Attempts to automate cartography are humbling in the way they reveal the science and art of human cartography. They reveal cartographic design to be a complex decision-making process, in which the solution is one of compromise, working within a multiple set of constraints and opportunities. Although technology may fundamentally change how we handle spatial information, the underlying principles of effective visualization and cartographic design remain the same. Generalization is but one essential piece of the jigsaw, integrated among a set of techniques and technologies that extend from capture and storage through to visualization and interaction. There are many challenging issues in generalization research: from modeling the complex decision-making processes inherent in the art and science of cartography, through to interaction models that support easy interaction and exploration of geographic phenomena across a continuum of scales. Moderately high levels of automation are now achievable but there remains an exciting set of research challenges that are evolving in response to the disruptive nature of IT. This chapter presents a set of challenges as a pointer to a research agenda that includes, among other things, generalization of global datasets, better evaluation methodologies, and better integration of reasoning about space in order to support autonomous solutions in map generalization.