What's New in Trustworthy AI in Julia (Taija) - JuliaCon 2024

Learn about the latest developments in Julia's Trustworthy AI ecosystem (Taija) in this 30-minute conference talk from JuliaCon 2024. Explore recent advancements in making AI models more reliable through various approaches including model explainability, algorithmic recourse, predictive uncertainty quantification, Bayesian deep learning, and hybrid learning. Discover updates to core packages like CounterfactualExplanations.jl, which now features improved computational speed, multi-threading support, and enhanced interoperability with Python and R. Examine the evolution of LaplaceRedux.jl with its new support for multi-class problems and sophisticated Hessian approximations. Gain insights into ongoing Julia Season of Code projects focusing on Conformal Bayes and Causal Recourse, as well as the development of TaijaInteractive.jl for web-based interactive model explanations. The presentation also covers student contributions to the ecosystem and their experiences working with Taija packages at TU Delft.