Can artificial intelligence truly hold consistent values? This question is at the heart of a fascinating new research paper, "Are Large Language Models Consistent over Value-laden Questions?" The study delves into whether AI models like the powerful Llama-3 and GPT-4 can maintain consistent viewpoints when faced with complex, ethically charged topics. The researchers devised clever tests, posing questions about everything from euthanasia to Thanksgiving, and throwing in paraphrases, translations, and different question formats to see if the AI could stay on track. Surprisingly, the larger AI models demonstrated a good deal of consistency, often rivaling human participants. However, things got interesting when the topics became more controversial. While AI held steady on less divisive issues, its consistency wavered on topics like euthanasia, mirroring the disagreements we see among humans. This inconsistency was even more pronounced in fine-tuned models, suggesting that the way we train AI significantly impacts its ability to hold a steady viewpoint. The research also explored whether AI could be steered towards specific values, using Schwartz's universal values as a guide. The results showed limited success, indicating that simply prompting AI with value-laden words doesn't easily sway its responses. This opens up a whole new avenue for research: how can we better guide AI towards desirable values and ensure consistency in its behavior? The study doesn't offer all the answers, but it raises crucial questions about the nature of values in AI and how we can build more reliable and ethically robust systems. As AI becomes increasingly integrated into our lives, understanding its capacity for consistent values will be essential for navigating the complex ethical landscape ahead.