A reflection on how Saussurean structuralism helps us to see what large language models are doing when they generate fluent text, and where their sign-to-sign mastery falls short of grounded, accountable human meaning.
Sepehr Haghighi (Sat,) studied this question.