A Thousand Brains

• brain is composed of a new part called the neocortex and comprising 70% of of size, and an old part made of distinct regions
• neocortex learns a model of the world
• old brain generates emotions and more primitive behaviors
• old brain takes control, causing us to act in ways we know we shouldn’t, making us also susceptible to false beliefs that can be viral

cortical columns

• first proposed by Vernon Mountcastle’s ideas in The Mindful Brain in 1982

“Put shortly, there is nothing intrinsically motor about the motor cortex, nor sensory about the sensory cortex. Thus the elucidation of the mode of operation of the local modular circuit anywhere in the neocortex will be of great generalizing significance.”

• Mountcastle proposed that throughout the neocortex, columns and minicolumns implemented a fundamental algorithm that is responsible for every aspect of perception and intelligence
• all manifestations of intelligence are manifestations of the same cortical algo
• the neocortex learns a model of the world and makes predictions based on its model

two tenets of neuroscience

1. thoughts, ideas, and perceptions are the activity of neurons
2. everything we know is stored in the connection between neurons (synapses)

existing view of neocortex

• neocortex is hierarchical, like a flowchart
• sensory input is processed in a series of steps, each step extracts more and more complex features from input
• continues until some region detects the complete object
• problem: treats things like vision like a static process, as if goal is to take a single picture

new view of neocortex

• all regions are capable of recognizing complex objects
• neocortex has many models of object, in different columns – not identical but complementary
• prevailing model is achieved by a form of voting amongst cortical columns, which is a solution to the binding problem (which sensory input defines the object)

3 key discoveries

1. neocortex learns a predictive model of the world and is always testing those predictions
2. predictions occur within neurons
3. the key to cortical columns are reference frames, by associating sensory input with locations in reference frames

4 parts of hypothesis of all knowledge using reference frames and thinking is a form of movement

1. reference frames are present everywhere in neocortex
2. reference frames are used to model everything we know, not just phyiscal objects
3. all knowledge is stored at locations relative to reference frames
4. thinking is a form of movement

Thoughts

The first part of this book details the Thousand Brain hypothesis, the second part goes through current state and future state of artificial intelligence, and the last part is a futurists plea for the salvation of humanity from existential threats and the tyranny of the old brain.

Old and new brain seems to map to some of the other psychology stuff I’ve read (id, ego-superego, or system 1 and system 2 thinking maybe). Also, haven’t really been able to wrap my head around thinking being a form of movement – what are the implications? Tried to pick up object and hold it perfectly still with eyes closed. If I didn’t know what the object was, would I be able to recognize without moving anything? What is movement? Seems to be bound in the types of cells we see in the brain, the grid-cells, and so perception requires a stream of info rather than a photograph in an instance.

Also, the AI chapters are interesting – is it really true that the only intelligence that could achieve articial general intelligence (AGI) would be something that mirrors the way human intelligence works? Might there be other ways to achieve “intelligence”? Hawkins never explicitly defines intelligence, just sketches around it. The idea loosely might be that intelligence is the ability to learn models about the world and make predictions around those models, and how intelligent you are is the speed and accuracy of those skills.

His chapters about uploading the brain are also fascinating thought experiments, although Hawkins is not a philosopher. I think the concept he is missing from uploading a brain is the continuity of self – the analogue to Theseus’s ship where if you replace all the parts of Theseus’s ship, and take those parts and build a new ship, is it still Theseus’s ship? All of this as an attempt for extending lives – some feasibility maybe to cybernetic existence, or extending our biology, which we have historically already done – hip replacements, heart surgeries, vitamins, etc.

I think the book gets less strong in part 3 – while I agree that viral false beliefs might be an existential threat, Hawkins seems to make knowledge a first-class citizen, as if, if the human race were extinguished, it is our knowledge that is important. If we built a fleet of AI robots and they continued our existence, I don’t think humans would figure that as a win.

Genes vs knowledge is also an interesting chapter – the goal of genes, and largely evolution (escaping Darwin’s orbit) is replication, and all that matters is how well a gene replicates. Knowledge is different – it is fundamental, a potential meritocracy. While knowledge can be memetic (viral) and shared through culture, the march of knowledge is towards truth. We will only make it so far as a species controlled by our old brains and the inexorable drive for our genes to replicate.

One other aside, I really liked the idea of both a wiki earth (a long-lived satellite orbiting the sun that we send our collective data to) or the idea of orbiting satellites around a star to send messages. Feels like an idea I saw in Cixin Liu’s work?

Brief Takeaway

As Richard Dawkins says in the foreword, this book “will turn your mind into a whirling maelstrom of excitingly provocative ideas”. I don’t know much about the brain and how intelligence works, but after having read this book (and Hawkin’s first book, On Intelligence in the past), I want to learn more. I’m still trying to wrap my head around the idea that thinking is movement, although, regardless of it’s veracity, I love the idea that the brain has thousands of models for every object vying for election to “true” (our perception is just a model!). Worth the read.