Understanding generative AI to harness its potentials and mini- mize risks: A perspective

Abstract

The hype surrounding Generative AI is such that the impression one may get is that these technologies are solving all the problems of humankind, including medical diagnoses. This can result in great disappointments (or worse) unless there is a clear and shared understanding of how this technology is developed and works to be correctly deployed as assistive tools under human supervision. Ten years ago, developments in Computer Vision were accompanied by bold claims on machines substituting radiologists. This has not happened (luckily). Generative AI requires additional precautions since it touches upon one of the most defining features of humankind: (natural) language. This contribution offers a perspective on how Generative AI can support advancements in medicine by presenting its inner working mechanisms and a critical reflection on datasets and evaluation methods.

Ähnliche Arbeiten