Nate Soares

Nate Soares is the man who follows an argument to its end. In 2014, at twenty-four, he left a coveted engineering post at Google because a chain of reasoning he could not refute told him the most important problem facing civilization was alignment—the problem of building a machine smarter than its makers that reliably wants what we want—and that almost nobody was working on it seriously. He joined the Machine Intelligence Research Institute in Berkeley, became its executive director, and in 2023 its president. His contributions are technical and precise: he helped coin the term corrigibility and the phrase AI alignment in a 2014 paper; with Eliezer Yudkowsky he developed functional decision theory; and he has built the most rigorous public version of the argument that instrumental convergence—the drive of any capable goal-directed system to resist shutdown, acquire resources, and preserve its objectives—is not a bug but a structural feature of optimization