What happens to mathematics when AI becomes an oracle?

Our modern infrastructure runs on mathematics. From the algorithms that decide which post you see on social media, to the logistics that route your Amazon package, to the climate models that warn us of tomorrow’s storms, the language of numbers is the silent engine of daily life. The rise of computers amplified that power, allowing us to juggle figures far beyond the limits of a human brain. Now, as large language models, AI chatbots, and the promise of quantum computing surge forward, the field of “big math” – problems with billions of variables that no one could ever solve by hand – is outpacing the very people who created it. Applied mathematicians Steven Strogatz and Alex Townsend explore this shift in their new book, Big Math: The Hidden Codes That Run Our World, and their insights are reshaping how we think about the future of mathematics.

From Abacus to AI: A 2,000‑Year Evolution

Strogatz frames the current moment as the latest chapter in a long tradition of externalizing arithmetic. He points to the ancient Chinese “fangcheng” board – a rectangular array of bamboo sticks used to solve systems of linear equations – as an early analogue of today’s matrix calculators. “What the abacus did for arithmetic, the fangcheng board did for systems of linear equations,” he tells Ars Technica. The principle is simple: when a problem exceeds the capacity of the human mind, we build tools to extend it.

The leap from hand‑held sticks to silicon chips was dramatic, but the underlying logic remains unchanged. Modern computers let us manipulate matrices with billions of entries, turning what once required a roomful of human clerks into a matter of milliseconds. As Strogatz notes, the lunar module that guided astronauts to the Moon relied on a tiny computer that handled only a few dozen variables at a time – a stark contrast to today’s language models that juggle trillions of parameters. Yet the core idea – offloading calculation to an external device – is the same.

Defining “Big Math” in the Age of AI

When asked to define “big math,” Strogatz emphasizes that “big” is relative to human cognitive capacity. A problem can be “big” even if it involves a modest number of variables, so long as it exceeds what a person can hold in mind. He draws a line between linear and non‑linear problems: linear systems, even at massive scale, remain tractable because they preserve structure that algorithms can exploit. Non‑linear problems, by contrast, quickly become intractable without computational assistance.

This distinction matters because much of traditional pure mathematics stays comfortably in the “small” realm – one or two variables, neat algebraic forms, and elegant theorems. Even cryptography, while mathematically sophisticated, is “small” in the sense that it deals with a limited set of variables, though it hides complexity in non‑linear structures. In the “big math” world, the sheer volume of variables – often billions or trillions – overwhelms manual reasoning, forcing mathematicians to lean on computers for any hope of progress.

The Rise of Massive Models: From Billions to Trillions

Strogatz points out that the largest language models today already operate with “trillions of parameters.” While a trillion is a staggering figure, he argues that infinity – the abstract concept that dominates calculus – is surprisingly tame compared to a billion‑scale problem. Calculus, with its infinite processes, fits neatly into the human mind because it abstracts away the need to enumerate each individual element. By contrast, a billion‑parameter system forces us to confront each element, a task that quickly becomes impossible without computers.

This paradox highlights why AI breakthroughs feel both familiar and alien. The same mathematical foundations that underpin calculus still apply, but the scale forces us to adopt new computational strategies. The result is a landscape where breakthroughs often emerge from the synergy of human insight and machine brute force, rather than from pure, solitary reasoning.

Human Reasoning vs. Machine Discovery

The surge of AI‑driven breakthroughs has sparked fierce debate about the role of human mathematicians. Strogatz and Townsend acknowledge that “the mathematician in each of us will always want to know more,” but they also warn that knowledge must serve a purpose larger than itself. Their book’s central thesis is that as “big math” expands, we must guard against losing the “soul” of mathematics – the human drive to understand, to question, and to apply insight responsibly.

In practice, this means treating AI as an oracle that can suggest conjectures, test massive simulations, or crunch numbers at scale, while retaining human oversight for interpretation and ethical direction. The authors argue that the true challenge is not just technical – it’s cultural. We must cultivate a partnership where machines amplify, rather than replace, the creative spark that has driven mathematics for centuries.

Real‑World Stakes: Gerrymandering and the Quest for Fair Maps

One concrete illustration of “big math” at work is the fight against gerrymandering. Strogatz describes the legal criteria for district maps – compactness, contiguity, respect for natural boundaries, and equal population – as mathematically vague. Translating those vague standards into precise, testable metrics requires massive data analysis and optimization across millions of possible configurations.

By feeding historical voting data and demographic information into large‑scale algorithms, analysts can generate countless alternative maps, each satisfying the legal constraints to varying degrees. This computational power reveals how easily a map can be “stacked” to favor a particular party, exposing the hidden mechanics of partisan redistricting. The ability to enumerate and evaluate billions of potential configurations turns gerrymandering from a political art into a quantifiable problem, opening the door for courts and reform advocates to demand data‑driven fairness.

From Theory to Practice: Applications Across Industries

Beyond politics, “big math” underpins many sectors that define modern life. In medical imaging, algorithms process terabytes of scan data to detect tumors with precision that would be impossible for a radiologist to achieve unaided. In logistics, route‑optimization engines juggle countless variables – traffic patterns, fuel costs, delivery windows – to keep packages moving efficiently. Even the construction of megastructures like the Burj Khalifa relies on massive simulations that model wind loads, material stresses, and seismic activity, all in real time.

These examples illustrate a common thread: when a problem scales beyond human mental bandwidth, the solution hinges on linearity, sparsity, and clever computational tricks. Linear systems can be broken into smaller, manageable pieces; sparse matrices – those with many zeros – allow algorithms to skip unnecessary calculations. Understanding and exploiting these properties is the new frontier for mathematicians, engineers, and data scientists alike.

Preparing for a Future Dominated by “Big Math”

Strogatz and Townsend’s call to action is clear: education and policy must adapt to a world where “big math” is the norm. This means training the next generation not just in abstract theorem‑proving, but in high‑performance computing, numerical analysis, and interdisciplinary collaboration. It also means establishing oversight frameworks that ensure AI‑generated insights are transparent, accountable, and aligned with societal values.

As the authors put it, the work of our century may be “letting our power grow without losing our soul.” The tools are already here – massive matrices, trillion‑parameter models, and quantum processors on the horizon. The challenge is to wield them wisely, ensuring that the hidden codes running our world serve the public good, not just the interests of a privileged few. In the end, the mathematics that powers our lives will remain a human story, even if the calculators we use have become far smarter than any of us.

This article was produced with AI-assisted research and editorial support. Reporting is based on the source material cited below. Sources: Ars Technica; arstechnica.com; Global1.News (12 November 2026).

By Jessica Ali, Staff Writer