I actually think that human language is unreliable at expressing what's going on inside a persons mind[1]. My native language is not English, I have only introductory-level knowledge in the field of pragmatics[2], which makes me fully aware of the many ways in which I could fail to write a compelling sentence to support my argument. I can use language to only approximate thoughts in my head, and when it comes to abstract concepts and ideas, words alone, I assert, are never sufficient. It isn't even necessary to step outside our main knowledge area to illustrate this point. How many around here have read a Monad tutorial, without any hands-on experience, and how many of those have understood what Monads are/or how they work from words alone?
My entire paragraph from before, just to set the stage on a simple question. How can you even formulate a question, for a multi-billion parameter language model, to evaluate that it can understand in an abstract/conceptual way something/anything? Heck, how can you do that with other people? I think if we'd have an answer here, we actually could evaluate easily experience/expertise with anyone we'd interview; instead of requiring credentials, references, tests, trials, etc.
[1] Lots of poets from the romanticism era liked to touch upon this topic. One that comes to mind, and one of my personal favorites, is Silentium by Fyodor Tyutchev https://culturedarm.com/silentium-by-fyodor-tyutchev/