PERSON

Rudolf Kalman

The Hungarian-American engineer who gave the twentieth century its mathematics for knowing the unknowable—whose 1960 predict-correct filter flew men to the Moon and now returns, sixty years later, as the hidden-state architecture challenging the transformer at the frontier of AI.

Rudolf Kalman built the most widely deployed estimation algorithm in the history of engineering, and he spent the last third of his life insisting that most of the people using it did not understand what it was for. His 1960 paper introduced the recursive filter that now runs inside every phone’s GPS, every autopilot, and the Apollo guidance computer before them: a loop so simple it can be stated in two beats—predict what you expect to see, correct against what you actually observe—and so deep that the field is still mining it. The filter is the exact, closed-form solution to a Bayesian inference problem under linear Gaussian assumptions: it maintains not merely a best guess about a hidden state but a full probability distribution over that guess, expanding when it predicts and contracting when it measures. In the same years he formalized observability—whether a system’s full internal state can be reconstructed from available measurements—and its dual,

Keep reading with YOU ON AI