Skip to main content Link Menu Expand (external link) Document Search Copy Copied
I2 Intro Neuro/AI

Unit 11: Building a Brain: Networks and Systems

It would be negligent not to begin by acknowledging the inadequacy of this chapter. The connectivity and wiring of the brain as well as the genetics and epigenetics that comprise it are far from fully understood, and these resources are merely scratching the surface of a profound ocean of literature and undiscovered territory of building the systems of a mind. Nonetheless, we begin with a lecture investigating the genetic and learned nature of human behaviors.

Task 1: 10. Development, Nature & Nurture I

Synthesis Questions:

  • What is one possible mechanism for building knowledge from parts of the brain present at birth? What kinds of knowledge might be innate and "precursory"?
    • Why might it be useful for those precursors of knowledge to be innate rather than learned?
    • Google and find 1-2 other possible rudimentary or innate processes in the brain. Do any machine learning techniques use similar innate predispositions or ingrained knowledge?
  • What is perceptual narrowing? Are there analogous processes, protocols, or side effects in machine learning? What are possible downsides and upsides of the effect?
    • What might you wish you could learn faster, and what might be a cost to having neural architectures that support that kind of learning?
  • If humans are uniquely suited to recognizing faces, what might be a useful analogous skill in machine learning which may support more human-like behavior?

Task 2: Next, briefly skim the questions below then indulge in this scintillating lecture describing brain modules and the discovery of their connected wiring!

Start watching at 8:40, 21. Brain Networks

For your reference, a voxel is a 3D pixel. Think of it as a minecraft block of the brain. Also a term used in computer graphics.

Synthesis Questions:

  • What is white matter? Why does it matter?
  • What is a connectivity fingerprint?
    • Why might we care about the connectivity similarities of other species and what implications might findings in this research have?
  • Describe three large connections between major regions in the brain and what those regions do, using Google if necessary.
  • What is the default mode network? What purpose is it hypothesized to serve?
    • If you were to build an analog of the DMN in a machine learning network placed in your favorite video game as an environment, what function would you design it to serve?
  • Hypothesize how multiple demand regions might work and why they might be necessary. I don't know the right answer here so go nuts.

Project Spec:

There is no programming for this project. Instead, we have provided a LaTeX template for you to fill out.

  • If you are unaware of what $\LaTeX$ is, you can read about it here.

I would recommend taking a look at Overleaf to edit/compile $\LaTeX$ code. Simply copy the code in the template into a blank Overleaf project and type your answers into the TODO areas. Be sure to hit the “Recompile” button to see your work.

GH Link: Unit 11 Template (40 min)

The questions in the template are also written below:

Ohhh this one’s good. I’m tapping my fingers together like Dr. Evil.

You are designing an egg. It will become a lifeform on a new planet. The planet is almost completely covered in water. Small islands made from underwater volcanoes dot the surface. The gravity is stronger on this planet and the biodiversity is lower than on Earth. There is less direct sunlight and fewer living organisms in general.

Your task is to design the brain and the phenotypes of a successful organism on this new planet, using some of the neuroscience and machine learning principles you have learned so far. Your designed brain and nervous system will be translated into genetic instructions and inserted into an egg that will be sent to the new planet. Be as specific as you can so your egg doesn’t become a mutant!

If you are stuck, here are some options for getting started

  • What do you consider a successful organism?
  • What kinds of traits would support this success; what kinds of organisms on Earth have these traits?
  • What parts of the brain support these traits and what kind of machine learning algorithms might also exhibit these abilities?
  • How might these different traits and abilities need to work together, or connect?