This is a book of hope for humanity. But more than just hope, Jeff Hawkins has a plan–a plan for all humanity, and beyond. Each of its three sections is largely independent, and yet in the final two, short chapters Jeff Hawkins unites the book into a breathtaking estate plan for humanity.

This book is very short, less than 300 pages on my Kindle, but the scope is truly universal. Hawkins explores several cosmic struggles, such as good vs. evil, threats from aliens, killer robots, and the emergence of cyborgs–part human, part intelligent machine. Not shy of difficult topics, he tackles consciousness, religion, fake news, wars, artificial intelligence, SETI, origin and purpose of emotions, neuroscience, and a new theory of how the brain actually works.

All this he does with an easy, “aw, shucks,” conversational writing style that makes the reading itself a pleasure.

The book is structured in three sections, each somewhat independent until the thrilling, surprise conclusion.

• Part 1: A New Understanding of the Brain
• Part 2: Machine Intelligence
• Part 3: Human Intelligence

A New Understanding of the Brain

This brief description does not in any way summarize the entire part 1. Hopefully, it contains just enough information to entice you to read the book.

Hawkins unfolds the theory of a thousand brains by telling the story of Vernon Mountcastle, who wrote a small book, The Mindful Brain. In Hawkins’ words, “Mountcastle’s essay had an immediate and profound effect on me, and, as you will see, his proposal heavily influenced the theory I present in this book.”

Mountcastle proposed that the neocortex is not fundamentally separated into different functions, but is really only a collection of similar functional units. Hawkins draws a parallel to Charles Darwin’s idea. “Darwin proposed that the diversity of life is due to one basic algorithm. Mountcastle proposed that the diversity of intelligence is also due to one basic algorithm.”

Mountcastle proposed that this basic algorithm, indeed the fundamental unit of intelligence, was located in the neocortex, and named it the “cortical column.” It looks a little like a grain of rice, or a short piece of spaghetti. Each column extends the entire width of the neocortex, about 2.5 mm. These columns are stacked vertically throughout the entire cortex, about 150,000 columns in a single brain.

These cortical columns all have a similar structure. Most have some sensory input at the top, nearest the skull. Each has several layers of neurons in the middle, with the connections running mostly vertically through these layers.

Mountcastle proposed that the brain works on a universal learning mechanism. At that time there was simply not enough known about the brain, particularly the neocortex, to develop the mechanisms and structures to support this principle, only the stunning ability of human brains to learn anything.

Jeff Hawkins actually met Mountcastle and asked him to sign Jeff’s copy of Mountcastle’s book, which had started Jeff on this journey many years earlier. Eventually, Hawkins founded Numentum specifically to develop a model of the brain, which developed into the theory of a Thousand Brains. In his words:

“The brain creates a predictive model. This just means that the brain continuously predicts what its inputs will be. Prediction isn’t something that the brain does every now and then; it is an intrinsic property that never stops, and it serves an essential role in learning. When the brain’s predictions are verified, that means the brain’s model of the world is accurate. A misprediction causes you to attend to the error and update the model.”

We all know the joy of hearing a tune and predicting the next note just before it is played. This is the prediction machine at work. Similarly, we all know the surprise when the prediction model fails. This is the basis of humor; the punch line is a surprise. More often, the prediction failure causes some stress, such as when the car keys are not where we thought, or when a person acts differently than we predicted.

This stress may be an integral part of learning, and a small motivator to update the prediction model. Hawkins ties learning and prediction to movement, initially physical movement. The view of a room as we walk through it and the feel of a coffee cup as we rotate it to reach the handle.

Neurons and Connections

Hawkins proposes two tenets regarding neurons:

• Thoughts, ideas, and perceptions are the activity of neurons.
• Everything we know is stored in the connections between neurons.

These two tenets have profound implications for everything from how we think, to our concept of self, and even the existence of a free will. These neurons and connections are like the turtle that carries the world on its back. It’s neurons and connections all the way down.

