Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs