Perspectives on Large Language Models for Relevance Judgment