Curvature is the common driving force for grain boundary motion in all polycrystals. However, models and simulations derived from curvature-based motion cannot predict irregular, albeit commonly observed, grain growth behavior. To build better predictive models, we need to employ new tools to understand what governs grain growth. First, I will demonstrate how high energy x-ray diffraction microscopy (HEDM) can be used to observe grain growth in real 3D polycrystalline systems. In a grain growth study employing HEDM of strontium titanate, we find that curvature is a poor predictor of grain boundary migration. Instead, anisotropic grain boundary properties are hypothesized to override the contribution of curvature. Second, I will describe how reinforcement learning, a machine learning tool, can capture the underlying behavior of an evolving Markov decision process and “teach” it to maximize the “rewards” regarding the agreement between prediction and simulation. To demonstrate the feasibility of this approach, we built a deep reinforcement model that emulates grain growth by training on Monte Carlo Potts grain growth simulations. The developed reinforcement model was validated on different microstructural architectures to ensure that it captures the underlying physics. The accuracy of our short and long-term predictions will be evaluated. Then, I will discuss how HEDM and our machine learning model can be combined to understand how anisotropic grain boundaries migrate in 3D polycrystals.