Sirawit Pongnakintr

สิรวิชญ์ พงศ์นคินทร์

Back to the main page

Why do I choose to study Math/CS?

Before answering that question, let me declare that my view towards mathematics is aligned to mathematical platonism. This is in contrast to some of my friends' views that mathematics is constructed by humans, or that mathematics is just a convenient tool to solve problems. For me, it is more than just a construction or a tool. Mathematics is a realm of unknowns and their connections. The study of mathematics deserves its own purpose, independent of technological advances and independent of applications. However, this doesn't restrict my goals to studying only the pure mathematics. Sometimes I also study mathematics for its applications.

One can characterize my "reason to study" as a combination of simply two things:

• I solve problems because they are fun to think about. It's like a puzzle game where I can spend time thinking on solutions to the puzzle. This is the same reason as why I like to play puzzle games, and this is more CS-oriented.
• I study deeper mathematics because I want to uncover the underlying truth. I want to know more about something that I don't know. It's fun when I learn something I don't understand before. This is the reason why I study topics in pure math.

Seeing open problems and conjectures, these two motivations push me forward to think about things and to go try solving them for fun (even if I almost always failed).

Current Opinion and Position on AI

Recently (at the time of writing this), there are incredible progress going on in mathematics and theoretical computer science coming from AI. For example, Planar unit distance problem from OpenAI on May 20, 2026, Jacobian conjecture counterexample from levent on Jul 20, 2026, ten advances from OpenAI on Aug 1, 2026, Riemann zeta density lower bound from Anthropic on Aug 10, 2026, Hadamard matrices of order up to 2000 from levent on Aug 12, 2026, a solution of Navier–Stokes from OpenAI on Sep 8, 2026 , and a release of 372 result families in 719 manuscripts from OpenAI on Oct 6, 2026 . I congratulate them all (both the people prompting the AI and the people developing new AI capabilities). I hold the simultaneous position of being excited with these new results and being depressed with the possibility that there might be nothing left for humans to be done in the near future. This led me to the current decision to give up on the dream of becoming a mathematician (since there might be no such thing as a mathematician anymore). This may sound depressing, but I can assure that I do not give up on mathematics. I am, and I will, still do mathematics, not because I want to achieve a result, but because I like it and I have fun doing it. I therefore retract my dream of becoming a contributor of mathematics and instead choose the lesser path of becoming simply a person who loves mathematics and want to know more about mathematics. Here is a more detailed explanation of my opinions on the PhD.

The next question is whether I support or oppose AI in the current state. My current opinion is that the way the AI is disrupting mathematics is doing some harm and people are now acknowledging the harm. However, I do not think stopping AI, nor actively decelerating the progress, would be a better choice. I think AI should be accelerated as much as the economic invisible hand wants it to be, not externally accelerated nor decelerated. And I also think AI progress is inevitable: there might be one day when AI can do effectively everything better than human mathematicians. The emphasis, in my opinion, should rather be: how do we adapt to the changes. My dream is adapted, as stated in the above paragraph. I argue that now institutions should adapt as well. In the end, I hope that the society will still let me study mathematics, even if it makes effectively no material benefit to the society. I doubt whether the majority would bite that bullet, but for me, I've decided to commit to this hope.

Now, how do I use AI in my work? Recently I've used a lot of Codex for coding tasks outside of competitive programming. I've also used ChatGPT in the usual Chat mode to talk to when I have questions or confusions about a particular topic or idea in mathematics. My general rule is: for any task that I do not (currently) wish to understand deeply, I offload the most of the work to AI (this includes ML/AI programming, website development, literature search, daydream idea checking, grammar and spell checking). For the tasks that I wish to really understand deeply, I try to prevent myself from asking AI (or even humans) immediately. I try to instead spend some time thinking on it on my own, until the progress plateau, then I ask AI or humans for hints, and then for the full solution. These include competitive programming, competitive mathematics, attempts on understanding basic well-known mathematics, solving classical exercises, parsing complicated ideas in old published papers, etc. In particular, I actively try not to let AI do research for me and claim the result as my own. More precisely, I try to decouple research into two layers: the understanding layer and the exploring layer. My main job is to try to understand things. The AI's main job is to try to explore things. If the AI finds something new and it's worth my time and effort, then I'll spend time and effort trying to understand the idea before communicating publicly. If the result is important enough to be communicated, but I fail to understand, then I will put a big disclaimer out there that I do not understand it.

Stupidity doesn't stop me from doing math. I may try and I may fail. And at least I'm proud to fail rather than not even give it a try. I dare to dream. To be a great mathematician who stands alongside the frontiers of humanity. To be a person doing mathematics with happiness, and learning theories at the forefront of the humanity's knowledge. And even if I fail, I'd rather die trying hard.