Do Large Language Models Defend Inferentialist Semantics?: On the Logical Expressivism and Anti-Representationalism of LLMs