griffiths presents these three traditions—rules and symbols, neural networks and Bayesian inference—not as mutually exclusive dogmas but as complementary approaches to a stubborn problem. Cognitive science has not achieved the theoretical consolidation of physics. Disagreement persists. Yet he suggests that recent work hints at convergence: after centuries of fragmentation, a more coherent account of the Laws of Thought may be emerging.
The book’s contemporary relevance is hard to miss. Artificial intelligence systems now perform tasks once deemed uniquely human, from strategic gameplay to fluent conversation. Their successes are impressive, yet they typically require prodigious quantities of data. Humans do not. Understanding why may illuminate both the promise of automation and its boundaries.
In the age of AI, this book is more than an intellectual history. It is a sober assessment of how far we have come in mathematizing the mind—and how far we have yet to go.
May 2, 2026
This was a book I really wanted to read. To get it I had to drive 60 miles roundtrip to the Seattle library and will have to do the same to return it. Was it worth the drive? Yes and no. It is great that there is a book written for a general audience on the current state of our understanding of cognitive psychology and the search for a mathematical model of the mind. Griffiths takes us thorugh the history of some of that research. After looking at logic and features, he turns to neural networks and probabilities. Clearly, we can hear the baying of the hounds in the background as scientists are hot on the trail of the fox of a computational model of human thought to let them create a truely general AI.
However, for me at least the book proved a bit of a let down. Perhaps it was because this was a historical survey, in part, and so there was insufficient space to explain everything for the general reader with limited math background. Sometimes after reading a section of this book and coming away somewhat less than enlightened, I'd turn to Claude or ChatGPT for other sources. For example, ChatGPT suggested “Neural Networks and Deep Learning” – Michael Nielsen http://neuralnetworksanddeeplearning...., a free online book. The chapter on neural networks that I read there was very clear and helpful without being longer than Griffiths section on the same topic. This was true also for understanding Bayes and Bayesian probability, too. And then the discussion of Shepard's psychological rule of generalization was, well, mystifying. The figures 13.1, 13.2, and 13.3 that he introduces in Chapter 13 are never adequately explained, and he really should have said more about psychological space and how generalization works with more examples. (Though an earlier part of the book discussed geometric representations in the work of Eleanor Rosch and others, more could have been said in Chapter 13 about how Shepard used this type of representation and how he arrived at the insight that it was a universal psychological rule of generalization.) Perhaps Shepard is just hard to get for a general reader? The author notes that Isaac Asimov, no idiot, read Shepard's initial study in Science and struggled to understand it. I felt the same after Griffith's explanation. The general idea is not hard to get, but the details and why this was such an insight were far from clear.
Small side note, I think there is a crucial typo on page 269 when the author is explaining the logic of a card game in a Harvard study. The player is told that we have a set of two sided playing cards with letters and numbers on the cards. The only rule is that if there is a "D" on one side there must be a "3" on the other side. How do we test that rule? If we see four cards on the table showing a D, a 3, a B, and a 7. The logical rule is that where there is a "D" visible, we should find a "3" on the other side, so to test this we would check the card showing "D", but also the cards showing a "3" and a "7". If there is a D on the other side of the "3" card, we have some confirmation of the claim, but if there is a "D" on the other side of the "7" card we know the rule is wrong because only a "3" can be opposite a "D". Instead Griffiths tells us that we are checking the "7" to see if there is a "3" on the other side. Obviosuly that is wrong. There is no rule that requires a 3 to only occur on cards with Ds. After puzzling over that for a while, I checked with ChatGPT and it helpfully confirmed that we are looking for a "D" on the other side of the "7" not a "3". Kind of meta to use ChatGPT and Claude as part of reading a book on, in part, such programs. I can't help but wonder what the Bayesian probability is that Griffith made other errors that confused his readers? I almost dropped him to 2 stars for the uncaught error, but he is afterall only human, right? I did like this one enough that I'd be interested to read his earlier book on algorithms.
