137. 对洪乐潼的4小时访谈：AI for Math、把数学变成Lean、数学天书中的证明、直觉、被创造与被发现的

Axion's AI achieved a *perfect score* on the Putnam Math Competition, demonstrating AI's ability to solve complex mathematical problems and even discover proofs in ways distinct from human intuition.

The core vision is to transform mathematics into a formal programming language like Lean, enabling *automated formal verification* and ensuring 100% correctness in proofs, a critical step for applications like chip design and code verification.

While AI can accelerate discovery and proving, human mathematicians will remain crucial for *intuition, conjecturing, and identifying significant problems*, operating at higher levels of abstraction.

Hong Letong, despite not seeing herself as a "genius," built Axion by embracing difficult problems, navigating challenging fundraising (seed round valuation *tripled*), and assembling a diverse team of AI, math, and compiler experts.

AI for Math promises a shift from a "math poor" to "math rich" society, with *exponential growth in theoretical discoveries* and the potential for AI scientists to continuously improve themselves.