“Grant Sanderson predicts it would be surprising and disappointing if there were no economically valuable improvements directly resulting from AI progress in mathematics over the next 5 years.”