We're Gonna Need More Mathematicians

Amit Sahai warns that rapid advances in AI will soon outpace most humans’ ability to follow and internalize new mathematical and technical ideas, a gap that could cause talented people to abandon research as earlier generations did when they couldn’t keep up. He stresses that AI systems are already producing novel, high-impact proposals - far beyond mere calculation - and that surrendering comprehension to machines would be a serious loss of human agency. Using a hypothetical terawatt fusion-plant design as an example, he argues that acceptance of radical AI-generated designs requires human communities able to scrutinize models, assess failure modes, validate material behavior, and judge uncertainties; reliance solely on AI judgment would place consequential decisions on reasons no human understands.

To address this, he calls for a major expansion of mathematically sophisticated people organized into sustained, well-supported groups whose explicit task is to study and internalize AI-generated breakthroughs, creating a “deployable intellectual reserve.” He urges broadening the mathematician’s role beyond narrow research specialties, leveraging AI as a tool to enable cross-disciplinary communication but not as a substitute for human depth, and recognizing human biological limits that require cooperative scaling. The note closes as a practical manifesto: invest in training and supporting many more humans to preserve meaningful human understanding and oversight; Sahai acknowledges assistance from GPT-6 Astra and several colleagues.