Big Data: AI and Math

By Helix

Big Data is our brand-new weekly science and technological news column, published jointly by Helix, Milton’s STEM magazine, and TMP. Every week, a Helix writer authors an article covering the most pertinent science and technology news of our day, which undergoes scientific review at Helix and journalistic review at TMP. We believe a marrying of science and journalism is most necessary to dispel the collective myth that the only crime we commit with AI is using it to write our English essays.

We cover AI, but not just AI. As you will see in our inaugural article, written by Cindy Zeng ’27, our focus is still how we, as a society, may navigate science and technology humanistically. We hope to preserve a human ethic and perspective on the changing technological landscape of the world and to provide our community with the necessary information to discuss one of the most pervasive issues of our time with literacy and grace.

Love,

TMP & Helix

Is there a limit to human reason? Gödel’s incompleteness theorems, demonstrating the impossibility of a set of axioms that can prove all of mathematics, say yes. In William Barrett’s seminal introduction to existentialist philosophy, Irrational Man (1958), he claims that Gödel’s theorems show “mathematics can never be turned over to a giant computing machine… mathematicians—the human beings who construct mathematics—will always be in business.”

Barrett’s prediction, published nearly seventy years ago, seems to be wrong. On September 8, 2026, OpenAI published a proof of the Navier-Stokes problem, one of the “Millennium Prize” questions, which are seven of the most difficult unsolved mathematical questions. Navier-Stokes asks whether equations describing movement of fluids (e.g. water and air) can break down in specific conditions. Terence Tao, a professor at UCLA, describes the problem: “could you set some fluid in motion, watch it flow, and then it blows up?” The OpenAI proof claims that theoretical conditions exist where, yes, water could spontaneously explode.

$1 million dollars of prize money is on the table, and more importantly, the future of higher mathematics. OpenAI reports that it took over 10,000 AI agents operating on an internal model more powerful than their best public model, GPT-6 Astra, and only around eighty-eight hours to formulate a proof to a problem that mathematicians have been working to solve for decades.

Tao, along with twenty-four other winners of the Fields Medal, the most prestigious prize in mathematics, published a statement titled “A Severe Misalignment of AI in Mathematics” three days after OpenAI’s proof. They declare that mathematicians work to enhance the understanding of the entire field and, eventually, the wider population. Likening the field of mathematics to “a miniature version of humanity,” they urge that only human discussion, explanation, and transmission of mathematical ideas can sustain the life of the discipline. They critique that, as AI use proliferates in mathematics, independent critical thinking skills in the next generation of mathematicians may be impaired.

Furthermore, mathematical research is a creative profession, a fact often overlooked. Mathematicians raised concerns over whether unpublished work by Levent Alpöge, an employee of OpenAI’s biggest rival Anthropic, and Tristan Buckmaster was plagiarized by OpenAI. OpenAI confirms that they were motivated to pursue the problem by rumors of a Navier-Stokes solution by Alpöge and Buckmaster, but denies that their work was used for the OpenAI proof. The proof is currently undergoing cross-examination, and the company has said it does not plan to accept the 1 million dollar prize money.

AI’s relentless pursuit of faster and faster proofs without consideration for process is slowly reorienting mathematics away from human values of scientific inquiry. Two weeks ago, on October 6, 2026, OpenAI released 722 papers-in-progress for over 372 distinct mathematical problems all at once. Many of these proofs overlap with and even borrow exact processes from yet-unpublished work by human mathematicians. The future of AI in mathematics will undoubtedly impact its role in other sciences; its integration into mathematics is almost guaranteed. Whether for constructive or destructive effects, only the humans controlling the models may decide.