1870: Jean Perrin is born

On 30 September 1870, Jean Baptiste Perrin was born in Lille, France. A physicist whose meticulous experiments helped convince the scientific world that atoms were real, Perrin belongs to the deep prehistory of artificial intelligence through the statistical ideas his work helped establish.

In the early twentieth century many scientists still doubted the physical existence of atoms. Albert Einstein had published a theoretical explanation of Brownian motion—the jittery dance of tiny particles in liquid—in 1905, treating it as the visible effect of countless invisible molecular collisions. Perrin set out to test the theory. Between 1908 and 1913 he performed elegant experiments measuring the displacements of suspended particles. His results matched Einstein’s predictions so precisely that skepticism about atoms largely collapsed. For this work he received the Nobel Prize in Physics in 1926.

Why does this matter for AI? Perrin’s confirmation cemented statistical mechanics as a legitimate description of nature. Ludwig Boltzmann and Josiah Willard Gibbs had already shown that heat and entropy could be understood as averages over myriad microscopic states. Once atoms were accepted, this probabilistic style of thinking became foundational. Claude Shannon later borrowed the same mathematical language when he created information theory in 1948, defining information in terms of entropy. From information theory it is a short step to the probabilistic models that dominate modern machine learning.

Today’s AI systems are steeped in these ideas. Boltzmann machines, named in honour of the statistical physicist, were early generative neural networks. Stochastic gradient descent, the workhorse of deep learning, relies on noisy estimates of averages. Diffusion models and energy-based models used in image generation draw directly on the mathematics of statistical physics. Even the way large language models represent uncertainty traces back to the same intellectual current.

Perrin never imagined electronic computers or neural networks. Yet by giving experimental solidity to the statistical view of the world, he helped prepare the conceptual ground on which artificial intelligence now stands. On this day we remember a scientist whose careful measurements of dancing particles quietly shaped the probabilistic soul of modern AI.