History and Philosophy of Science Seminar (cross-listed with the LCSSP series)

Abstract: This project attempts to undermine the well-known distinction between generative and non-generative artificial intelligence (we call this distinction ‘The Generative AI Myth') by arguing that all statistical learning is generative in nature. To do so, we define a generalized statistical learning problem, and we then show that every such problem can be framed as a generative learning problem. This mathematical equivalence forms the backbone of our argument. We then examine the pernicious consequences of the Generative AI Myth, and we consider whether a variant of it can be salvaged (we conclude that it cannot be).

Co-authored work with Alec Kirkley, Institute of Data Science, University of Hong Kong.