Computational learning theory through a new lens: scalability, uncertainty, practicality, and beyond