Sahai begins by recalling his undergraduate days, when some talented peers felt they could not keep up with the fastest learners and eventually left research mathematics. He notes that the current community largely consists of those who either never felt that way or overcame it through hard work, but now everyone may experience the same feeling as AI systems generate ideas faster than humans can follow. He observes that AI is already creating beautiful concepts that go beyond mere calculation, and future systems may produce even more astonishing results. If researchers respond to this pace by withdrawing, the community would abandon its responsibility to humanity to understand and shape the discoveries that will change the world. Instead, Sahai proposes that struggle be shared: research groups with sustained support should spend months or years studying extraordinary AI‑generated ideas, aided by AI tools. This collaborative effort, he says, could become some of the most important mathematical work ahead and requires a significant increase in the number of mathematically sophisticated people worldwide. To illustrate why society needs this expertise, Sahai imagines an AI‑designed one‑terawatt nuclear fusion power plant. While the design promises abundant clean energy, its novelty means no existing operational experience can guarantee safety. Humans would need to understand the underlying model, experimental evidence, uncertainties, and possible failure modes before approving construction. He stresses that grasping a theorem requires understanding its model and its limits, a task that demands deep mathematical training. Although one could let AI decide whether to build the plant, Sahai warns that handing over consequential decisions to machines whose reasoning no human community understands would erode human agency. He argues that retaining the ability to meaningfully consider alternatives and decide what kind of world to build is a fundamental value worth the effort. To meet this challenge, Sahai suggests broadening the notion of a mathematician’s role, envisioning a “deployable intellectual reserve” of mathematically skilled communities that can be called upon to comprehend AI‑enabled breakthroughs across domains. He acknowledges that AI might make individuals more effective, but emphasizes that biological limits on depth of understanding necessitate cooperation; through shared effort, humanity can achieve more than any individual could alone. Ultimately, if AI fulfills its promise of delivering ever more beautiful and consequential ideas, society must respond by cultivating thriving human communities capable of understanding them together.
