This symposium will feature presenters whose research, jointly considered, may propel machine learning (ML) beyond statistical pattern matching to dynamical multi-scale interpretations. Key gaps in ML include the exploration-exploitation tradeoff, unsupervised learning, and task transfer/generalization, which are solved by the brain but challenge our computational understanding. How can we translate decades of animal studies into useful spatial, sensory, and motivational signals to explain foraging strategies? The panel will also be encouraged to discuss how recent advances in experimental and data sciences might advance theoretical neuroscience. Our panel will clarify and prioritize research gaps for the future of theory- and data-driven dynamical neuroscience and ML.