Steven Strogatz warns AI could turn math into an “oracle”

In short: Mathematician Steven Strogatz says AI tools may soon give math answers that humans cannot fully explain or prove.

What's going on

Steven Strogatz, a Cornell mathematician, spoke with Ars Technica about his new book, Big Math: The Hidden Codes That Run Our World, co-written with applied mathematician Alex Townsend. The book looks at “big math,” meaning math problems that are too large for a person to hold in their head, even if the numbers themselves are not enormous.

Strogatz argues that computers have already changed math in stages. First, computers acted like “servants,” doing calculations people already understood. Later, they became “scouts,” exploring problems where researchers did not know what they would find, like early computer simulations of physical systems.

Now, he says AI can start acting like an “oracle.” By that he means a system that gives useful results, but cannot clearly explain how it got them. He compares this to tools like Google Translate and AlphaFold, which can produce strong outputs, but do not necessarily provide human-style explanations that teach new underlying rules.

Strogatz also points to practical uses of large-scale math, like testing whether voting district maps are unusual outliers. In these cases, math can work like a “control experiment” by generating many possible maps that meet legal rules, then checking how extreme a proposed map is.

What to watch

If AI produces results that seem correct but are hard to prove, Strogatz suggests math may rely more on testing and verification, similar to how medicines are checked through trials even when the exact mechanism is not fully understood. That shift could bring benefits, but it could also reduce the traditional role of proofs, which are like step-by-step receipts that show why an answer must be true.

Source: Arstechnica