XAI and Beyond: Towards Trustworthy AI
Friday 11/03/22, 2:00pm
Benoît Frenay (UNamur)
Abstract:
Nowadays, AI systems are spreading everywhere, causing radical changes in some domains. Yet, this (r)evolution does not come without any risk. Safeguards must be put in place and properties like explainability and interpretability have become paramount. In this talk, I begin by discussing explanaible artificial intelligence (XAI) from both theoretical and legal points of views. Then, through several recent works in our human-centered machine learning (HuMaLearn) team at UNamur, we will explore how challenges linked to trustworthy AI can be tackled: explainability, interpretability, interactivity, constraint enforcement, etc. I will also advocate the interest of user studies and illustrate their importance.