An Essay review of Jeff Hawking's book, A Thousand Brains: A new theory of intelligence

Elements of a Paradigm Shift in thinking about thinking

While neuroscience languishes in a holding pattern, forever seeming to be hovering over the intractable mind-body problem, Jeff Hawkins, an Engineer turned self-taught neuroscientist, plows full-steam ahead.

Using his engineering knowledge and "constraint-guided" deductive thought experiments (piecewise confirmed by simulation analysis), Hawking makes subtle but powerful refinements to his pattern recognition-prediction framework introduced in "On Intelligence."

This, "A Thousand Brains," his second book, Hawking, following in the footsteps of Nobel Laureate Santiago Ramon de Cabajal and Vernon B Mountcastle, proceeds step-by-painful-step to discovering that the neocortex makes thousands of copies of the same brain circuits displayed in de Cabajal's Nobel prize winning diagrams of cortical mini-columns found deep within the structure of each neuron.

These dendritic-driven circuits are the brain's mini-processors. They receive inputs and make predictions. The circuits are repeated for every one of de Cabajal's 150,000 columns across every sensory region of the neocortex.

Their most important function is to use their locally-based predictions to prime parent neurons into casting their vote for the most likely whole, integrated systemwide prediction based on the parent neuron's own higher level shared inputs.

The single coherent integrated prediction that results, defines what we know about the world. As neurons are activated, they become our thoughts, consciousness, indeed our mental image of the outside world.

And, as startling a discovery as this may seem, it is not the crowning achievement of this important mid-course correction to Hawking's earlier theory.

The crowning achievement is discovering that in the course of making piecewise predictions, each neuronal circuit creates its own reference frames responsible for building these hundreds of thousands of sensory-mode-specific prediction models of the outside world.

What does it all mean?

If these updates hold up under peer reviewed scrutiny, modestly, Hawking will have successfully launched a new theoretical paradigm of how we think, and of how the brain becomes intelligent.

The neuroscience community then will have no choice but to give up pursuit of the phantom mind-body problem, and accept that most of the cells in our neocortex are dedicated to creating and manipulating the reference frames that the brain uses to build models for planning and thinking.

These hundreds of thousands of reference frame-generating mini-processors, store everything we know throughout our lives as a systemwide distributed network of dynamically changing neuronal connections.

The upshot of Hawking's work is that there is no longer a mind-body problem because there is no mind, only a body.

It is neuronal activity all the way down.

Thus, having cleared a path across a minefield of difficult constraints, Hawking's refinements to his "On Intelligence" pattern recognition prediction-framework, fills-in the brain's picture of the world.

And, in doing so, begins to provide answers to some of neuroscience's most difficult questions; such as: How do our varied sensory inputs get unified into a singular experience? What is happening when we think? What is consciousness? How can two people reach different beliefs from the same observations? What is abstract thinking? How does the brain process language? And why do we have a sense of self?

A Few of the Important Details

The first person to look carefully at the details of the circuitry inside the neocortex was Santiago Ramon de Cabajal who received a Nobel prize for uncovering the structure of neocortical columns. The first thing de Cabajal observed was that neurons are arranged in layers according to differences in size, how closely packed they are, what they are connected to, and how they behave.

As we learned in "On Intelligence," inputs enter at the bottom rung of de Cabajal's six-layer hierarchy. And whenever predictions are confirmed, they become apart of the distributed knowledge base of our experiences. If not, they ascend the cortical ladder learning and becoming more abstract at each higher level. At the highest end of this hierarchy lies language and generalized abstract thinking like mathematics.

As Mountcastle had forewarned in one of the most important papers in the neurosciences ("An Organizing Principle for Cerebral Function: The Unit Module and the Distributed System." A reprint can be found in "The Mindful Brain," by Gerald M Edelman and Vernon B Mountcastle, MIT Press, 1982), that although circuits look the same everywhere in the neocortex, this may in fact be telling us something important: that every region in the neocortex is using the same algorithm for carrying out its respective functions, functions that probably do something far more complex and more important than just extracting sensory related features.

And thus, like a dog chewing on an old bone, Hawking grabbed hold to this idea and refused to let go until he discovered what this more complex and more important secret was.

After fives years of toiling away in the wilderness, using thought experiments and computer simulation studies primarily, he finally hit pay dirt.

It turns out that Mountcastle's secret is this: The connections in our brain recognize patterns that make predictions, and then stores them as distributed dynamic reference-frame-based models of the world, models created, learned and stored through everyday experiences.

With every experience, new pieces of knowledge are added to the model by forming new synapses. The small percentage of neurons that are active at any point in time represent and account for our perceptions, thoughts, and consciousness.

In the end, they all are only what it feels like to have an active neocortex. The rest, when the neocortex is inactive, is the memory that makes up the distributed knowledge-base of our stored experiences.

The author, thus in the end, uncovered Mountcastle's secret by simple deduction: The neocortex stores everything we know as a systemwide distributed network using map-like reference frames to build its prediction models.

The brain uses these models to engage in survival enhancing thoughts. Intelligence is how we learn to use these models to enhance our survival.