On 11 September 2026 the best mathematicians in the world published a declaration. It essentially says that AI labs solving math problems don't help the field of mathematics. Over 8000 scientists endorsed it, including Fields Medal winners.
Hi, my name is Tom Smykowski. I've spent 19 years building production systems and now focus on AI-assisted development. On this blog I write about the intersection of AI and software engineering.
I had some trouble reaching that declaration. It's not something media covered very much, and if they did, they softened what the declaration says.
The context of the publication is important here. Some time ago an AI lab was accused of stealing research mathematicians did to publish their results, taking away proper attribution to people who worked on the problem for years.
Now, when one of the labs published supposable solutions to some math problems, Fields Medal owners with over 8000 endorsers point out how harmful AI is for their field, and way beyond.
Why Discoveries Don't Impress Scientists
People may wonder, ok, but what's wrong in AI making discoveries, isn't it all about it?
As it turns out, making a discovery isn't impressing scientists as we thought it would. They point out that these solutions don't contribute to the mathematics field at all.
The full article explores the jungle analogy that explains why AI discoveries fail to advance science, the credit problem facing researchers, verification challenges that may take generations to resolve, and what this means for the future of research.