Review: A Thousand Brains: A New Theory of Intelligence

A Thousand Brains: A New Theory of Intelligence by Jeff Hawkins
My rating: 3 of 5 stars

The book offers an intriguing exploration into an alternate architecture of intelligence, drawing from Hawkin's extensive background in neuroscience and computing (he helped design the Palm Pilot). The book is divided into three parts, each tackling a distinct aspect of brain function and its implications for artificial intelligence and beyond.

The book's first part is a compelling dive into the world of cortical columns. It focuses on the importance of the brain's cortex (the crumpled outer layer), which is thought to have evolved later than the older 'reptilian brain', which the author dismisses throughout the book (despite the fact it keeps him and other humans alive). Hawkins suggests that if you spread the brain's 'newer' neocortex out on a table into its 2.5mm thick extra-large pizza size, you would be looking at 150,000 similar 'columns' that extend through the six layers of the cortex. He builds on the ideas of Vernon Mountcastle, who, in 1978, put forth a hypothesis that a single unifying computational principle might be operating across the entire cortex. There are anatomical similarities across cortical areas, and some experiments have shown that these can be repurposed. If inputs from the optic nerve are diverted via the auditory thalamus to the auditory cortex, the auditory cortex can develop the properties associated with the visual receptive field.

Hawkins presents a detailed and engaging description of these structures, proposing that they function as self-contained predictive processing components in the brain. He further hangs his theory on the idea that each column has the equivalent of 'place' and 'grid' cells, thought to exist in the 'older' brain structures, that can help create a reference model for part of the external world. His explanation of how these columns work together to form a model of the world is both accessible and thought-provoking. This section alone makes the book a worthwhile read for anyone interested in the mechanics of the brain and the nature of intelligence. There are some areas of overlap and agreement with work done by Rajesh Rao, Karl Friston and Andy Clark - legends in the field.

However, the second part, which delves into artificial intelligence, falls short of the high standard set by the initial chapters. While informed by his understanding of brain function, Hawkins' insights into AI seem neither groundbreaking nor remarkably cohesive. The discussion meanders, lacking the clarity and depth that characterize the book's opening.

The final part of A Thousand Brains unexpectedly turns philosophical, and Hawkins' tone becomes markedly more strident. His alignment with Richard Dawkins (who wrote the fawning introduction) and the other New Atheists (Sam Harris, Daniel Dennett and Christopher Hitchens) is evident as he dismisses with unnecessary arrogance the faith of billions as a "viral false belief." This sweeping generalization feels out of place in a book grounded in scientific inquiry. Hawkins' approach here may alienate readers who hold religious beliefs or seek a more nuanced discussion of faith and reason.

A Thousand Brains is a book of mixed strengths. Its first part shines with insightful analysis and accessible explanations of complex neuroscience concepts. In my opinion Hawkins should have stopped here. The subsequent sections lack the same rigour and coherence, with the third part straying into an unexpectedly dismissive and dogmatic critique of religion. He also skips lightly over consciousness, clearly taking a dig at both Koch (Christof Koch) and David Chalmers and echoing the 'there is no hard problem of consciousness' akin to Anil Seth (Being You: A New Science of Consciousness).

Despite these shortcomings, Hawkins' exploration of cortical columns offers valuable food for thought, making the book a worthwhile, if uneven, read.

I gave the book three stars.


View all my reviews