My Notes About On Intelligence Book By Jeff Hawkins. Part 2.

Thu Sep 17 2020 23:12:01 GMT+0200 (Hora de verano romance), David Ibáñez

My Notes About On Intelligence Book By Jeff Hawkins. Part 2.

on intelligence.jpg

Amazon page of the book

"On Intelligence"" By Jeff Hawkins @JeffCHawkins on Twitter and Sandra Blakeslee @bysblakeslee is a very inspiring book for me.

This post contains the second part of my notes about the book and includes the chapter 3 of the book.

You can find the Part 1 Here

Chapter 3. The Human Brain

Intro

  • Intelligence resides in neocortex.

  • A human is much more than an intelligent machines. An intelligent machine need not have sexual urges, hunger, a pulse, muscles, emotions, or a humanlike body.

  • Main intelligence regions: Neocortex with important roles played by other regions like thalamus and hippocampus.

Neocortex

  • 2mm thick, 6 layers. Extended size-like a large dinner napking.
  • Humans are more intelligence than other animals because the size of our neocrtex relative to body size.
  • Estimation: 30.000.000.000 neurons.
  • Has an concrete structure, is not just an amourphous blob of cells.

Neocortex malfunctions

  • Parietal lobe (right): Lose the ability to perceive anything on the left side of his body and space around (left half).
  • Left frontal region (Brocca's area): Lose the rules of grammar, but keeps the ability to understand words
  • Fusiform gyrus: Lose ability to recognize faces.

Functional areas

  • Same cortical sheet

  • What makes one region “higher” or “lower” than another is how they are connected to one another.

  • Lower areas feed information up to higher areas by way of a certain neural pattern of connectivity.

  • Higher areas send feedback down to lower areas using a different connection pattern.

  • Lowest Functional Areas: Where the sensing stimulus arrive. Process the information at raw, most basic level.

  • Vision: V1, V2, V4 and IT.

    • V1: Basic features, tiny edge-segments, smallscale components of motion, binocular disparity (for stereovision),
    • V1 feed to V2, V4 and IT
    • V4: Medium complexity.
    • MT: Motion of objects.
    • In the higher echelons of the visual cortex are areas that represent your visual memories of all sorts of objects like faces, animals, tools, body parts, and so on.
  • Audtory: From A1 to higher levels.

  • Somatosensory: From S1 to higher levels.

  • All these regions goes further to association areas.

  • Motor output: Sends connections to spinal cord, driving muscles. Follows the same path but in the opposite direction. From higher regions sophisticated motors commands to M1.

Feedback

We have to add feedback in the whole process. There're more feedback connection than forward.

Noecortex layers

Formed by a variation in the density of cell bodies, cell types and connections.

Neurons

  • Structure: Dendrites, cell body and axon.
  • Axon -> Dendrite: Synapse. Can be excitatory or inhibitory.
  • New synapses can be formed.
  • Hebbian learning: Increasing strength of these connections (synapses) due to neurons spiking nearly or at the same time.
  • Formation and strength of synapses = memories.
  • 80% of neurons are pyramid cells. Several thousand synapses. It varies in cells, layers and regions.
  • Connections can be in the neighbourhood or lengthy (To other regions or other areas like thalamus)

Neocortex structure

Vernon Mountcastle: “An Organizing Principle for Cerebral Function.” - Paper 1978.

  • Mountcastle says: Neocrotex is uniform in appearance and structure. All regions are similar. All regions pergorm the same basic operations. Same computational tool for everything. What differs from one region to another is the way they are connected with one anoother and to other parts of the brain. A common algorithm for all the neocortex.

  • Brain regions specialized functions based on the wind of the information that flows into during the development.

  • Genes dictates the overall structure and the specific of what regions are connected together, but is highly flexible.

  • Different types of inputs, all end up being the same. All inputs ends up being spiking (action potentials) patterns. Perception and knowledge about the world are build from these patterns. The brain is the only part of the body thad has no senses.

    • Vision: Million fibers optic nerve -> Thalamus -> V1
    • Auditory: 30K fibers auditory nerve -> Older brain -> A1
    • Touch: Million fibers through spinal cord -> S1

Evolution

  • Brain is not born with special functions.
  • It learns. It can adapt to environments didn't exists until very recently.
  • Extremely flexible.

Spatial and temporal patterns

Vision

  • Spatial patterns: Coincident patterns in time. Vision creates spatial patterns in the brain, not copies of the real images (vectors of information). There're no pictures in the brain.
  • Temporal patterns: Pattern changing over time. Vision is more like a song than a painting.
    • Saccades: Auto eye movement. 3 times per second.

Hearing and touching work as spatial patterns and temporal patterns. Vector of frequencies agt the same time. We feel touch by changing what we sense. We need to move our fingers over things and surfaces to recognize what they are.

Sense. More than 5

  • Vision: Motion, color, luminance.
  • Touch: Pressure, temperature, pain, vibration.
  • Proprireceptive system: Joint angles and body position, balance.

Cortex doesn't really know or sense the world directly. the brain is in a dark quiet box with no knowledge of anything other than the time-flowing patterns on its input fibers. Brains are pattern machines.

  • Flexible feeling of self: If we use a tool for grasping we can feel soon that is your own body.
  • Brain could work with any other input type of sense with no problem like radar or sonar.

Building intelligent machines: If we can descipher how neocortical algorithm works, we can add any kind of sense to it.

Setting parameters -> train and educate -> billions of neurons taking care of complex thoughts.

Sources and other resources

Amazon page of the book

Jeff Hawkins @JeffCHawkins on Twitter

Sandra Blakeslee @bysblakeslee on Twitter

Numenta Homepage

About me

I'm David Ibañez from Barcelona. Feel free to get in touch, here or:

  • You can get in touch with me on Twitter

Tags

Follow us

Latest blogs
My Notes About On Intelligence Book By Jeff Hawkins. Part 2.
Thu Sep 17 2020 23:12:01 GMT+0200 (Hora de verano romance)
My Notes About On Intelligence Book By Jeff Hawkins. Part 1
Fri Sep 04 2020 15:49:05 GMT+0200 (Hora de verano romance)
New Blog Section... Neuroscience
Wed Sep 02 2020 19:43:02 GMT+0200 (Hora de verano romance)
My Notes About Intro to Tensorflow Course
Sat Jun 22 2019 19:43:02 GMT+0200 (Hora de verano romance)
How to train YOLOv3 using Darknet on Colab notebook and speed up load times
Wed Apr 10 2019 14:57:55 GMT+0200 (Hora de verano romance)

Latest comments
Hi David, Thank you very much, this blog saved my life!
mai
Thu Apr 16 2020 11:06:33 GMT+0200 (Hora de verano romance)