Mathematician Bartosz Naskręcki and co-authors release the Leiden Declaration calling for global guidelines on AI in mathematical research · Digg

/Tech27d ago

Mathematician Bartosz Naskręcki and co-authors release the Leiden Declaration calling for global guidelines on AI in mathematical research

The manifesto aims to preserve rigor in scientific workflows.

--0--

#226

Original post

Patrick Shafto#1621

Bartosz Naskręcki@nasqret

AI technology is on the rise. So are its applications in industry, science, and society at large. At

https://leidendeclaration.ai

we, a group of 16 scientists, including mathematicians, computer scientists, philosophers, and historians, have drafted a declaration calling for action on the challenges posed by the use of artificial intelligence in mathematical research.

Over several months, we carefully weighed the promises and risks of introducing AI into research workflows and beyond. The declaration has already been endorsed by international communities, most prominently by the International Mathematical Union, as well as by a large group of mathematicians and other scientists.

We invite our colleagues to sign the declaration and support its recommendations on how AI should be adopted in our workflows, and how we should establish and maintain relationships with AI labs and industry. One of its central points is the need to preserve balance, rigor, and sound judgment while acknowledging the extraordinary pace of development in the AI community.

We hope this declaration, a deeply human act, will serve as a guiding document for those who recognize the human imperative and the need for governance in scientific discovery, especially mathematical discovery.

5:59 AM · Jun 2, 2026 · 6.3K Views

Sentiment

Positive users endorse the Leiden Declaration on AI in mathematics for promoting human-led ethical approaches, while negative users criticize its writing quality, group composition, and terminology like 'stochastic parrots'.

Pos

50.0%

Neg

50.0%

9 comments with sentiment.

Cluster Engagement

Views

Comments

Reposts

Bookmarks

Expand data

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

LEIDENDECLARATION.AIVia

Posts from X

Most Activity

VIEWS17.4KBOOKMARKS39LIKES102RETWEETS21REPLIES6

Ken Ono@KenOno691

I unequivocally support the Leiden Declaration. As a former AMS VP, current Editor-in-Chief, and someone bridging academia and industry, I see its message as urgent. AI will shape mathematics. We must uphold rigor, transparency, attribution, and culture. https://leidendeclaration.ai/

27d17.4K10239

Tal Linzen@tallinzen

On the margins of this but I don't think anyone intended the term "stochastic parrots" to refer to any AI system that would ever be created in the future. It was certainly more apt for 2020 when the paper was written than for current systems that mix next token generation with reasoning, tool calling, retrieval, etc.

Will Kinney@WKCosmo

Mathematicians are no longer calling AI "stochastic parrots" or "next-word prediction machines."

https://leidendeclaration.ai/

27d5.2K296

Thomas Bloom@thomasfbloom

Indeed, the word mathematics itself comes from Greek, meaning roughly 'that which is to be learned'.

With no humans to learn it, it is not true mathematics.

26d329141

Thomas Bloom@thomasfbloom

If a machine is writing down theorems and proofs that will never be read by a human, I don't care about that.

This otherworld exists and needs no proofs; proofs are a way to make sense of it to the human mind, and without that human mind there is no point to the proof.

26d320171

Aran Nayebi@aran_nayebi

About time!

Will Kinney@WKCosmo

Mathematicians are no longer calling AI "stochastic parrots" or "next-word prediction machines."

https://leidendeclaration.ai/

27d1.6K52

Thomas Bloom@thomasfbloom

For example maybe the machines will invent new branches of mathematics to explore some branch of physics that will lead to unlimited clean free energy. Maybe this mathematics is too hard for our human brains, but it doesn't stop the physical consequences benefiting humanity.

26d1597

Aran Nayebi@aran_nayebi

And here's why the moniker was never accurate (aside from the obvious one of LLMs literally learning a compressed representation to generalize to new data, which they clearly are doing!):

27d1071

Thomas Bloom@thomasfbloom

I am not going to turn that down, even if I did prefer that such a discovery come from humans.

But it would still be a shame if we lost our ability to understand and explore such maths ourselves.

