My book review this week is on Jeff Hawkins’ book, A Thousand Brains: A New Theory of Intelligence. Hawkins, a neuroscientist, engineer, and businessman, co-developed Palm Computing and created some of the earliest smart phones. After he sold his interest in the Company, he has pursued his passion to solve the mystery of how the brain works. This book is quite complex and I won’t try to dig deeply but I read it because Bill Gates praised the book saying it contained “tantalizing clues about the future of intelligent machines.”
The book is about the old, elsewhere sometimes referred to as the reptilian brain, versus the new brain. Hawkins spends considerable time trying to explain a very complicated mystery and was clear that he has not yet solved all of the brain’s mysteries.
Think of the neocortex, or new brain, if it could be laid flat, as a thick dinner napkin. Different sections do different things and each region is “divided into thousands of columns.” From there, think of each column as a collection of tiny mini-columns with “over one hundred cells each.” The neocortex wraps around the old part of the brain and “is like 150,000 short pieces of spaghetti stacked vertically next to each other.” (p. 40)
One of the biggest discoveries that Hawkins made is that the brain makes a reference map (or model) of everything it knows, stores that data in reference frames, and uses that data to make predictions. Reference maps are incomplete and sometimes wrong, and so people can easily think they know things that are incorrect. Several more takeaways include:
- The brain learns through movement. It is only through movements (sometimes small) that the brain can create new reference frames, which Hawkins thinks are the secret to understanding how intelligence works.
- According to Hawkins, most existing neural networks in computers do not make these models, which is a limitation that likely needs to change before we can get to true generative AI. And deep learning machines cannot continue to learn while they are doing. They have to be continuously trained.
- For those who wonder about the seemingly random thoughts that come to them at strange times, Hawkins says they are not random. “What we think next depends on which direction we mentally move through a reference frame, in the same way that what we see next in a town depends on which direction we move from our current location.” (p. 99)