在《美国医学会杂志》发表的一篇争议性文章中，肿瘤学家埃泽基尔·伊曼纽尔与创投家维诺德·科斯拉提出大胆预测：到2030年，全自主AI在病史询问、诊断、检验开立、处方及慢性病管理等核心医疗任务上的表现，将全面超越人类医师以及「人机协同」模式，甚至主张人类的介入反而会降低AI的医疗水准。

此论点引发医疗界的强烈反弹与讨论。美国医学会执行长约翰·怀特指出该研究存在模拟实验偏差与真实病患沟通限制，坚称医疗决策必须由医师主导；而加州大学旧金山分校的罗伯特·沃克特则提出「门房谬误」，认为即便诊断被AI取代，人类医师在同理心沟通、传达预后坏消息等无可替代的情感价值与照护功能依然不可或缺。

然而，随著AI快速普及，医学界也面临「去技能化」的深层隐忧，新一代实习医师可能因过度依赖AI而丧失基础诊断能力。这场关于AI全面接管医疗决策的争论，已超出医疗领域本身，演变为人类在高度自动化时代中究竟还能保留何种专业价值与角色的全社会核心诘问。

In a provocative article published in JAMA, oncologist Ezekiel Emanuel and venture capitalist Vinod Khosla argue that by 2030, autonomous AI will surpass both human doctors and doctor-AI hybrids across fundamental medical tasks—such as taking histories, diagnosing, and prescribing treatments—asserting that human involvement may actually degrade AI performance.

This thesis has sparked significant pushback across the medical community. AMA CEO John Whyte criticized the paper's reliance on simulations and argued that care must remain physician-governed, while UCSF's Robert Wachter cited the 'doorman fallacy,' emphasizing that vital human qualities like empathy and delivering difficult prognoses ensure clinicians remain essential even if diagnostic tasks are automated.

Nevertheless, the rapid rise of AI raises urgent concerns over medical 'de-skilling,' as trainees risk losing core diagnostic capabilities through overreliance on algorithms. Ultimately, the question of what remains for doctors to do mirrors a broader societal debate about defining human purpose and professional value in an increasingly automated world.