26d1647

Thomas Bloom@thomasfbloom

It is now clear that AI can produce interesting mathematics, and greatly accelerate research.

But we need to make sure that humans are not left behind, and that this after this massive change there are plenty of humans able to think about and understand mathematics.

26d1676

Thomas Bloom@thomasfbloom

Now it can be argued that maybe we don't care about the human involvement in 'mathematics' in the long run, as long as it benefits humans down the line somewhere.

26d1666

Bartosz Naskręcki@nasqret

UK Academy for Mathematical Sciences, an umbrella body for the UKs math-related learned societies https://www.acadmathsci.org.uk/announcement-of-leiden-declaration-on-artificial-intelligence-and-mathematics/

Isaac Newton Institute https://www.newton.ac.uk/news/ini-news/mathematicians-call-for-action-on-ai-to-protect-the-future-of-the-discipline/

New York Times https://www.nytimes.com/2026/06/02/science/ai-mathematics-leiden-declaration.html

Scientific American https://www.scientificamerican.com/article/mathematicians-sign-declaration-to-rein-in-ai-use/

27d4906

Thomas Bloom@thomasfbloom

@prz_chojecki I agree the answer is not 'keep the status quo' but it is also not 'throw away all existing institutions and structures and invent something ad hoc'.

Like most things, the best path is somewhere in the middle. And when so much of the coverage is a breathless focus on the new,

26d862

Thomas Bloom@thomasfbloom

I am not saying this is inevitable, or even likely or possible. Just that we should try our best, as early as possible, to make sure that AI and humans are aligned on what 'good mathematics' means, both abstractly and in its impact on society.

26d1815

Thomas Bloom@thomasfbloom

...which are, for some reason, just incompatible with the way our brains work. We don't have the capability to reason about such objects. But the AI prefers them, so builds theories about them, which are correct, and proves RH that way.

26d1385

Thomas Bloom@thomasfbloom

But maybe there were other proofs of RH possible. Proofs that better fit our own human intuition and perspective. But maybe the AI prefers another way, and once it is done, the desire to find a human proof for ourselves will fade.

26d1984

Thomas Bloom@thomasfbloom

Concretely, let me respond to your "If AI can make math progress faster, then it is only for good of mathematics."

This is not necessarily true. Perhaps AI will decide that the most efficient route for -it- to, say, prove the Riemann Hypothesis, is to invent new theories...

26d1444

Bartosz Naskręcki@nasqret

International Mathematica Union:

https://www.linkedin.com/posts/international-mathematical-union_the-imu-endorses-the-leiden-declaration-on-activity-7467545920860262400-LBhV

IMU Circular: https://www.mathunion.org/fileadmin/documents/2026-06/IMU_AO_CL_8_2026.pdf

IMU Editorial: https://www.mathunion.org/imu-news/archive/2026/imu-news-137-may-2026

27d6055

Przemek Chojecki | PC@prz_chojecki

Indeed, I understand mathematics in a platonic sense. Discovering mathematics is not reserved for humans only.

However I agree with the rest of your argument fully. Mathematics, AI, science in general, as we do it should be fully human-centric.

You conveyed better what I wanted to say - if there's free unlimited energy to be discovered and mined (Dyson spheres like) - I don't really care about whether it's human-discovered or AI-discovered. Just let's make sure it has positive impact on our civilization as a whole.

But to make that leap into unknown, we need to rethink incentives in science, academia. And that's the part I don't agree with when it comes to Leiden declaration - it looks more like trying to save the status quo.

26d1192

Rajarshi Maiti@formaliser_dude

the discoveries that AI models are currently making are based on the millions of lines of training data that mathematicians have written.

The same reason why there is still a ceiling on how current models can get, and why we will always need human mathematicians, is essentially the same concept.

My belief is that you cannot synthetically automate iteration loops and expect breakthroughs. It is only through humans at every part of the loop.

26d10

Juan Valle Lisboa@LentoBio

@tallinzen The name stochastic parrot shows a lack of understanding of underlying principles. The intelligent part is creating conplex internal representations that can support all sorts of responses that preserve knowledge. In this sense, even gpt2 was not a stochastic parrot.

27d1391