The cognitive sciences have always educated educators by providing a pedagogical framework as a guide. However, the standard cognitive sciences are being challenged by a new paradigm, embodied cognition, in which learning is part of a dynamical system. In this paradigm, virtual embodiment (VE) is the new artificial intelligence (AI). This thesis is an application of VE, introducing an approach to developing a virtual 3D agent that has the potential to achieve “strong AI” status. I believe such agents can mature into AI educators. And that the development of a great AI educator starts with the development of a humble AI child. My methodological approach is a metasynthesis of a broad range of disciplines and consists of (1) the use of empirical research to ground my ideas, (2) the integration of dissimilar research to construct new ideas, and (3) the use of thought experiments to uncover the fundamental nature of learning within an embodiment paradigm. As a result, this thesis introduces a virtual 3D agent, the virtual autonomous learner (VAL), along with key elements of its ecological construct. With an embodied cognitive perspective, VAL seeks to find its own affordance and that of its environment. I conclude that (1) the construct for VAL needs to accommodate different cognitive architectures if we are to make full use of its methodology; (2) a rigorous virtual curriculum must be developed, and efficient pedagogical tools should be designed and developed to implement this curriculum; and (3) an educational perspective is paramount for this project.