A recent paper from Harvard researchers argues that while AI excels at optimizing known problems, it struggles with the “hard problem” of independently discovering and defining new scientific questions, a task still reliant on human creativity and insight.