Back to How Do We Learn Dec 2025 Aliveline

Day 14

What do I hypothesize/believe about the C. Elegans brain right now?

• Each individual’s brain has its own coordinate system to implement the same behaviors
  • “Macroscopic universality, microscopic degeneracy”
• There exists a locomotion population subspace; projecting each worm brain into its locomotion subspace & associated manifold allows us to predict locomotion
  • Why do I think this?
    • It feels that there has to be some universal coordinate system across worms when they are developmentally identical – all of them move and it doesn’t feel likely that all of their brains independently happen to discover movement
    • Natural selection works at the level of behavior, so it makes sense that there would be some way to universally compare key behavior like locomotion across worms

  • What would alternative explanations of the above look like?

• The locomotion manifold has a conserved shape across worms at the start of their life (it is somehow developmentally canalized), but it changes as the worm learns & based on the context/task
  • How does the manifold change?
• In freely moving worms in their natural contexts, the locomotion subspace/manifold is just one of possibly many subspaces/manifolds present as the worm is doing many things, and they interact
  • How do they interact?

The brain is tiny (302 neurons). As per the Liang paper it has 1-4 neurons per cell type, so little to no neuron redundancy. How might it do error correction / avoid catastrophic forgetting?

• Their proposal is “context-gating”, ie encode something about the context+stimulus combo rather than changing the global neuronal response for a stimulus
  • I think all brains obviously do this?