Iqbal Latif
You report that 25 winners of the Fields Medal have signed a declaration against the way AI companies are using mathematics, and you open with Socrates, who feared that writing would make people lazy about remembering.
Dear Economist, you report that 25 winners of the Fields Medal have signed a declaration against the way AI companies are using mathematics, and you open with Socrates, who feared that writing would make people lazy about remembering.
It is a good comparison, but it cuts the other way. We know what Socrates feared only because Plato wrote it down. Writing did not kill inquiry. Nor did the printing press, the slide rule, the calculator or the chess engine. More people play chess today than in 1997, when the machine beat Kasparov.
So let us keep this in proportion. On 8 September OpenAI said about 10,000 agents working for 88 hours had settled the Navier Stokes problem. By the company’s own count, the machines wrote 130 billion tokens to get there, which is roughly 100 billion words. The proof is still unverified. Suppose it holds. Engineers have designed aircraft, ships and pipelines with those equations for more than a century, and they work. A proof that an ideal fluid can run to infinity on paper changes no wing, no turbine and no weather forecast. The Poincare conjecture was settled by a man, Grigori Perelman, more than 20 years ago, and no bridge was built differently afterwards. Of the 7 Millennium problems only P versus NP reaches into daily life, because the codes that guard our money and our messages rest on it. The others are beautiful, they are hard, and they are exercise for the mind.
Then why bother? Because curiosity deserves to be settled. A question that has stood open for a century is an itch in the human mind, and there is nothing wrong with scratching it, by hand or by machine. Settling it also teaches us something plain and useful. It lets us say out loud which of these famous questions would change our lives and which would not. That sorting is knowledge too, and nobody offered it to the public before.
The medallists are right on one point. Understanding is the one thing a machine proof does not hand over. A proof nobody understands is a locked box. They are right to ask for credit, for care and for time to check. But that is a complaint about haste and manners in a few companies. It is not the death of mathematics, and their own declaration says AI could help mathematics move faster.
And let us be honest. Some professions will be hit hard, and the reading of scans may be among the first. Mathematics will not be among the dead. Mathematics is the language of the universe, and 0 and 1 are the ground the machine itself stands on. The best programming may soon be done by machines, but the algorithms were designed by human beings. The machine is a human product, and our trust in it is a human decision. A machine can write 100 billion words. It cannot want to know. Wanting to know is still ours.
So do not fear the machine. Fear what is left unasked and unresolved. Stop 100 people in the street and ask them to name the 7 problems of the Clay Institute. Not one will. Ask 10,000 and you may find one. The machines spent 130 billion tokens on Navier Stokes, and nearly every report told me a prize had been won. Hardly any told me what the question was. Most of us do not know, and worse, most of us do not want to know.
For a century these problems lived behind a wall of notation, and every guild likes its mysteries kept inside the guild. That wall is down. An ordinary person like me can now walk in, ask, argue and enjoy it. I did exactly that. Yesterday I spent 3 hours arguing with Claude Fable 5.1 about these very problems. I did not ask it to solve anything. I asked what the problems are, why they matter and where the limits of our knowing lie. What came out is a guide of 101 questions in plain words, for anyone who left school with a little maths and a lot of curiosity. Please read it:
The machine gives every person the chance to ask the right question, to reason on the answer and to argue with it. If they do not use it, that is the problem. If they use it for TikTok and fake memes, that is the problem. My biggest worry is the machine left unused. Used properly, the sky is the limit of your understanding. There will be haves of AI and have nots of AI. The tool is nearly free, so the line between them will be curiosity. I intend to be among the haves, and I want every reader of yours there with me.
The machine did not do the inquiry. The questions did, and the questions were mine.
Intellectual inquiry has never died. It dies only in the person who stops asking.
