Because most useful things in our real world aren't expressed in symbolic logic. Also Google with alpha proof has shown that LLMs trained on Natural language can produce pretty good symbolic logic as output.