It is becoming apparent that generative AI has significant implications for education. However, previous technologies that have had a large impact, such as calculators, do not provide a suitable model for understanding how generative AI can and will be used in learning. Drawing on research on human-computer interactions, we map out a typology of possible student-to-generative AI relationships to afford a more nuanced discussion about what these new technologies can and should be used for in learning. Our contribution in this article is to offer a typology for considering the range of possible interactions across two dimensions of relationships. In doing so, we argue that there is not a single metaphor for the relationship between humans and AI in learning, but many.