What is the most important unsolved problem in Artificial Intelligence? It may well be the problem of creating a unified model of the universe, to which all observations of the system can be related to. This requires representations that are universal and dynamic, combine tacit and propositional knowledge, and can be grounded in predictive models of real-time sensory data. The demands of online learning, dynamicity, and self-organization may require the re-engineering of large parts of the current stack of machine learning tools and a change in perspective on the design of suitable AI architectures. This presentation discusses some prerequisites for systems that model their own agency and their relationship to the environment and contrasts existing approaches with the way in which humans model their own self and autonomy.