Since movement is fundamental to evolution and survival, the brain also develops frames of reference for everything in a person’s world. When walking, where is your foot relative to the floor, the stair, the ground, and especially the root you just noticed a foot in front of you? When drinking coffee, where is the cup in relation to your hand, your mouth, and in relation to gravity? Is the coffee about to spill?

This concept of position within a frame of reference, almost a cortical GPS, turns out to have a literal manifestation in an older section of the brain, which does keep track of location in a grid system similar to the letter/number grid on a map. A separate set of neurons called place cells, keeps track of what is at each grid. The final element in the reference frame is orientation—which way am I facing? Together, they form a basis for physical predictions. For instance, If I am in the living room looking at the front door, I know the kitchen door is on my right.

This structure for a frame of reference easily generalizes to a structure for all knowledge, not just physical location. In fact, we often use the term ”frame of reference” specifically to refer to how a person understands something, the lens through which they perceive reality.

When the reference frame is weak or incomplete, the predictions are not very good. The only way to improve the predictions is to move through the reference frame, whether physically or conceptually, and usually some combination of both. Each movement in the reference frame creates a host of predictions about the future, the next location, and the next view within the frame. At the next moment, each prediction is tested against reality, and good predictions are reinforced while predictions that don’t match are adjusted, corrected if you will, to align with the perceived reality. This is exactly a learning machine. For humans to learn, the best way is to practice, practice, practice.

Hawkins also demonstrates how frames of reference can be nested recursively, to provide very compact, reusable elements of knowledge. For instance, language is made of up collections of letters, nested inside collections of words, nested inside collections of sentences, and nested inside collections of paragraphs.

In part 1 of the book, Hawkins extends this theory much further into the physical operation of the brain. Then, he extends it to the organization of people into enterprises such as business, with truly surprising insights into fault tolerance robust operations, development of consensus, focus and hierarchy, and effective collaboration, among others. Each set has a taxonomy, within which elements can be combined to produce unlimited results.

Part 2: Machine Intelligence

In this part, Hawkins reviews the current state of artificial intelligence and how the Thousand Brains theory might provide a direction to achieve artificial general intelligence or AGI. To date, even the triumphs of deep learning in the creation of AIs that have beaten humans at chess, Go, and Jeopardy are not truly intelligent. Each one is specialized and cannot generalize the learning to other areas, dare we say other frames of reference.

Hawkins proposes that the model for AGI systems should be the only AGI system we know—the human brain itself, specifically the model in the Thousand Brain theory. He suggests that AI development is following the path of computer CPU development. Initially, computer CPUs followed many specialized models for different tasks, such as encryption/decryption machines, decimal number calculating machines, and even analog computers, mostly for calculating trajectories of artillery shells. Eventually, the universal Turing machine came to dominate and universal, general-purpose computers emerged. While ARMS or X86 CPUs may not be the very best for any specific task, the powerful universality of a Turing machine has won the evolutionary race for computer CPU designs.

What will the universal AGI look like? Hawkins proposes four guidelines, each directly from the Thousand Brain theory:

1. Continuous Learning–The AGI never stops learning because the world never stops changing. The AGI continuously adapts to new inputs and new information.
2. Learning via Movement—The AGI learns by moving through frames of reference, or models of the world. It continuously makes and updates predictions based on how the inputs change as it moves.
3. Many models—The AGI moves through many models, or frameworks simultaneously—visual, auditory, linguistic, feel, etc. Similarly, the AGI moves through many conceptual models simultaneously, such as fluid flow, electrical, thermal flow, group dynamics, etc. Most environments and problems will be a simultaneous combination of several physical and conceptual spaces.
4. Store Knowledge in Frames of Reference—A frame of reference, or taxonomy, creates knowledge from a disparate collection of facts by virtue of its structure. For example, The wide variety of chemical interactions is difficult to understand as individual observations, but easy to understand and predict when the elements are organized in the Periodic Table of Elements. Many data observations make sense only when organized chronologically or as a statistical distribution. The frame of reference creates and stores knowledge.

