Schütze B, Schlieter H
Arbeitsgruppe "Datenschutz und IT-Sicherheit im Gesundheitswesen" (DIG), Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (GMDS), Industriestraße 154, 50996, Köln, Deutschland.
Fachgruppe Digital Health, Gesellschaft für Informatik e. V. (GI), Bonn, Deutschland.
Radiologe. 2019 Dec;59(12):1091-1096. doi: 10.1007/s00117-019-00599-9.
Artificial intelligence (AI) is not intelligent in the human meaning-AI works on the basis of algorithms created by humans and the results of AI are based on real-world data. AI is regulated by all legal frameworks that also apply in other scenarios, in particular data protection law and medical device law. AI delivers results that are as good as the database that formed the learning content of the AI enables. If, for example, the data on which the AI is based contain discrimination, this discrimination is "learned" by the AI; since the AI does not really think, the AI cannot reflect its own results either-it can only make a comparison within the parameters of the specifications by the software developers. If AI is used, the question of the algorithms implemented and the database on which the decision is based must therefore be clarified beforehand and a check must be carried out based on this to determine whether the AI is suitable for use. Due to the missing possibility of real human thinking, AI cannot replace physicians, but can assist them in their daily work.
人工智能(AI)并非人类意义上的智能——人工智能基于人类创建的算法运行,其结果基于现实世界的数据。人工智能受所有适用于其他场景的法律框架监管,特别是数据保护法和医疗器械法。人工智能得出的结果与构成其学习内容的数据库所能达到的效果一样好。例如,如果人工智能所基于的数据包含歧视内容,这种歧视就会被人工智能“学习”到;由于人工智能并不真正思考,它也无法反思自己的结果——它只能在软件开发人员规定的参数范围内进行比较。因此,在使用人工智能时,必须事先明确所实施的算法以及决策所基于的数据库问题,并据此进行检查,以确定人工智能是否适合使用。由于缺乏真正人类思考的可能性,人工智能无法取代医生,但可以在日常工作中协助他们。