When 10,000 Minds Work for 88 Hours: What AI Just Did to Mathematics
I am writing this from my small desk here in Walnut Creek, at 91 nearing 92 years young, with the morning light coming through the window. For more than six decades I have watched mathematics move at the pace of a human life, one blackboard at a time, one letter, one proof, one quiet doubt. Last week, something happened that makes me sit back and simply breathe in wonder.
In October of this year, an announcement came that I never imagined I would read in my lifetime. I want to share it with you not as a technical expert, but as an old man who still believes that wonder is the most important thing we can pass on.
1. What happened
According to OpenAI, over a single long weekend -88 hours, they set loose not one AI, but 10,000 AI agents working together. Not chatting, not writing poems, but doing mathematics. Real mathematics. The kind that has kept the best minds awake for a century.
Their target was one of the seven Millennium Problems set by the Clay Mathematics Institute in the year 2000 - the Navier- Stokes existence and smoothness problem. It is a question about water, air, and flow itself: can we prove that the equations that describe fluid motion will always remain smooth and well-behaved, never breaking into nonsense?
The system did not just produce one draft. It produced 722 complete manuscripts, touching 372 distinct families of problems, lemmas, counter-examples, extensions, and bridges I am still trying to understand. Within that forest of work sits what they claim is a full proof of Navier-Stokes. The Clay Institute has not yet finished its long review as it should not, these things must take years but even the claim, with the weight of checking behind it, has shaken the world of mathematics.
Think of it: 10,000 minds that do not sleep, do not eat, do not forget what the 9,999 others have just discovered, working for 88 hours. That is roughly 880,000 hours of concentrated thought compressed into less than four days.
2. The human story
When I first read this, I thought of my professors in the 1950s, smoking pipes, erasing a blackboard for an hour to get one line right. I thought of my classmates who spent an entire summer on a single inequality.
Some will say this is the end of the human mathematician. I do not believe it for a moment.
What I see is not a replacement but a mirror. For the first time, we can see what mathematics looks like when memory is perfect and stamina is infinite. It shows us how much of our work was limited not by our intelligence, but by our time by the need to sleep, to raise children, to grow old. The young mathematician in Nairobi or in Mindanao who has an idea at 2 a.m. can now ask 10,000 tireless colleagues to test it with her.
The proof, if it holds, still had to be asked for by a human. The question was human. The courage to declare one of the Millennium Problems solved that, too, is human. Machines can produce manuscripts; only we can feel the trembling of the hands when we read them.
3. What remains for us
At 91 nearing 92, I have learned that the important questions are rarely "Can we?" but "What now shall we do?"
First, humility. We must wait. A proof is not true because it is large, or because it is produced quickly. It is true because it survives the patient, sometimes cranky, careful reading of other human beings. That review will take years, and that is good.
Second, gratitude. If you have ever stared at a flowing creek and wondered why it curls the way it does, mathematics is trying to honor your wonder. Whether it is proved by a man with chalk or by 10,000 agents humming in a data center, the beauty belongs to all of us.
Third, invitation. There is more room now, not less, for the human voice. We need explainers, doubters, poets of logic, teachers who can take these 722 manuscripts and tell our grandchildren what they mean. At my age, that is the work I love most sitting with you, dear reader from wherever you are reading, and saying: Look what we humans have dreamed up to do.
The sea is still out there beyond the Chateau. It still flows by rules we are just beginning to name.
- OpenAI — Navier-Stokes announcement — Official announcement of 88-hour, 10,000-agent run
- Clay Mathematics Institute — Millennium Problems — The seven problems including Navier-Stokes
- Northeastern University — AI solving math problems — Reporting on 722 manuscripts and 372 problem families
- Smithsonian Magazine — How AI is changing mathematics — Context on human-machine collaboration
No comments:
Post a Comment