Odd: when I think of "learning AI" I think I'd be constructing an LLM that has added value over what's currently available - such as removing hallucinations within a limited scope.