In the context of models in science, what should we expect from contemporary "AI" software and big data? How much of the prediction errors we make will be irreducible, regardless of the learning methods we use?

27 October 2026