近日，纽约科恩儿童医学中心的三位儿科医生对大型语言模型 ChatGPT 的儿科疾病诊断能力进行了评估，结果发现其准确率仅为 17%。研究人员让 ChatGPT 诊断了 100 个随机的儿科病例，结果发现其只有 17 次回答被评为“正确”。尽管 ChatGPT 在医疗诊断领域仍有很大的提升空间，但研究人员表示，通过更具针对性的训练，其准确性可能会得到提高。目前，ChatGPT 或许可以作为一种辅助工具，用于整理文件、协助撰写研究文章或为患者提供术后护理指南等。
Title: Accuracy of ChatGPT in pediatric disease diagnosis仅为 17%
Keywords: ChatGPT, pediatric disease, diagnostic accuracy
News content:
Recently, three pediatricians from the New York Coen Child Medicine Center evaluated the diagnostic ability of the large-scale language model ChatGPT in pediatric diseases, and found that its accuracy was only 17%. Researchers diagnosed 100 random pediatric cases with ChatGPT, and only 17 of its answers were rated as “correct”. Although ChatGPT still has much room for improvement in the field of medical diagnosis, researchers suggest that its accuracy may be improved through more targeted training. Currently, ChatGPT may serve as an auxiliary tool for organizing files, assisting in writing research articles, or providing post-operative care guidelines for patients.
【来源】https://www.ithome.com/0/743/448.htm
Views: 1