Symbolic Logic Based LLM

Why big AI labs not building LLM that fully trained on Symbolic logic instead of Natural language?