François Jacob

The most important sentence ever written about artificial intelligence was written about lungfish. “Evolution does not produce novelties from scratch,” François Jacob declared in 1977. “It works on what already exists, either transforming a system to give it new functions or combining several systems to produce a more elaborate one.” He was describing how a fish’s swim bladder becomes a lung—and he was, with uncanny precision, describing how the transformer architecture was assembled from attention mechanisms borrowed from translation systems, combined with residual connections borrowed from image recognition, stacked in a configuration that happened to train well. Jacob earned this insight through one of the twentieth century’s great careers in molecular biology. Working at the Pasteur Institute with Jacques Monod and André Lwoff, he cracked the logic of gene regulation—the operon model, by which bacterial cells switch genes on and off in response to their environment—and shared the 1965 Nobel Prize in Physiology or