Do they pass it? I doubt a current LLM can fool anyone to not realize they are an LLM for a long enough discussion session.
The question is only if future LLMs might be good enough to trick anyone in most iterations, whether we would be forced to admit they understand meaning.
This isn't Turing test we are talking about. Of course LLMs pass understanding tests. ChatGPT can easily explain overloading or polymorphism and answer related questions.
The question is only if future LLMs might be good enough to trick anyone in most iterations, whether we would be forced to admit they understand meaning.