In 1948, Norbert Wiener named a new science. He called it cybernetics, after the Greek word for steersman. His book of the same name proposed a unified study of control and feedback in living and mechanical systems. By the 1970s the field had largely dissolved.

Wiener was MIT's resident genius and absent-minded professor: child prodigy, Harvard PhD at eighteen, the kind of mathematician colleagues spotted continuing his train of thought in the parking lot for ten minutes before going inside. After the war he gathered a research team that included Warren McCulloch and Walter Pitts. In the early 1950s, Wiener abruptly cut all contact with the group. Pitts was devastated, his career declined, and the reasons remain disputed.

Cybernetics is being rediscovered without being named. Modern reinforcement learning is a study of how agents learn through feedback from environments. Modern agentic AI is a study of how systems regulate behavior in response to outcomes. These are Wiener's problems with new mathematical apparatus.

Wiener anticipated several problems modern AI is now encountering for the first time. The control of self-modifying systems. The stability conditions for feedback loops involving humans. The risk of optimizing the wrong objective in a system that learns. He named these in 1948 and 1950. The current alignment community is rediscovering them because the institutional memory did not pass through computer science.

Reading The Human Use of Human Beings (1950), Wiener's accessible follow-up, is closer to reading 2026 alignment discourse than any computer science textbook from the same period.