However, for me at least the book proved a bit of a let down. Perhaps it was because this was a historical survey, in part, and so there was insufficient space to explain everything for the general reader with limited math background. Sometimes after reading a section of this book and coming away somewhat less than enlightened, I'd turn to Claude or ChatGPT for other sources. For example, ChatGPT suggested “Neural Networks and Deep Learning” – Michael Nielsen http://neuralnetworksanddeeplearning...., a free online book. The chapter on neural networks that I read there was very clear and helpful without being longer than Griffiths section on the same topic. This was true also for understanding Bayes and Bayesian probability, too. And then the discussion of Shepard's psychological rule of generalization was, well, mystifying. The figures 13.1, 13.2, and 13.3 that he introduces in Chapter 13 are never adequately explained, and he really should have said more about psychological space and how generalization works with more examples. (Though an earlier part of the book discussed geometric representations in the work of Eleanor Rosch and others, more could have been said in Chapter 13 about how Shepard used this type of representation and how he arrived at the insight that it was a universal psychological rule of generalization.) Perhaps Shepard is just hard to get for a general reader? The author notes that Isaac Asimov, no idiot, read Shepard's initial study in Science and struggled to understand it. I felt the same after Griffith's explanation. The general idea is not hard to get, but the details and why this was such an insight were far from clear.
Small side note, I think there is a crucial typo on page 269 when the author is explaining the logic of a card game in a Harvard study. The player is told that we have a set of two sided playing cards with letters and numbers on the cards. The only rule is that if there is a "D" on one side there must be a "3" on the other side. How do we test that rule? If we see four cards on the table showing a D, a 3, a B, and a 7. The logical rule is that where there is a "D" visible, we should find a "3" on the other side, so to test this we would check the card showing "D", but also the cards showing a "3" and a "7". If there is a D on the other side of the "3" card, we have some confirmation of the claim, but if there is a "D" on the other side of the "7" card we know the rule is wrong because only a "3" can be opposite a "D". Instead Griffiths tells us that we are checking the "7" to see if there is a "3" on the other side. Obviosuly that is wrong. There is no rule that requires a 3 to only occur on cards with Ds. After puzzling over that for a while, I checked with ChatGPT and it helpfully confirmed that we are looking for a "D" on the other side of the "7" not a "3". Kind of meta to use ChatGPT and Claude as part of reading a book on, in part, such programs. I can't help but wonder what the Bayesian probability is that Griffith made other errors that confused his readers? I almost dropped him to 2 stars for the uncaught error, but he is afterall only human, right? I did like this one enough that I'd be interested to read his earlier book on algorithms.
October 14, 2025
Can we use the laws of mathematics to describe the many ways human beings think? The author of this new book definitely believes so. Griffiths is a professor at Princeton focusing on information technology. His research is on interdisciplinary questions at the intersection of psychology and computer science. He received real national attention with his previous book, Algorithms to Live By: the Computer Science of Human Decisions, published by Henry Holt & Company in 2016. Griffith’s research explores connections between human and machine learning, using ideas from statistics and artificial intelligence to understand how people solve the challenging human questions and problems they encounter in everyday life. He favors introducing ideas from computer science and cognitive science to wider general audiences. In this new book, Griffiths identifies three major approaches of rules and symbols, neural networks, and probability and statistics. The idea of “the laws of thought” began with the 19th century philosopher and mathematician, George Boole. Boole’s mathematical work in algebra and logic is essential to computer programming and helped lay a foundation for today’s information age. The very idea, however, can be better credited to the Greek philosopher Aristotle and his ideas underlying modern science and logic. As informed conversations about thought, language, and learning become ever more pressing in the age of artificial intelligence, The Laws of Thought is an essential resource for anyone interested in the future of technology. . The final chapter – appropriately titled “putting it all together” – is an excellent summary of the state of such applied efforts at understanding human decision making. Highly recommended.
April 25, 2026
Whenever experts in a field can zoom out and summarize the birds eye view perspective for me it is so useful. Cognitive science is the field that I am supposedly an expert with my Bachelors, Masters, and soon to be PhD in this field haha. Yet this book summarized the field in a way I wouldn't have fully been able to put together!
