The Design of Trustworthy AI System: a Deep Look into the Transparency of Data, Models, and Decisions
Résumé
Transparency is a cornerstone of trustworthy AI, a key mandate of the European approach to AI and a crosssectional principle of FUTURE-AI guidelines in health imaging. In this talk, we will critically overview the strategies to guarantee an appropriate level of transparency, mainly in terms of data provenance and usage, AI system development and functioning interpretation. Current approaches, hindrances, and opportunities will be highlighted and discussed, along with the strategies adopted in the EU H2020 ProCAncer-I project.
Origine | Fichiers produits par l'(les) auteur(s) |
---|