It seems clear to me that AI technologies — specifically LLMs such as ChatGPT and Claude and LLM-based agents — are having and will continue to have disastrous effects on learning, research, and teaching within the mathematical sciences. This is not an accident: the ideology underlying these technologies is fundamentally at odds with the primary goal of furthering our own understanding of mathematics. As such, I do not use AI technologies in any capacity.

Writings on the Purpose and Practice of Mathematics

Mathematicians have written about this issue in thoughtful and elegant ways. I recommend the following texts on the purpose of mathematics, its essential nature, and the way in which AI is harming our discipline:

• Jacques Hadamard (1945) “The Mathematician’s Mind: The Psychology of Invention in the Mathematical Field” [link]
• William Thurston (1994) “On Proof and Progress in Mathematics” [link]
• Max Weinreich (2026) “The Crisis of AI Generated Mathematics” [link]
• Math and AI (2026) “A Severe Misalignment of AI in Mathematics” [link]

Getting Involved 

It can feel as if we as students, teachers, researchers, and practitioners of mathematical sciences have no control over the current state of affairs, that we are simply along for the ride. Nothing could be further from the truth. Our words and actions can make a difference in how our discipline is practiced. 

If the opinions expressed here resonate with you in any way, consider joining the Association for Human Mathematics — don’t get mad, get organized!

Frequently Asked Questions

Q: What do you perceive to be the problem with using AI in a research or teaching capacity?

The primary purpose of AI is to replace the human faculty of thought. This is neither a secret nor is it an exaggeration — it’s how these technologies are explicitly being advertised. We are being asked to use AI when we feel confused, anxious, or unsure of how to proceed. As scholars, we ought to reject this proposition and instead exercise our own faculties of thought.

One of the most important aspects of mathematics is developing intuition. This is not merely being able to “guess what should be true” but instead involves constructing your own personal understanding of concepts, their relationships, examples, counter-examples, etc. These are all collected into your personal constellation of working mathematical knowledge. Ideally, this mental constellation is constantly being refined through hard, active work: grinding through details of arguments, seeking out concrete examples, and comparing these with your current understanding. Most importantly, it is from and through this mental constellation that we engage with the discoveries of others and develop new lines of inquiry. In short, this is how I believe that mathematical progress is actually made.

The primary problem with AI is that it seeks to replace your personal mental constellation. I claim that this replacement has far-reaching consequences at both an individual and community level: understanding becomes more shallow, the natural process of mathematical inquiry is stunted, and social trust within the community breaks down. If we believe that our students are stunted by using AI, why should researchers and teachers be excepted from this effect?

This is only a very short, incomplete description. If I ever write an essay or blog which explains my perspective in further detail, I will link to it here.

Q: What does your AI-free commitment actually mean?

The commitment means that I do not willingly or knowingly prompt AI tools in my work, whether it be research or teaching.

The nature of the commitment is similar to that of “committing not to eat plastic”. I would not knowingly or willingly eat plastic, but given the current reality of food production, small pieces of plastic will inevitably end up in my body. Similarly, while I do not prompt AI systems, it may be the case that my phone’s autocorrect uses AI technology or that I see an AI summary during a Google Search. In this sense, the commitment should be understood then as being one of principle, rather than a literal statement.

Q: Do you collaborate with people who use AI?

First and foremost, I value building mathematical understanding thru collaboration. I respect the decisions of my co-authors and I do not wish to force my own beliefs upon them.

That being said, I ask my co-authors to not use AI in our projects. I do not wish to carry out research that is driven by AI output rather than human insight. In such cases, I will remove myself as an author of the paper.

Q: Do you read or otherwise engage with AI generated papers?

I have been sent several AI generated manuscripts or notes which were not previously reviewed by a human. In my experience, reading these kinds of texts in detail is a frustrating experience and ultimately a waste of time. For this reason, I will be declining to read fully AI generated text. I am willing to read a human written summary or have a face-to-face discussion about the generated text.

Q: Could it be that AI use in applied fields such as Statistics or Computer Science ought to be different from AI use in pure Mathematics?

While there is of course a distinction between the specific aims of pure and applied mathematical disciplines, I believe that the primary purpose of these disciplines as scholarly pursuits is the same: the improvement of our understanding of phenomena, techniques, and tools. Regardless of whether the specific discipline in question is categorized as being “pure” or “applied”, the fact that AI aims to replace our faculties of thought is what places it in direct opposition to the primary goal of improving our actual understanding.