Pages in this blog

Thursday, August 15, 2019

Jeff Hawkins on the brain [Thousand Brains Theory of Intelligence]

Hawkins invented the Palm Pilot and then took his money and became a neuroscientist (he founded Numenta). Steve2152 over at Lesswrong summarizes a podcast Hawkins gave about his Thousand Brains Theory of Intelligence (cf. my post on the Busy Bee Brain). Here's a couple paragraphs I found particularly interesting:
He brought up his paper Why do neurons have thousands of synapses?. Neurons have anywhere from 5 to 30,000 synapses. There are two types. The synapses near the cell body (perhaps a few hundred) can cause the neuron to fire, and these are most similar to the connections in ANNs. The other 95% are way out on a dendrite (neuron branch), too far from the neuron body to make it fire, even if all 95% were activated at once! Instead, what happens is if you have 10-40 of these synapses that all activate at the same time and are all very close to each other on the dendrite, it creates a "dendritic spike" that goes to the cell body and raises the voltage a little bit, but not enough to make the cell fire. And then the voltage goes back down shortly thereafter. What good is that? If the neuron is triggered to fire (due to the first type of synapses, the ones near the cell body), and has already been prepared by a dendritic spike, then it fires slightly sooner, which matters because there are fast inhibitory processes, such that if a neuron fires slightly before its neighbors, it can prevent those neighbors from firing at all.

So, there are dozens to hundreds of different patterns that the neuron can recognize—one for each close-together group of synapses on a dendrite—each of which can cause a dendritic spike. This allows networks of neurons to do sophisticated temporal predictions, he says: "Real neurons in the brain are time-based prediction engines, and there's no concept of this at all" in ANNs; "I don't think you can build intelligence without them".
You can run on over to Numenta for a paper on this theory,  A Framework for Intelligence and Cortical Function Based on Grid Cells in the Neocortex:
Rather than learning one model of the world, the Thousand Brains Theory of Intelligence states that every part of the neocortex learns complete models of objects and concepts. Long range connections in the neocortex allow the models to work together to create your perception of the world.
Here are the high points of the theory:
  • Every cortical column has a location signal that we propose is implemented by grid cells.
  • We propose an extension of grid cells, called “displacement cells”. Displacement cells enable us to learn how objects are composed of other objects, also known as object compositionality.
  • Learning an object’s behavior is simply learning the sequence of movements tracked by displacement cells.
  • A location-based framework can be applied to concepts and high-level thought in the same way it can to physical objects.
  • We discuss how the “what” and “where” pathways of the brain can be thought of as performing the same computations, but modeling different object centered and body centered location spaces.
  • Our hypothesis that every cortical column can learn complete models and the brain creates thousands of models simultaneously, rather than one big model of the world, leads to a rethinking of hierarchy in the cortex. We refer to this idea as the Thousand Brains Theory of Intelligence.
Color me skeptical about location signals, but YES to the idea that concepts and high-level thought are implemented in the same neural constructs as physical objects (and processes). 


    No comments:

    Post a Comment