Recent advancements in large language models, such as ChatGPT, have demonstrated significant potential to impact various aspects of human life. However, ChatGPT still faces challenges in providing reliable and accurate answers to user questions. To better understand the model's particular weaknesses in providing truthful answers, we embark an in-depth exploration of open-domain question answering. Specifically, we undertake a detailed examination of ChatGPT's failures, categorized into: comprehension, factuality, specificity, and inference. We further pinpoint factuality as the most contributing failure and identify two critical abilities associated with factuality: knowledge memorization and knowledge recall. Through experiments focusing on factuality, we propose several potential enhancement strategies. Our findings suggest that augmenting the model with granular external knowledge and cues for knowledge recall can enhance the model's factuality in answering questions.

The most likely explanation for these results is that ChatGPT was trained on content containing political biases. ChatGPT was trained on a large corpus of textual data gathered from the Internet. Such a corpus would probably be dominated by establishment sources of information such as popular news media outlets, academic institutions, and social media. It has been well-documented before that the majority of professionals working in those institutions are politically left-leaning (see here, here, here, here, here, here, here and here). It is conceivable that the political leanings of such professionals influences the textual content generated by those institutions.