Griffiths describes cognitive science as: “…using mathematics to express precise hypotheses about how thought and language work, and then using information about human behavior to evaluate those hypotheses.” He outlines the giants in the field and how each has brought unique discoveries that have evolved into the AI revolution we're seeing today. This whole field of neuroscience influencing AI and vice versa is so interesting to me. How can we look at how the brain processes information so efficiently and apply those techniques to the data-hungry AI models? And how can we then simulate what we think the brain is doing in these models and tweak parameters to form new hypotheses in neuroscience?
I feel so lucky to have taken Griffith's Computational Models of Cognition" class when he was a professor at Berkeley. I was truly stumped at this question of how humans are able to learn so quickly with so little information. A child only has to hear a few sounds, and suddenly appears to learn complex syntax and how those arbitrary sounds correspond to meaningful concepts. This book is more or less the complete answer to this question and how researchers have attempted to model this behavior. From logic and rules, to probability distributions, we have started to approximate what our brains are doing and that is so COOL. Only knocking a star because there were some convoluted parts (even for someone in the field) and it was a lotttttt of name / institution dropping that I understand is very science-like, but got distracting from the main points.
Griffiths describes cognitive science as: “…using mathematics to express precise hypotheses about how thought and language work, and then using information about human behavior to evaluate those hypotheses.” He outlines the giants in the field and how each has brought unique discoveries that have evolved into the AI revolution we're seeing today. This whole field of neuroscience influencing AI and vice versa is so interesting to me. How can we look at how the brain processes information so efficiently and apply those techniques to the data-hungry AI models? And how can we then simulate what we think the brain is doing in these models and tweak parameters to form new hypotheses in neuroscience?
I feel so lucky to have taken Griffith's Computational Models of Cognition" class when he was a professor at Berkeley. I was truly stumped at this question of how humans are able to learn so quickly with so little information. A child only has to hear a few sounds, and suddenly appears to learn complex syntax and how those arbitrary sounds correspond to meaningful concepts. This book is more or less the complete answer to this question and how researchers have attempted to model this behavior. From logic and rules, to probability distributions, we have started to approximate what our brains are doing and that is so COOL. Only knocking a star because there were some convoluted parts (even for someone in the field) and it was a lotttttt of name / institution dropping that I understand is very science-like, but got distracting from the main points.
April 19, 2026
Tom Griffiths’ The Laws of Thought presents a structured attempt to formalize cognition through mathematical frameworks, positioning the mind as a system that can be described using principles analogous to the laws governing physical reality. The book’s core ambition is to unify historical and modern approaches to understanding thought under a coherent theoretical lens.
The work is strongest in its comparative framing of three dominant paradigms: symbolic reasoning systems, neural network models, and probabilistic inference. By tracing their intellectual lineage and practical implications, Griffiths highlights both their explanatory power and their limitations when used independently to model human cognition.
What emerges is not a single definitive theory, but a disciplined mapping of competing frameworks that collectively approximate how thought might be structured. The book is particularly relevant in the context of modern AI, where these paradigms intersect but still fail to fully replicate human-like reasoning, underscoring the gap between computational simulation and cognitive reality.
The work is strongest in its comparative framing of three dominant paradigms: symbolic reasoning systems, neural network models, and probabilistic inference. By tracing their intellectual lineage and practical implications, Griffiths highlights both their explanatory power and their limitations when used independently to model human cognition.
What emerges is not a single definitive theory, but a disciplined mapping of competing frameworks that collectively approximate how thought might be structured. The book is particularly relevant in the context of modern AI, where these paradigms intersect but still fail to fully replicate human-like reasoning, underscoring the gap between computational simulation and cognitive reality.
