Imagine spending decades studying mathematics, only for a 23-year-old with no advanced training to crack the same problem after typing one prompt into ChatGPT.
That is what has stunned the maths world this week.
Liam Price, an amateur maths enthusiast, reportedly used ChatGPT Pro to solve a 60-year-old conjecture linked to legendary mathematician Paul Erdos. The puzzle had resisted attempts from some of the brightest minds in the field.
Price says he was casually feeding Erdos problems into the AI, something he often does for fun, when one response looked surprisingly solid.
He checked it. Then others checked it.
And suddenly, experts realised this was not just another flashy AI claim.
WHY THIS CASE FEELS DIFFERENT
Artificial intelligence has already made headlines for solving older maths problems. But many previous wins came with caveats. Some were on easier problems. Others used methods humans already knew. A few looked more impressive than they actually were.
This case appears different.
According to mathematicians who reviewed it, ChatGPT did not simply grind through calculations faster than humans. It approached the problem using a method researchers had not thought of applying here.
That matters because in mathematics, finding a new path can be more valuable than finding one answer.
Terence Tao, one of the world’s most respected mathematicians, said many researchers had started the problem in the same traditional way for years. The AI appears to have avoided that trap entirely.
WHAT WAS THE PROBLEM ABOUT?
The puzzle involved “primitive sets,” special collections of whole numbers where no number can divide another exactly.
Think of it as a group where no member can be built neatly from another member.
Paul Erdos studied how these sets behave and proposed several deep questions about them. One of those remained unresolved for around six decades.
Price’s ChatGPT-assisted solution reportedly tackled that open conjecture.
For everyday readers, the exact formula matters less than this fact: experts had tried for years and got stuck.
AI’S RAW ANSWER WASN’T PERFECT
Before we imagine robots replacing mathematicians overnight, there is an important catch.
Researchers said ChatGPT’s original proof was messy and hard to follow. It still needed human experts to inspect, verify, simplify and rewrite the argument into a cleaner form.
So this was not AI working alone in a vacuum.
It was closer to AI generating a surprising idea, then humans doing the rigorous polishing mathematics demands.
That may be the real model of the future.
THE BIGGER STORY: CAN AI SPARK HUMAN GENIUS?
This moment is bigger than one solved puzzle.
For years, people have asked whether AI can truly create new knowledge or only remix what humans already know. Cases like this suggest the answer may be somewhere in between.
AI may become a tool that helps humans escape mental habits, see hidden links and test bold directions faster than ever before.
In other words, the machine may not replace the mathematician.
It may become the brainstorming partner who never gets tired.
WHAT THIS MEANS FOR STUDENTS AND YOUNG PEOPLE
There is also something quietly radical here: a young outsider with curiosity and a subscription accessed territory once limited to elite academia.
That does not mean expertise is dead. It means access has changed.
A motivated student, coder or hobbyist now has tools that can help them explore advanced ideas from home.
For young people watching this story, the message is simple.
You may not need to be inside the old gates to knock on the door of discovery anymore.
