In this in-depth interview with Numenta Co-founder Jeff Hawkins, host Matt Taylor dives deeply into concepts of location and object representation in the neocortex. In Part 1 of this 2-part interview, they discuss location, unique spaces, object compositionality & behavior, movement and learning, sequence memory, and the definition of “space” itself. In Part 2, they have an in-depth discussion on how HTM sequence memory builds object representations in space through movement.
For educational material regarding the topics discussed in this podcast episode, you have a few options. If you want to be fed the theory, watch HTM School. If you need more details, there are many papers to explore. If you want to talk to others about HTM theory, get involved in our community. For a primer on grid cells, have a look at this interactive blog post. For more on how we think grid cells work within the neocortex, watch Jeff’s recent talk, Does the Neocortex Use Grid Cell-Like Mechanisms to Learn the Structure of Objects?
- [1:36] Introduction of Jeff
- [2.02] Location in the brain – what have we learned over the past year?
- [3.05] Exploring the idea of grid cells
- [4.21] Jeff explains Numenta’s big idea related to grid cells: how a location signal is generated
- [7.24] An insight that Numenta Research Engineer Marcus Lewis had: every point in space is unique to a particular object
- [10.48] An idea we learned from grid cells that was not in our 2017 paper: locations themselves are unique
- [11.18] Compositionality: how we can represent an object as not just a set of features, but as a set of other objects
- [13.31] An implication of our latest research: Objects have behaviors
- [16.39] Misconceptions about hierarchy
- [18.40] Modeling objects, space, and abstract concepts
- [20.30] Object learning: a thought experiment
- [21.16] Sparse distributed representations
- [23.30] How grid cells interplay with minicolumns, with HTM systems
- [29.06] How grid cells in sensory input coordinate