The main goal of the project is to extend and improve on our VBMC framework for efficient probabilistic inference with moderately-to-very expensive models, published in multiple papers, available in MATLAB and recently released for Python. We aim to perform Bayesian inference for parameters of complex, expensive state-of-the-art models in fields such as cognitive science and AI. An example is the AI-inspired model of human gameplay from Wei Ji Ma's group (van Opheusden et al., Nature 2023). The project includes funding for research visits to international collaborators such as Wei Ji Ma at New York University and Michael Osborne at the University of Oxford. We also have many local collaborators, such as Antti Honkela for applications of sample-efficient inference to privacy, and our team is highly involved in the thriving & highly collaborative community of probabilistic ML/AI researchers — PhDs, postdocs, PIs — in the Finnish Center for Artificial Intelligence FCAI, on top of many ongoing national and international collaborations in cognitive science and computational neuroscience.