Consciousness, Qualia, Emotions, Killer AGI, SETI and other Distractions

With this model of a successful AGI, Hawkins extrapolates how the model might apply to concepts such as consciousness, qualia, emotions, and other ideas that have troubled historic philosophers and modern AGI philosophers. His reasoning is a joy to follow, and his conclusions can be surprising.

Part 3: Human Intelligence

Human intelligence is the peak of evolutionary progress, at least it appears so to us humans. But, even with this intelligence, humans can go far off track. In the first two chapters of this part, Hawkins explores how and why humans can go off track.

Humans can easily believe things that are not true, often because our perceptions are limited. The model of flat earth actually worked very well for millennia when humans traveled by foot. The flat earth model even works pretty even well for car travel, except for long trips that cross time zones. Physically perceiving the curvature of the earth is very difficult, except for observant sailors.

Our perceived model of the earth as the the flat is not due to ignorance or stubbornness, but due to the design of our model, our frame of reference. In other areas, we have no direct observation at all. We cannot see viruses, bacteria, or even cells. We cannot perceive magnetic waves, electricity, atoms, or even most of the frequency spectrum. Visible light is only a narrow slice of the spectrum. With our intelligence, we have discovered worlds that are completely invisible to our direct perception.

Without direct observation to contradict and correct false models, these false models and beliefs easily take root and spread. With the advent of better faster communication and especially the internet, false beliefs can spread more rapidly.

This becomes an existential conflict between intelligence, represented by the neocortex, and prehistoric survival instincts, represented by the old, reptilian brain. The reptilian brain directly controls the body, including chemicals such as testosterone and adrenaline that the old brain uses to accomplish its hardwired goals to survive and reproduce. “Fight or flight” is truly hardwired into the reptilian brain. These behaviors are necessary for the genes to survive, but they are actively harmful to the survival of the human species now. In the fight between the neocortex and the old reptilian brain, the old brain almost always wins because it controls the body. With the tools our neocortex has created, from weapons of war to the engines that change the world, the old brain can kill humanity.

Hawkins calls this “The Big Idea.” In his words:

The point I want to make is that our intelligence, which has led to our success as a species, could also be the seed of our demise. The structure of our brain, composed of an old brain and the neocortex, is the problem.

Our old brain is highly adapted for short-term survival and for having as many offspring as possible. The old brain has its good side, such as maturing our young and caring for friends and relatives. But it also has its bad side, such as anti-s9cial behavior to garner resources and reproductive access, including murder and rape. Calling these “good” and “bad” is somewhat subjective. From a replicating gene’s point of view, they are all successful.

Out neocortex evolved to serve the old brain. The neocortex is a model of the world that the old brain can use to better achieve its goals of survival and procreation.

Hawkins explores three possible solutions:

Colonize new worlds, new stars. Aside from the technical challenges, this strategy simply spreads the old brain to more locations and perpetuates all its problems farther into the universe.

Modify Our Genes. This path is perhaps possible, and almost certainly uncertain and unknown. Most likely it leaves humanity with the flawed old brain still in charge.

Create intelligent machines and abandon our carbon bodies. This approach removes the old brain and the genes, thus eliminating the hard-wired genetic drives for procreation and survival. It puts the neocortex in charge.

Estate Planning for Humanity

After painting such a dark picture of the threat, Hawkins proposes two objectives which might worthy of humanity:

• Preserve knowledge.
• Create new knowledge.

Perhaps these may be the ultimate goals for a rational neocortex, which is driven by an insatiable curiosity to discover and learn everything, rather than the genetic drives of procreation and survival. Along this path, we and our non-carbon children might preserve enough knowledge to share with and benefit other intelligent beings.