> I want to understand how things work. AI feels like a black box to me. The amount of papers I’d have to read and mathematics that I’d have to ingest to really understand why a certain prompt X results in a certain output Y feels overwhelming.

First, no paper will make you understand why prompt X made result Y. But you can understand the architecture of these systems and try to understand the prompt answer relations somewhat intuitively (e.g., by learning your own models with various text collections).

I think we need to accept these AIs as creatures of their own kind. As complex actors that we don't fully understand.

Humans have been breeding dogs since tens of thousands of years and we don't understand them fully. But we understand enough to employ them in useful ways and mitigate the risks.