Investigating Cultural Alignment of Large Language Models