> It's mean spirited, but the phrase "she's what stupid people think smart people sound like" struck me at the time.
The phrase is a 'poetic' comeback. It's successful in the sense that it's exactly what everyone wishes they could come up with when put on the spot, so it's easy to remember. Like you say, it is unfortunately rather mean spirited, and I say unfortunate because I suspect that it contains truth and usefulness.
My running theory is that the structure of natural language and jargon is encoded such that it carries with it useful information. And this sort of makes sense. The problems we encounter in reality have a certain structure to them. And the natural progression of encoding is to make frequent things short and infrequent things long (thanks information theory). Over time communities shape language and jargon into the structural shape that is most useful to solve the problems they encounter.
Primarily, I've started down this road to try and get my head wrapped around LLMs. Why are they useful at all? And how can something like prompt engineering even exist? The answer being (maybe) language structure sometimes matches problem structure. And prompts from people who "know" how to speak "right" carry additional structure that helps the LLM to provide better results.
However, I think the same can be applied to "what stupid people think smart people sound like". If I'm right then there's a certain sense in which (for extremely well known problems) talking "correctly" helps you to be successful (although all things being equal actually knowing is going to be a better scenario).
And if you abstract this, then there's a benefit to not knowing what the truth is (it frees up your time) but knowing what the sentence structure of "smart" approximately is (it allows you to pick winners to invest in).
So perhaps we can de-meanify the phrase: "She's what non-experts think experts sound like." Or to maybe up the poetry: "Bach and pop use the same notes, but as the children mature so does their taste."