Terence Tao tells Brian Keating that AI systems can suggest approaches to mathematical problems but still invent references and produce unreliable reasoning. Useful contributions need to pass through expert judgment and rigorous verification rather than being accepted because an answer looks polished.
Terence Tao contrasts the efficiency of mathematical language with the nuance of ordinary language, then considers how AI changes teaching. Since tools can already solve many undergraduate exercises, students need practice evaluating answers and identifying mistakes, not simply reproducing solutions.
Terence Tao sees opportunities in pairing imperfect AI systems with strong verification and applying them to large collections of moderately difficult problems. He also favors more collaborative research structures that allow contributors to supply smaller, independently checked parts of a mathematical project.
Watch on YouTube
