Terence Tao says: AI will create a math crisis larger than Godel’s

Fields medalist Terence Tao (considered by many to be the best mathematician alive) wrote an essay to the International Congress of Mathematicians about how AI may cause a crisis similar to the one in the early 1900s when Godel’s theorems required a fundamental reshaping of mathematics. He likens this to the crisis of 1900–1930, where Russell’s paradox and the incompleteness theorems revealed deep assumptions built into all of mathematics.

7 out of 10 problems from the First-Proof project got “positive evaluations” from AI for a few dozen dollars each. Dozens of AI-generated proofs sit in the Erdős database, never checked by humans. AI-generated texts are “too well-polished” to be stress-tested for correctness.

Math relies on humans to understand and check proofs. If AI can create perfectly polished texts that are easy to read but difficult to stress-test, then we lose our self-checking process, and there may be a crisis of confidence in all of mathematical knowledge.