引用pella · 2025-01
“(o1-preview) LLMs show promise in clinical reasoning but fall short in probabilistic tasks, underscoring why AI shouldn't replace doctors for diagnosis just yet.”
引它说明 LLM 有临床推理潜力但概率任务不足,AI 暂不应取代医生诊断
多项实验和急诊室实测里,LLM 在诊断与管理推理上超过医生基线。
论文用具有挑战性的临床病例,把一个大语言模型与数百名医生组成的基线作比较,做了五项实验,覆盖鉴别诊断生成、诊断推理展示、分诊鉴别诊断、概率推理和管理推理,由具有验证过心理测量学指标的医生专家评判。作者还在波士顿一家大型三级学术医疗中心急诊室做了真实世界研究,比较人类专家与 AI 的第二意见。在所有实验中该 LLM 都表现出“超人”的诊断与推理能力,作者据此呼吁尽快开展前瞻性试验。
评论者引用它讨论 LLM(如 o1-preview)在临床推理上能否超过医生,既有摘要里的“超人”结论,也有人指出它在概率类任务上仍有不足、暂不应取代医生诊断。
“(o1-preview) LLMs show promise in clinical reasoning but fall short in probabilistic tasks, underscoring why AI shouldn't replace doctors for diagnosis just yet.”
“Link: Superhuman performance of a large language model on the reasoning tasks of a physician. — Adam Rodman, a co-author of the paper [LINK]”
“Our study suggests that LLMs have achieved superhuman performance on general medical diagnostic and management reasoning”