Reviewing 'A Brief History of Intelligence'

How would you figure out how to make intelligent machines? The historical precedent for building intelligence sounds like a good start. What if that precedent is a disordered mess of papers and insights across several fields and decades that may require a couple of textbooks to start to understand? In that case, you organize the mess into a neat evolutionary story, filling gaps with proposed hypotheses along the way. Max Bennett did a fantastic job tidying up his story.

Bennett’s book is a delightful tour of the evolution of intelligence through the specific path from the first animals to homo sapiens. Other forms of animal intelligence are mentioned but not covered in the book. The framework depends on identifying 5 breakthroughs in the evolutionary history of human intelligence, and pondering which one will be the 6th. Any excuses for metacognition are welcome, but this book is a particularly fun one.

More than fun, it’s an impressive endeavor. As far as I know, there’s no similarly accessible synthesis on intelligence out there. It successfully integrates work in reinforcement learning, linguistics, biology, and neuroscience. The proposed 5-breakthrough path is the result of a lot of reading and deep familiarity with the research discussed.

This book is likely my favorite of 2024. It doesn’t feel as outdated as it could be given the pace of robotics advances this year and accounts for the capabilities LLMs had as of July 2023. I hope Bennett decides to publish more books, for this is a great start.

This review is also on Goodreads.