Jeff Hawkins on the Neocortex, Thousand Brains Theory, and the Path to True AI

Jeff Hawkins, founder of Numenta and the Redwood Center for Theoretical Neuroscience, articulates his conviction that understanding the human brain is not merely an academic pursuit but the most direct and fastest route to achieving truly intelligent machines. He posits that current artificial intelligence paradigms, such as deep learning and convolutional neural networks, possess inherent limitations and cannot bridge the "huge gap" to human-level intelligence without incorporating the fundamental principles by which the brain operates. His research primarily focuses on reverse-engineering the neocortex, the largest and most recently evolved part of the mammalian brain, which he believes is the computational substrate for all higher cognitive functions. Hawkins emphasizes the neocortex's remarkable structural uniformity across its entire expanse and across different mammalian species. This observation underpins the concept of a "common cortical algorithm," suggesting that the neocortex operates on a singular computational principle, with its specific function (e.g., vision, hearing, language) determined by the sensory inputs it receives. He contrasts this with the older, more specialized parts of the brain responsible for autonomic functions, basic behaviors, and emotions. Hawkins also discusses his earlier theoretical framework, Hierarchical Temporal Memory (HTM), which highlighted the crucial roles of time-based processing, the brain's ability to learn and store a comprehensive model of the world, and hierarchical organization in intelligence. While still foundational, HTM has evolved into a more comprehensive understanding, such as the Thousand Brains Theory. Hawkins characterizes the current state of neuroscience as a "pre-paradigm science," rich in empirical data but lacking a unifying theoretical framework to assimilate and explain it. His team's approach involves using this vast body of empirical data as strict constraints for developing theories, arguing that a theory's ability to explain numerous disparate observations simultaneously provides high confidence in its correctness. He likens these theoretical breakthroughs to historical "aha moments" in science, where complex problems suddenly become clear, drawing parallels to discoveries by Copernicus, Darwin, and Watson & Crick. This rigorous, constraint-driven methodology allows for easy falsification and refinement of theories through collaboration with experimental labs. The broader implications of Hawkins' work are profound for the future of AI, advocating for a shift from purely empirical, black-box machine learning to biologically inspired architectures. By deciphering the neocortex's computational principles, his research aims to unlock the secrets of general intelligence, paving the way for machines that can learn and adapt with human-like versatility. He views the evolutionary development of the neocortex as a significant leap, underscoring its unique capacity for learning complex models of the world, including the temporal aspects of objects and events. This perspective offers a compelling vision for a future where AI development is deeply intertwined with a fundamental understanding of biological intelligence.

my primary interest is understanding the human brain no question about it but I also firmly believe that we will not be able to create fully intelligent machines until we understand how the human brain works

I think the quickest way of bridging that gap is to figure out how the brain does that and then we can sit back and look and say oh what do these principles that the brain works on are necessary and which ones or not

There's not have been a single thing we've ever humans have ever put their minds to so we've said oh we reached the wall we can't go any further it just people keep saying that

the neocortex it's extremely uniform it's not visually or anatomically or it's very sucky I always like to say it's like the size of a dinner napkin about two and a half millimeters thick and it looks remarkably the same everywhere

all the evidence we have and this is an idea that was first articulated in a very cogent and beautiful argument by a guy named Vernon mal Castle in 1978 was that the neocortex all works on the same principle

it's data without it's in the language of Thomas Kuhns a historian it would be a sort of a pre paradigm science lots of data but no way to fit in together

we are blessed with the fact that we can test our theories out the ying-yang here because there's so much on a similar data and we can also falsify our theories very easily which we do often

when the answer comes to you and everything falls into place it's like oh my gosh that's it that's got to be right

I got interested in brains by reading an essay he wrote in 1979 called thinking about the brain and that is when I decided I'm gonna leave my profession of computers and engineering and become a neuroscientist just reading that one essay from Francis Crick