May 9, 2026
3.5 rounded up; i loved how this was a non-linear history of artificial intelligence. it was grouped by concept so you could really see how the individual contributions of different researchers around the world came together to be where we are now. i was going to round this DOWN because I will say it was a lot of names being thrown around and hard to keep track, but I listened to Griffiths’ me, myself and AI podcast episode (lol) and feel like I really aligned with his view of AI as a complement to human intelligence so i’m rounding up.
i think it’s really interesting to see AI not as a representation of human intelligence, but as a completely different SPECIES of intelligence. breaking it down to the fact that LLMs require the equivalent of 10,000 hours of conversation to learn a language, whereas we are able to learn a new language as children with very limited experience on earth is almost comforting. i appreciated that he makes the point that there is something special about us that can’t quite be replicated!!!
i think it’s really interesting to see AI not as a representation of human intelligence, but as a completely different SPECIES of intelligence. breaking it down to the fact that LLMs require the equivalent of 10,000 hours of conversation to learn a language, whereas we are able to learn a new language as children with very limited experience on earth is almost comforting. i appreciated that he makes the point that there is something special about us that can’t quite be replicated!!!
Review of advance copy received from Goodreads Giveaways
In The Laws of Thought, author Tom Griffiths explores the central question: how do we measure and model the mind mathematically, not only to understand human thinking but also to recreate it through artificial intelligence.
This fascinating intellectual history begins with Aristotle, Leibniz, and Boole, guiding readers through the foundations of logic, formal systems, language, and much more, exploring how we organize thought in ways that can be analyzed and predicted.
Along the way, Griffiths adds context by sharing short biographies of the key figures mentioned; where they came from, how they got interested in the mind, and how their ideas developed.
A compelling blend of cognitive science, language and history of artificial intelligence, Highly recommended.
Many Thanks to Holt and Goodreads for the ARC.
This fascinating intellectual history begins with Aristotle, Leibniz, and Boole, guiding readers through the foundations of logic, formal systems, language, and much more, exploring how we organize thought in ways that can be analyzed and predicted.
Along the way, Griffiths adds context by sharing short biographies of the key figures mentioned; where they came from, how they got interested in the mind, and how their ideas developed.
A compelling blend of cognitive science, language and history of artificial intelligence, Highly recommended.
Many Thanks to Holt and Goodreads for the ARC.
May 8, 2026
True, this book is a great exposition of the Mathematics of thought, as the subtitle claims.
However, not of the Philosophy of thought. For those seeking something along these lines better read Chomsky and, especially, Jerry Fodor, who, by the way, wrote two books named "The Language of Thought", which Griffiths doesn't even bother to mention in a book called "The Laws of Thought"!
I'm not surprised in the least however. If he did, the book would turn to ontology and metaphysics. And this is of course anathema in Western (non) thinking/civilization!! And then these people believe they are discovering the laws of thinking. What they are instead discovering is solely the Compositionality of thinking, not of Thought itself.
At least, the last few pages attest that there are still, thank God, people who think along real lines!
However, not of the Philosophy of thought. For those seeking something along these lines better read Chomsky and, especially, Jerry Fodor, who, by the way, wrote two books named "The Language of Thought", which Griffiths doesn't even bother to mention in a book called "The Laws of Thought"!
I'm not surprised in the least however. If he did, the book would turn to ontology and metaphysics. And this is of course anathema in Western (non) thinking/civilization!! And then these people believe they are discovering the laws of thinking. What they are instead discovering is solely the Compositionality of thinking, not of Thought itself.
At least, the last few pages attest that there are still, thank God, people who think along real lines!
May 19, 2026
Disappointing. A supremely important topic - we made a bunch of progress and people don't realise! we grew a new sort of mind! - but he gives neither enough maths nor diagrams nor brilliant analogies to carry the load.
Very focussed on the revolt against behaviourism and the linguistics wars, which now seem a little parochial and irrelevant to me. (We no longer have to fight a rearguard action against Skinnerites nor Chomskyans.)
If this is your first encounter with linguistics / cognitive science / AI it's fine and well. And I did learn some good stuff, like Dummett's brutal takedown of Boole's many errors.
