摘要
具有变革性潜力的多模态模型在医疗领域的快速发展和广泛应用,标志着医疗人工智能进入了一个全新的大模型时代,通过整合和分析大量的文本、图像和语音数据等,在预防、诊断和治疗各个环节推动了医疗服务的个性化和精准化。然而,随着大型多模态模型在医疗健康领域的快速发展和应用,医疗领域的监管机构和传统规则体系也面临着全新挑战。其中,虚假信息、“情感操纵”、算法偏见和侵权责任的模糊都是亟待解决的核心问题。为应对这些挑战,一是落实人工智能全生命周期安全措施,打造可信的医疗大模型应用;二是践行伦理嵌入设计的AI伦理治理理念,实现医疗大模型价值对齐;三是明确医疗大模型的产品责任规则的适用,确保对受害人的有效救济。共同推进医疗人工智能安全和有效应用,为构建更加健康、公正和智能的医疗生态系统作出贡献。
The rapid development and widespread application of transformative multimodal models in the medical field mark the advent of a new era of large models in medical artificial intelligence.By integrating and analyzing vast amounts of text,image,and speech data,these models are advancing the personalization and precision of medical services across all stages of prevention,diagnosis,and treatment.However,the swift progression and application of large multimodal models in the healthcare sector also present new challenges to regulatory bodies and traditional rule systems.Core issues that urgently need to be addressed include misinformation,“emotional manipulation,”algorithmic bias,and ambiguous liability for infringements.To address these challenges,it is essential to:first,implement safety measures throughout the entire lifecycle of artificial intelligence to develop trustworthy applications of large medical models;second,practice AI ethics governance concept by embedding ethical design to align the values of large medical models;and third,clarify the application of product liability rules for large medical models to ensure effective remedies for harmed individuals.Collectively,these efforts will promote the safe and effective application of medical artificial intelligence,contributing to the development of a healthier,fairer,and smarter healthcare ecosystem.
作者
曹建峰
徐艳玲
CAO Jianfeng;XU Yanling(University of International Business and Economics,Research Center for Digital Economy and Legal Innovation,Beijing 100029,China;South China University of Technology,Law School,Guangzhou 510006,China)
出处
《中国医学伦理学》
北大核心
2024年第9期1023-1029,共7页
Chinese Medical Ethics
关键词
通用基础模型
大型多模态模型
价值对齐
可信AI
产品责任
general-purpose foundation models
large multimodal models
value alignment
trustworthy AI
product liability