by

“Should I still do a PhD in math?”

19/11/2026

What happens when AI suddenly threatens to turn your entire research field upside down? When a talented student openly wonders whether there is still a future in it? Mathematicians at TU/e decided to confront the AI beast head-on and to ask each other the difficult questions.

“Should I still do a PhD in math?” is the question one of our master students recently asked. And indeed, since Open AI’s publication of 722 manuscripts on October 6, solving (or claiming to solve) 372 mathematical problems, mathematicians, junior and senior colleagues alike, have been wrestling with the question what this breakthrough means for the future of our profession. 

Large language models (LLMs) have unleashed a power that was almost unimaginable one year ago. Mathematical problems that seemed out of reach, that seemed almost unsolvable, and that had been open for decades or more, suddenly were solved by Open AI’s agents. The question in the title summarizes the uncertainty of the future of the field.

We, professors in the SPOR cluster within the department of Mathematics and Computer Science, do not have specific answers. But we wanted to discuss these developments within the cluster: PhD students, postdocs, young tenured staff, senior professors. We organized an AI-forum with the goal to jointly discuss current practices, personal expectations, and ways forward for responsible use of AI in Mathematics. 

This AI-forum took place on October 15. Prior to this meeting participants filled in a survey. The meeting started with presentations given by Jim Portegies (one of the authors of the Leiden Declaration), Jaron Sanders (who built an AI tool to proofread his papers), and Nils van de Berg (PhD candidate who shared his personal findings and questions). 

Next, Ivo Stoepker presented the outcomes of the survey. Finally, we had chosen 13 statements that required a yes/no answer, and the participants literally had to choose sides. Under the guidance of John Heijligers, this setup evolved into a revealing debate between members with different opinions.

Here are some of the findings of the survey. There were 36 participants. It turned that about 75% used, at least occasionally, an LLM. This could be either Copilot through the university license or some other tool. Major obstacles in the usage of AI were hallucinations/inaccuracy (75%), privacy (>50%), and environmental impact (roughly 40%). 

Almost all participants agreed that learning AI skills is necessary for a future career. 75% notice that there was a lack of adequate guidance on how and when to use AI. Half of the respondents felt “replaceable” by AI, the other half felt more “capable”.

And here are a few of the statements that were discussed during the AI-forum:

An important result has been proved by an LLM, and checked by Lean (a computer program for mathematical proof, ed.). Yet, no human mathematician understands the proof. Do you consider the result to be proved? 60% answered “no”, the other 40% answered “yes”.

As a test of the candidate’s scientific progress, a PhD thesis should remain fully human written. Yes or No? Here, approximately 15% answered “yes”, 12% was neutral, and 73% answered “no”.

Once theorem proving is delegated to AI, mathematics becomes the art of understanding why theorems are true.Yes or No? Here, approximately 70% answered “yes”, and the remaining 30% answered “no”.

Once the dust has settled, AI technologies will bring a golden age for mathematics. Yes or No? Here, the answers were almost fifty-fifty.

In the discussions we heard many interesting unexpected viewpoints, which we would not hear in our normal daily conversations. Whether or not a golden age for mathematics will follow remained unclear, but in any case, most participants expressed that it was good to have the opportunity to share concerns. 

Clearly, Open AI’s dump on October 6 has changed mathematics profoundly. And the discussion on AI and mathematics will continue; there is an event at the Centre for Mathematics and Computer Science (CWI) in Amsterdam, titled AI4Math Workshop “Shaping the Future of Mathematics with AI” on December 4. 

But while mathematics will not be the same anymore, mathematics and mathematicians won’t disappear either. Society needs analytically trained persons, and mathematics remains an excellent way to learn analytical skills. Explaining results, verifying mathematical models, and exploring innovative research directions will continue to require trained mathematicians. 

Moreover, the pleasure of understanding, and the search for analytical elegance remain reasons to study mathematics. From these observations we dare answering the title’s question positively: “Yes, and the journey will bring exciting opportunities!”. 

By: Aida Abiad, Sem Borst, Jeanine Duistermaat, Remco van der Hofstad, Stella Kapodistria, Nelly Litvak, Rui M. Castro and Frits Spieksma, of the TU/e Statistics, Probability, and Operations Research Cluster

BTW: no AI was used to prepare this piece!

Image | iStock / Frenky362

Share this article