Understanding Causality with Large Language Models: Feasibility and Opportunities