Every year, on 14 March, the global community celebrates the International Day of Mathematics (IDM), a date chosen for its elegant connection to the mathematical constant (Pi ≈ 3.14). What began as an informal “Pi Day” celebration has evolved into a worldwide initiative led by the International Mathematical Union (IMU), involving schools, universities, museums, and science centres in more than a 100 countries. The theme for 2026, "Mathematics and Hope" draws on the wisdom of Greek philosopher Thales, who noted that hope is the most universal of human possessions; today, we recognise that mathematics is equally universal. Yet, as we commemorate this day, a profound question looms over the festivities: in an era where artificial intelligence (AI) can solve Olympiad-level problems and verify complex theorems, what is the future of human mathematics?
A brief history of Pi and participation
The origins of this celebration are rooted in the playful observation that the date 14 March, when written in the month/day format used in many countries, matches the first digits of Pi. Over time, this evolved into the formal IDM, a project supported by numerous international and regional organisations to highlight how mathematics is everywhere. Past challenges have invited the public to create math remixes and mathematics for a better world posters, and even mathematical sculptures, proving that the discipline is as much an art as it is a science. However, the 2026 theme of ‘Hope’ feels particularly poignant as we stand at a technological crossroads.
The AI disruption: From hallucinations to silver medals
The relationship between mathematicians and AI was one of skeptical amusement. Early large language models (LLMs) like Generative Pre-trained Transformer Four were notorious for hallucinations, generating plausible-sounding but logically incoherent proofs. These models were essentially next-word predictors or auto-complete on steroids, capable of flubbing simple arithmetic while excelling at linguistic patterns.
This skepticism was challenged in July 2024 when Google DeepMind’s AlphaProof and AlphaGeometry Two achieved a silver-medal standard at the International Mathematical Olympiad (IMO). Together, these systems solved four out of six exceptionally difficult problems, including the hardest problem in the competition, which only five human contestants managed to solve. Unlike traditional LLMs, AlphaProof utilises a neurosymbolic approach, combining the intuitive pattern recognition of neural networks with the rigourous logic of a formal proof language called Lean (https://lean-lang.org/). This breakthrough demonstrated that AI could move beyond mere text prediction to trustworthy reasoning.
A research paper published by Google DeepMind on 6 March titled Towards Autonomous Mathematics Research demonstrates how AI may soon assist mathematicians at the level of genuine research. The study introduces an AI research agent called Aletheia, which can iteratively generate, verify, and refine mathematical arguments while consulting the existing literature. In experimental studies, the system produced several publication-grade mathematical results, including solutions to a number of long-standing problems from the Erdős conjecture database (https://www.erdosproblems.com/). It also demonstrated the ability to collaborate with human mathematicians by proposing proof strategies and conjectures that researchers later developed into full mathematical papers. While the authors emphasise that AI has not yet matched the creativity of expert mathematicians, the work shows that machines may increasingly function as research collaborators rather than mere computational tools.
The secret sauce: Formal verification and Lean
The survival of mathematics in the AI age depends largely on a concept known as formal verification. Tools like Mathlib, a community-driven library of formalised mathematics written in Lean, provide a bedrock of over 210,000 theorems that machines can use to verify new proofs with absolute certainty. As British mathematician Andrew James Granville notes, Lean acts like an obnoxious colleague who pesters you until every tiny logical gap is filled.
This technology is transforming how mathematics is done. One of the world's most renowned mathematicians, Australian-American mathematician Terence Chi-Shen Tao has utilised Lean to crowdsource complex projects like the Equational Theories Project. By breaking a massive problem into 22 million micro-tasks, Tao was able to collaborate with 50 people, many of whom were not professional mathematicians because the Lean compiler acted as an automated grader. This decentralisation of mathematics breaks the trust barrier, allowing contributions from anyone as long as their code passes the machine's rigours.
Will AI replace the mathematician?
As machines begin to write their own proofs, some fear a disgusting future where human-readable, elegant proofs are replaced by incomprehensible, billion-line computer outputs. If a machine solves a problem but no human understands why it is true, has mathematics really progressed?
The consensus among experts like Tao and Canadian-American mathematician Ravi Damodar Vakil is not one of replacement, but of augmentation. AI is expected to become a universal translator and a co-author. It will handle the long tail of medium-difficulty problems that humans simply don't have the manpower to attack, freeing human geniuses to focus on bold new approaches and creative theory-building. As Vakil suggests, while the generation trained in a pre-AI world might find these tools clunky, the next generation will speak AI as a first language, using it to discover connections that were previously invisible.
The phase change: Mathematics and hope
We are currently waiting for a phase change in mathematics, much like the one seen in chess when computers began to beat grandmasters. Interestingly, despite computers being "better" at chess, more people enjoy playing and studying the game today than ever before. The same is likely true for mathematics. Even in a utopian future where AI can prove any theorem, humans will continue to do math because it is the most beautiful thing and provides a unique sense of epiphany.
The theme of "Mathematics and Hope" for 2026 reminds us that math is not just about the final proof, but about the human search for truth and definitions. AI will enlarge the pi of mathematical knowledge, create more economically feasible objectives and allow us to explore millions of problems at once rather than just one.
As we celebrate Pi Day, we should look at AI not as the end of mathematics, but as the beginning of its most expansive chapter. The "traditionalist" discipline that still loves blackboards and chalk is finally embracing "big science". Whether it is using SAT (Boolean Satisfiability solver) solvers to turn theorems into puzzles or using LLMs to suggest new conjectures, the future of mathematics is a collaborative effort between human intuition and silicon rigour. Mathematics will survive because it is a universal human possession, and as long as there are humans, there will be a desire to understand the hidden patterns of our universe. Machines may explore vast landscapes of possibilities, but, it is human curiosity that continues to ask profound questions. In this unfolding partnership between mind and machine, mathematics may grow richer, deeper, and more adventurous than ever before. As long as humanity gazes at the mysteries of nature, from the spirals of galaxies to the symmetry of number patterns, the timeless quest to uncover the hidden patterns of the universe will endure.
The writer is a Senior Mathematics Lecturer at the Peradeniya University
-------
The views and opinions expressed in this column are those of the author, and do not necessarily reflect those of this publication