How the brain works and how the neurones link up and develop is something that has intrigued science throughout history. What makes a neural network intelligent? Is it genes behind this process (connectivity) or is it the environment (learning)? Today science is busy trying to create ‘artificial brains’. But is it really possible to create artificial intelligence, i.e.something that so far only exists in biology?
To date nobody has built a human brain, despite various the attempts. A baby’s brain has to grow together with the rest of its body, to develop from nothing but the ‘genetic blueprint’ it inherits from its parents. For this it needs both its genome, and time and energy to build the adult brain. It is a process that needs to take place in the right order, and at the right time.
The focus of this book is around this developmental process and is designed to highlight the different viewpoints from biologists and AI specialists. Is there sufficient agreement between the different disciplines to establish that we know how the brain is assembling itself, or do we still lack too many pieces of the puzzle?
To deal with these issues the author has created a fictitious discussion between a neuroscientist, an AI researcher, a developmental geneticist, and a robotics engineer. (One explanation for this fictitious discussion is the author’s own experience of not being able to have real-life discussions with experts from these different disciplines.) This technique enables the author to effectively highlight the differences between the disciplines: “The idea that information unfolding based on genomic information cannot be mathematically calculated, but instead requires algorithmic growth or a full simulation thereof, is a core hypothesis of this book.”(p7)
The attempt to understand how the brain is able to assemble itself takes quite a bit of brain power. It is not ‘just genes’ that contain the blueprint for how to build a brain. Nor is it algorithms alone that lead to each phase of the development. “As a brain develops, all neurones run through their programs in an enormous coordinated and dynamic jigsaw puzzle…. The neurone lives in a dynamic normality that is characterised by continuous feedback with its surroundings and changing contexts that always seem to just right.” (p 154-155)
To help us understand how the genome encodes the growth of a brain, the author draws on live observation of the much smaller number of neurones in the developing brain of a bee and a worm at the very moment their neural network self-assembles and becomes a brain.
One conclusion is that this brain development happens in close contact with the environment, i.e. both nature and nurture plays a role. Even if the smallest components of a neural pathway system follow an algorithmic behaviour, the end result is not completely predictable. The author compares it with the outcome of an individual football match versus the overall performance of teams in a football league. Overall we can see and predict a trend but that does not mean that we can do the same for each football match.
So where does the information come from and how does it get into the brain in order to shape it? It takes connectivity in the neural network as well as time and energy. The network also needs to learn. Another time and energy consuming activity. Exactly how this is done is where biological and artificial systems disagree. The biology argument is that it grows, the AI claims there is an ‘on’ switch.
An important question which this book addresses is; can a self-taught AI device achieve ‘human-level AI’, without genes and development? Elon Musk, for instance, is already talking about AI implants in the brain as the next big thing. The final answer may well have a great impact on what our future will look like.
Published by Princeton University Press, 2022, 364PP, Paperback, ISBN 978-0691181226