[buzzovich]

You're right about substrate. But you're missing the point about utility.

I'm not debating the term is wrong. It IS wrong. I agree. THat's kind of my point. I'm saying a wrong term is yielding measurably positive results and it's being dismissed because it's not right.

For whatever reason, the term works. When "Cognition" is paired with Plato's modes and a single line of guidance, there's a big point improvement.

Measurable.

Replicable across models.

It seems, from what I can tell, to give the LLM the right lens to look through from the outset and therefore sends it down the right path early on.

If it's wax fruit, prove it. I am totally up for being proved wrong and someone showing me that those few words don't make a big difference.

[mallowfram]

It's wax fruit because we know the term cognition is bunk as intelligence:

https://pmc.ncbi.nlm.nih.gov/articles/PMC7415918/

[buzzovich]

Again, I agree. Cognition is way too vague a term.

You're actually making the argument FOR what I'm saying.

From what I can see, Buzsáki says "cognition" is bad for neuroscience BECAUSE it's philosophically loaded.

Doesn't that reinforce my point that it's good for LLM engineering PRECISELY because it's philosophically loaded?

Basically, if cognition is philosophically inherited then maybe that's why it's working for LLMs with Plato's model. Because training corpora are philosophy-heavy?