Explore the critical role of uncertainty quantification in artificial intelligence-driven physical simulations through this 24-minute conference talk by Dimitrios Tzivralis from CEA LPTMS. Delve into the challenges and methodologies for understanding and managing uncertainty when AI models are applied to simulate complex physical systems. Learn about the theoretical foundations and practical implications of incorporating uncertainty measures into AI-based simulation frameworks, examining how these considerations impact the reliability and interpretability of computational physics results. Discover current research approaches for addressing uncertainty propagation, model validation, and error estimation in the context of machine learning applications to physical modeling problems.