Terence Tao: Why I Co-Founded SAIR

In this new episode of On the SAIR, Fields Medalist Terence Tao joins Peter for a conversation on what “AI for science” actually demands: not hype, but methods scientists can trust. Tao shares why he helped co-found SAIR as the AI for Science: Kickoff 2026 approaches: The tools are ready to reshape scientific work, but there are far more ways to use them poorly than well. Getting it right means researchers stay deeply involved, set standards, and build workflows that keep outputs accountable. The discussion also dives into why mathematics may be a best-case testing ground: When AI produces confident claims, math has the culture and tooling to check them, including formal verification systems that force each step into a precise, machine-checkable form. Together, Terence and Peter explore: 🔹 Why Tao co-founded SAIR and why academia has to lead, not follow 🔹 The reliability gap in modern AI, and why “plausible” is not enough 🔹 How proof assistants and verification can keep outputs honest in mathematics 🔹 Why breadth is powerful, but still needs human judgment 🔹 What real progress looks like: interactive workflows, not one-click answers 🔹 Why “AI” isn’t one thing, and why scientists use different tools than the public As Tao puts it: “We didn’t just want the answer. We actually wanted the process as well.” SAIR is launching its public journey through AI for Science: Kickoff 2026 at UCLA — a global gathering of leaders across academia, technology, and research exploring the next frontier of AI-driven science. Learn more: https://sair.foundation/ Follow SAIR: LinkedIn:   / sairfoundation   X: https://x.com/SAIRfoundation