This paper is a companion to our research paper, “A Framework for Intelligence and Cortical Function Based on Grid Cells in the Neocortex,” which introduces a new theory for how the brain works. It is designed to explain in plain language The Thousand Brains Theory of Intelligence and the discoveries detailed in the research paper. Some may choose to read this companion piece alone, while others may use it as an entry point or guide to understanding the details in the research paper.
Rather than learning one model of the world, the Thousand Brains Theory of Intelligence proposes that every part of the neocortex learns complete models of objects and concepts.
Frequently asked questions
Q. What was the purpose of this paper?
How the brain works is a mystery that scientists have been trying to solve for more than a hundred years. Despite the details that scientists have amassed about the brain, there is still no accepted theory as to how it works and how the brain produces intelligence. The research paper proposes such a theoretical framework, which we believe can have significant implications across neuroscience, disease, education, and AI.
We felt that the ideas and implications in the research paper are significant, and we wanted to make them broadly accessible, so we created this companion piece.
Q. Who should read this companion piece?
We created this piece specifically for non-neuroscientists, though we invite anyone to read it. We’ve simplified some of the explanations to focus on the big ideas. Those looking for detailed explanations will want to read the research paper.
Q. What is the key takeaway?
The research paper proposes a basic explanation, or “framework”, for why cortical columns are complex and how they function. In brief, we propose that every cortical column learns models of complete objects through movement. We call this idea the Thousand Brains Theory of Intelligence because if every individual column is learning complete models, then our brain is not building one mega-model of an object, but rather thousands of models of the object in parallel. This hypothesis leads to a rethinking of hierarchy in the brain. If every cortical column is capable of learning complete models, then object recognition does not occur solely at the top of the hierarchy. It occurs in every column, at all levels.
Q. What are the implications?
We think the Thousand Brains Theory will have significant implications for neuroscience and AI. Our framework can offer experimental neuroscientists the opportunity to validate our hypotheses, fill in more details, and ultimately progress our understanding of how the brain is intelligent.
It offers AI and machine learning enthusiasts a model that may help them overcome the limitations of current techniques. We believe that truly intelligent systems require the ability to learn through movement, which is at the core of the Thousand Brains Theory.
Q. Where can I find out more?
We have several additional resources for people who want to learn more:
- A Framework for Intelligence and Cortical Function Based on Grid Cells in the Neocortex
- Locations in the Neocortex: A Theory of Sensorimotor Object Recognition Using Cortical Grid Cells and A Theory of How Columns in the Neocortex Enable Learning the Structure of the World – These two papers describe detailed network models for components of the framework.
- HTM School Episode 15: Framework for Intelligence – Host Matt Taylor breaks down some of the key ideas introduced in the paper in this video that’s part of our educational series.
- A video presentation of a talk that covered much of the theory: Jeff Hawkins at Johns Hopkins APL Colloquium Series
- Numenta On Intelligence Podcast: Research Update with Jeff Hawkins – Parts 1 and 2 – Jeff discusses many of the core concepts in the paper, including location and location spaces; object representation, compositionality & behavior; and movement and learning.
- HTM Forum – a great resource for further questions and discussion on the paper. The authors are active participants in the forum.