Good post... especially you attention to temporal context.

This one suggests my "rat boys". Tsao and moser?

Temporal context.

Memory...snipping nerons to freeze training.

Can be modeled as GLM.

Very efficient. Small...not too many neurons to simulate weights for connections.

Requires as you indicate a sort of timing circuit.

You have a great mind.. I enjoyed your post.

I myself work on ANN.. sort of dabble.

😜😁😎

I really like Dr fawad raufs work on local models.

I sort of view llm as soon to be extinct. 😎