Wiener's anti-aircraft predictor
In September 1940 Norbert Wiener offered his services to Vannevar Bush, who ran American war research, and took up the anti-aircraft gunner's problem: predicting where an aircraft will be when the shell reaches it. On 1 December the NDRC let MIT a contract, and with the engineer Julian Bigelow Wiener built a predictor that computed the target's future position from its past course.
Why it matters
The feedback idea that became cybernetics did not arrive from biology or from computing but from a gunnery problem. The atlas holds Wiener from 1943, by which point the idea was already general; this is the particular military problem it appeared on, together with the fact that the problem itself was never solved.
The predictor had to lead the target by twenty seconds; on the two courses the Army supplied for the test, the shell's flight lasted from 20 to 35 seconds. Wiener tried his first circuits on Bush's differential analyzer: what followed a straight line well overshot enormously as soon as the aircraft changed course. He turned to statistics and took as the best prediction the one with the least mean-square error. The NDRC printed this theory early in 1942, first as secret and later as restricted; its yellow cover and hard mathematics earned it the name Yellow Peril, and it appeared openly as a book in 1949. Bigelow, who joined in the early winter of 1940, built a test rig: a spot of light moved across a wall and an operator followed it with his own spot through a control that deliberately falsified the feel of his hand. On those records the predictor did predict. On real tracks from the Camp Davis range, with a twenty-second lead and ten seconds of past, the method scored 43 hits against 41 for the simpler memory-point method on one course and lost on the other, 48 against 53. In his final report of 1 December 1942 Wiener wrote himself that this was scarcely enough to justify new apparatus and that his method's advantage would show only at much shorter times of flight. The historian Peter Galison sums it up: Wiener "never got it to work more than a couple of seconds in advance." The project was ended in January 1943; the principle later went into data smoothing. What matters is what followed. The pilot and the gun-pointer were both links in one system, and Wiener and Bigelow concluded that feedback is central to voluntary human movement. They asked the physiologist Arturo Rosenblueth whether there is a disorder in which a hand reaching for a pencil overshoots and starts to oscillate, and he named purpose tremor. Their joint paper of 1943 grew out of this, and cybernetics after it. The record rests on Wiener's final report and his letter to Warren Weaver, declassified in 1960, on the Introduction to Cybernetics, on the chapter by Mindell, Segal and Gerovitch, and on the Galison interview. The 1942 monograph itself was not read here.