A Brief History of Intelligence: Evolution, AI, and the Five Breakthroughs That Made Our Brains

Eye opening work on the brain, its hierarchy of reinforcement learning and emergent intelligence
This is the best book I have read in a long time. I wasn’t quite sure what to expect but this exceeded all expectations and for me is in the must read category for the interested reader but in particular for those curious about AI, neuroscience and any part the spectrum between those fields. The work is ultimately an aggregation of results from neuroscience and computer science in a synthesis attempting to frame how brain functioning has evolved, how those cognitive functions developed led to a variety of physical manifestations allowing for our theory of mind and language and how these were not inevitable. This is not a work of neuroscience but is woven into break throughs in computer science and machine learning in particular. I know of no other book like this but wish there were more, it is an impressive overview of remarkable fields and interweaves their results with mastery.We live in remarkable times for technology but often forget how much more remarkable the brain is and not strictly for homo sapiens. The author starts at the beginning of life and describes how mobility became a competitive advantage and how such physical mechanisms provided the basis for early adaptive learning benefits for early multi-cellular objects and were fundamentally distinct from multicellular objects stuck in place like corals. From simple bipedal movement the author incrementally moves on to a story of evolution and the strategies employed to drive evolution forward. The earlier parts of the book are mainly about how simple neuron patterns formed that created positive and dampening feedback loops to stimuli and many famous early experiments are detailed that were critical for neuroscience to make progress on the nervous system. The author moves on to evolutionary gains in the brain as adding more supervision to the instinctual networks formed in the basal ganglia and likened them to supervised learning with a teacher. The author gives the reader computer science history lessons with the breakthroughs made in machine learning techniques that coincidentally seem to match what mammalian brains do now through their layered hierarchy and how such techniques allowed for solving of very difficult problems when rewards can often have a long time to emerge from actions taking. This phenomenon of the credit assignment problem is discussed and the author provides the reader with an overview of how our minds have parts of the brain seemingly dedicated to solving this problem.The author discusses how the large extinction event several hundred million years ago from what was almost certainly a large asteroid hitting the planet catalyzed the era of mammals and the evolution of the mammalian brain to which we all owe our current states of mind. The author describes how the neo-cortex evolved and some of the functions which provide a theory of mind that allowed for planning as well as tool building for multi-step optimization. The author interweaves the current state of machine learning and deep learning techniques highlighting similarities and differences. The current lack of internal representation of the world in language models is deemed to be a deficiency vs the way the brain works with its layers creating more complex models of the world (ie operating environment for the body). The human brain has uniquely created extensions of the neo-cortex associated with language and such a functionality has allowed for the effective hive mind of humanity where our knowledge scales with language lowering the friction on communicating what is in our heads. There are a number of neuroscience case studies weaved in about brain defects that impact a variety of cognitive abilities and such examples highlight the utility of certain functionalities but also that the mind has plasticity such that few parts of the brain are absolutely needed without substitution.There is really too much information in this book to really summarize quickly but the picture the author gives is a highly illuminating perspective on the mind, how it has evolved, what have been the ingredients of its intelligence and how there have been layers of abstraction and hierarchical modeling that have gotten us to where we are. Almost as importantly the author runs a parallel track of machine learning advances, how many mimic evolutionary advances in the brain and how such current models have closed many gaps but still have much ground to cover to effectively have a model of the world in which the questions being asked of them are simulated rather than merely inferred from a trained dataset. There are no claims on where the boundaries are and this is not a book about the limits of technology nor its inevitable dominance. It is a remarkable overview of extremely complex and interwoven results that are making major strides that we are witnessing in real time. This really is a must read.

Ambitious and Interesting, Though Some Framing Feels Unnecessary
This is an interesting and ambitious book, but I was puzzled by the author’s mentioning the term chauvinism hinting human chauvinism as I read it. The claim that we must avoid thinking modern humans are strictly superior to modern animals strikes me as a misframed argument.In evolutionary terms, humans are indeed just one species among many, adapted to a particular ecological niche. That point is uncontroversial. But outside of biological fitness, it is unclear why humans and animals should be compared at all. Humans are not merely biological entities; they are normative agents who create language, institutions, science, ethics, and history itself.Labeling human exceptionalism as “chauvinism” imports moral language that seems unnecessary and rhetorically loaded, as if a serious prejudice were being corrected. In reality, no serious reader needs to be warned against confusing evolutionary adaptation with moral worth. The result feels less like clarification and more like the construction of a problem that does not actually require resolution.The book has incredible strengths, but this particular framing felt conceptually imprecise and somewhat performative.

One of the best books about AI
This book should be read by everyone trying to understand, debate or build anything related to AI. Definitely a book to be read many times.

This book covers the evolution of human intelligence by describing 5 crucial breakthroughs, and how each built on the earlier ones. It’s a fascinating approach and I found it really enlightening. For example, in the 5th breakthrough, how human language acquisition and use distinguishes us from our close primate relatives.The scope of the book is huge – very impressive that it took only a year to write.Two points to mention:The concept of ‘intelligence’ isn’t defined anywhere in the book that I could find. I think it should be, given that it’s the basis of the book. We all know what it means, sort of, but different people would probably explain it differently. Similarly, an entry in the Glossary for ‘eukaryote’ could be helpful.The mixing of systems of units looks awkward – for example in Figure 1.4, microns together with inches. I think it would be better to stick with metric to be consistent with scientific writing (and use in many parts of the world).

From Bilaterians, to vertebrates, to mammals, to first primates, to early humans and modern humans. The book captures five breakthroughs (transitions) on the key features of the brain of those species mentioned that govern their survival, with detailed neuroscience and mixed bag of deep learning and reinforcement learning accounts in explaining how the brain works.The book is simply a refreshing read, provided I am solely a recent self-taught AI researcher without much background in neuroscience. It encourages me to go on and read more about the brain while I have more time, as I have now become more knowledgeable in those topics, which were once too scary to enter.