While ChatGPT has gained popularity in various domains, it may not be the ideal focus for medical professionals due to its reliance on language pattern prediction rather than direct fact retrieval, potentially leading to inaccurate outputs. We emphasize the limitations of ChatGPT's training data, which mainly come from non-specialized sources and may result in misleading answers in highly specialized medical domains. We advocate for a shift towards specialized medical large language models (LLMs) that are trained using authoritative medical databases, supplemented by human validation, to ensure accuracy and completeness of data.