Interpretable Inductive Biases and Physically Structured Learning | A NeurIPS workshop proposal