Explore a 27-minute conference talk from the Uncertainty in Artificial Intelligence event focusing on probabilistic circuits and their ability to handle out-of-distribution data. Delve into the research presented by Fabrizio Ventola and colleagues, which challenges the assumption that probabilistic circuits are inherently well-calibrated and robust to out-of-distribution scenarios. Learn about their proposed solution, tractable dropout inference (TDI), which offers a novel approach to uncertainty quantification in probabilistic circuits. Discover how TDI provides sampling-free uncertainty estimates in a single forward pass, improving the robustness of probabilistic circuits to distribution shift and out-of-distribution data. Gain insights into the experimental results demonstrating the effectiveness of this method on real-world data, evaluating classification confidence and uncertainty estimates.