← Back to Home
ON THIS DAY 2026-10-25

1811: Évariste Galois is born

On 25 October 1811, Évariste Galois was born in Bourg-la-Reine, near Paris. He would live only twenty years, dying after a duel in 1832, yet in that short span he reshaped mathematics so thoroughly that his name became a permanent fixture of the field: Galois theory.

Galois was trying to answer a deceptively simple question: when can a polynomial equation be solved using radicals — the familiar toolkit of roots and arithmetic operations? His answer required inventing a new kind of object, the mathematical group, to capture the symmetries of an equation's solutions. Whether an equation was solvable, he showed, depended entirely on the structure of this symmetry group. It was a radical shift: instead of computing answers, mathematics now classified the hidden structures that made answers possible or impossible.

Group theory, born from Galois's scribbled notes the night before his fatal duel, went on to become one of the great unifying languages of science. Physicists used it to describe conservation laws (via Emmy Noether's theorem) and particle symmetries. Cryptographers and coding theorists built on algebraic structures to protect and correct digital information. And in the last decade, the same ideas have resurfaced directly inside artificial intelligence.

  • Convolutional neural networks are, at heart, built to respect translation symmetry — a simple group action on images.
  • "Group equivariant" neural networks, formalized by Cohen and Welling in 2016, generalize this idea to richer symmetry groups, letting networks learn more efficiently from less data.
  • Systems like AlphaFold encode 3D rotational and translational symmetries (the group SE(3)) directly into their architecture, which is part of why they predict protein structures so accurately.

None of this was imaginable to Galois, who wrote for an audience of future mathematicians he hoped might understand him after his death. But the thread is direct: his insight that structure, not brute computation, is the key to understanding complex systems is exactly the principle guiding some of today's most capable AI architectures. When a modern model exploits symmetry to learn faster and generalize better, it is, in a quiet way, still working within a framework a twenty-year-old sketched out in a single desperate night.