On: Evidence of a predictive coding hierarchy in the h...

[IIAOPSW]

    |-----long timing loop / top of parse tree-----|
    |                                              |
    |-shorter / child node -|                      |-shorter / child node-|
    |                       |                      |                      |
    |highest freq|          |highest freq|         |highest freq|         |highest freq|

[politician → IIAOPSW]

Similarly, ChatGPT (gpt2-medium) has 12 heads of attention. GPT-3 has 96.

[IIAOPSW → politician]

I don't know how superimposed waves in finely tuned timing loops with non-linear interference translates into heads of attention, and honestly I suspect a lot of the things difficult to do with heads of attention (and other approaches of the past) come for free in a resonance based system.