Erdos Problem 1196: Can AI now solve maths that no human can?

It is said that AI could soon be coming for the jobs of artists, lawyers, and software engineers. But it might now also be threatening a role at the height of academia – are pure mathematicians safe?

Last month, a Stanford mathematician woke up to an email, claiming to have the solution to a problem he had been working on for seven years - a 60-year-old conundrum known as Erdos Problem 1196.

The answer had been generated in just 80 minutes - by ChatGPT.

Since the end of last year, AI has been providing solutions to a number of novel maths problems, but Problem 1196 is the first to raise eyebrows within the mathematical community.

In this episode, we talk to the mathematicians who have worked on Problem 1196 and find out what the rise of AI could mean for the future of their field.

Our contributors include Katie Steckles, Mathematician and communicator, Jared Duker Lichtman, Szegő assistant professor of mathematics at Stanford University and Liam Price, amateur mathematician.

Presenter: Charlotte McDonald
Producer: Josh McMinn
Sound engineer: Dave O'Neill
Editor: Richard Vadon