Introducing the world’s largest Math Olympiad database
Marketplace Tech
Marketplace
4.5 • 1.3K Ratings
🗓️ 28 April 2026
⏱️ 9 minutes
🧾️ Download transcript
Summary
The International Math Olympiad is a yearly competition for students, most of them high school age, who compete to solve six difficult math problems. They're chosen from a pool of math problems submitted by different countries that participate in the competition.
The problems that don't make the cut previously have mostly just been lost; there was no one place you could go to find them.
But now a team at MIT’s Computer Science and Artificial Intelligence Lab has gathered over 30,000 of those problems together in one dataset so both humans and AI models can look through and study them.
Marketplace’s Stephanie Hughes spoke with Mark Hamilton, a visiting researcher at MIT CSAIL who has been part of the work to gather problems. He’s also a Research Scientist at Google's DeepMind laboratory.
Transcript
Click on a timestamp to play from that location
| 0:00.0 | How do you help an AI get better at math? Well, first, you have to give it the right things to study. |
| 0:06.5 | From American Public Media, this is Marketplace Tech. I'm Stephanie Hughes. |
| 0:19.4 | The International Math Olympiad is a yearly competition for students, most of them high school age, who compete to solve six difficult math problems. |
| 0:27.8 | Those problems are chosen from a pool of math problems submitted by different countries that participate in the competition. |
| 0:33.6 | The problems that don't make the cut, Previously, they've just kind of been lost. |
| 0:37.7 | There was no one place you could go to to to find them. |
| 0:40.3 | But now a team at MIT has gathered over 30,000 of those problems together in one dataset, |
| 0:45.9 | so both humans and AI models can look through them and study them. |
| 0:51.1 | Mark Hamilton, a visiting researcher at MIT, has been part of the work to gather problems. |
| 0:55.9 | He's also a research scientist at Google's Deep Mind Laboratory. I asked him why it's valuable |
| 1:01.2 | to have all these problems in one place. |
| 1:04.0 | One of the things that's interesting is that it's just a central resource that you can go to |
| 1:07.8 | if you want a diverse collection of mathematical problems. It's a central resource that students can go to if you want a diverse collection of mathematical problems. It's a central |
| 1:11.6 | resource that students can go to if they like to learn about these problems, learn about the |
| 1:16.6 | different kinds of math in this dataset. And I think also in the AI side of things, if we want to |
| 1:24.2 | create systems that are capable of solving humanity's toughest mathematical challenges and really pushing the state of math forward, we need to give it a large collection of interesting, diverse math that it can study and it can learn from. |
| 1:38.1 | Talk a little bit more about that. |
| 1:39.7 | How would it be useful for AI models to learn about different problems, different approaches to |
| 1:46.1 | solving math problems? Yeah. I mean, there's a ton of independent ways that you can kind of make |
| 1:51.7 | a system actually leverage this. Like one of the ways that will work for both humans and machines |
| 1:57.1 | is just by searching. Like, let's say you're a mathematician and you encounter some kind of |
| 2:02.6 | challenge. Maybe you can describe that challenge and have our search engine going to pull up relevant |
... |
Please login to see the full transcript.
Disclaimer: The podcast and artwork embedded on this page are from Marketplace, and are the property of its owner and not affiliated with or endorsed by Tapesearch.
Generated transcripts are the property of Marketplace and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.
Copyright © Tapesearch 2026.

