Do people in ML consider large, opaque models interesting from an academic/theoretical standpoint? What do we learn from them?
@mc Why wouldn't it be interesting? Major fields of science exist entirely to study opaque things, like biology and macroeconomics