The concluding note on AI - that we are more sample-efficient and more general than AI (we have a motor policy AND a text policy) - is deluded and Disneyfied.
I was hoping to find something which displaces chapter 1 of Russell and Norvig or "The Universal Computer" or Gleick or HC von Baeyer or Dehaene as a single up to date treatment of the great programme. I didn't - but it is as clear as prose can get, and suitable for a ten year old.
Very focussed on the revolt against behaviourism and the linguistics wars, which now seem a little parochial and irrelevant to me. (We no longer have to fight a rearguard action against Skinnerites nor Chomskyans.)
If this is your first encounter with linguistics / cognitive science / AI it's fine and well. And I did learn some good stuff, like Dummett's brutal takedown of Boole's many errors.
The concluding note on AI - that we are more sample-efficient and more general than AI (we have a motor policy AND a text policy) - is deluded and Disneyfied.
I was hoping to find something which displaces chapter 1 of Russell and Norvig or "The Universal Computer" or Gleick or HC von Baeyer or Dehaene as a single up to date treatment of the great programme. I didn't - but it is as clear as prose can get, and suitable for a ten year old.
Review of advance copy received from Goodreads Giveaways
Wasn’t sure what to make of this book going in, but I ended up completely fascinated by not only the topic, but by how it’s presented. It’s topical, especially due to the inclusion of AI, and it’s given me a lot to consider in terms of the history of thought and philosophy and progress that’s been made in these realms. Technology isn’t generally something I seek to learn about (I usually find it overwhelming), but here it’s made relevant and not just discussed for the sake of it. I appreciated the anecdotal and personal accounts woven into this book, as they help ground things in the everyday. Still had to read the book slowly to digest what was being conveyed, but I’m glad I did! A solid paperback ARC to have won in a Goodreads Giveaway!
March 7, 2026
Perfectly sums up the history of cognitive science and the strive to uncover a computational, algorithmic and implementational mathematics-driven model of the human mind.
I thoroughly enjoyed the transitions and progression between concepts and theories prevalent in the early 20th century and the AI tools raging in the modern world.
Would recommend basic mathematics to understand the book in greater detail; fast-paced and amazing read nonetheless.
I thoroughly enjoyed the transitions and progression between concepts and theories prevalent in the early 20th century and the AI tools raging in the modern world.
Would recommend basic mathematics to understand the book in greater detail; fast-paced and amazing read nonetheless.
March 12, 2026
it was a good account of the history of logical thought. I enjoyed the comparisons and contrasts between human and machine intelligence including the roles of natural selection in priming human brains for language and the impact that has on the required training for language models. the explanation of hidden Markov models gave me an aha moment as to how llms work.
April 20, 2026
Engaging, entertainingly written history of (what Griffiths sees as) the key developments in cognitive science and adjacent fields seeking to understand and explain thinking with mathematics.
A lot of this was familiar to me — Boole, Shannon, the connectionists / PDP program — but I appreciated the contextualization within an intellectual history.
A lot of this was familiar to me — Boole, Shannon, the connectionists / PDP program — but I appreciated the contextualization within an intellectual history.
May 13, 2026
This book presents a careful history of the development of efforts to produce machines that can mimic the human mind. It is couched in terms of the thinking processes that can be expressed mathematically. The systemization of thought through rules, neural networks, and probability are presented clearly with illustrations and in an historical context.
March 29, 2026
great history of cog sci x ai
although i expected this book to introduce new material, it ended up being more of a historical synopsis of interactions between cog sci and AI, which was enjoyable in its own right
although i expected this book to introduce new material, it ended up being more of a historical synopsis of interactions between cog sci and AI, which was enjoyable in its own right
Review of advance copy received from Netgalley
A compelling blend of cognitive science, language and history of artificial intelligence, Highly recommended.
February 15, 2026
Intriguing framework for metacognition. Best consumed in smaller segments, for proper digestion. Will surely be ready again at a later point.
February 24, 2026
Accessible yet sophisticated.
Displaying 1 - 30 of 38 reviews