In September, OpenAI announced that its artificial intelligence technology solved the Navier-Stokes problem in less than a week, resolving a mathematical puzzle that had challenged top human experts for decades. This achievement, followed by a wave of additional findings, has stunned the mathematical community and sparked discussions about the future of academic research. However, the implications of these advanced AI capabilities extend far beyond theoretical mathematics into the critical realm of cybersecurity.
Silicon Valley is now developing AI systems designed to eliminate dangerous flaws, known as bugs, in computer software. The underlying logic mirrors the approach used for mathematical proofs: just as an AI can verify a complex theorem, it can theoretically prove that a software application is free of errors. This initiative, termed "software verification" by scientists, has become vital as the internet confronts its most significant cybersecurity threat in decades. With leading AI systems now capable of hacking software by identifying and exploiting vulnerabilities at unprecedented speeds, there is an urgent need for a new generation of AI dedicated to removing these flaws.
Among the companies pioneering this technology is &Truth, a Palo Alto-based start-up. Alongside other Silicon Valley firms like Axiom and Logical Intelligence, as well as major AI labs including Anthropic and OpenAI, &Truth aims to verify global software using Lean, a specialized programming language. This same tool enabled AI to solve million-dollar mathematical problems. Clark Barrett, a Stanford University professor and co-founder of &Truth, noted that the breakthroughs in mathematics have made previously impossible tasks achievable. He emphasized that the formal methods used to prove massive open conjectures can now be applied directly to verifying software integrity.
Sources: The New York Times Science.