Jay Forrester, SM ‘45, arrived at MIT in 1939 and fought World War II in the Servomechanism Laboratory. In the basement of Building 10, he helped design hydraulic controls for radar antennas and gun mounts. As the war in Europe drew to a close, many of the Servo Lab’s military projects began winding down. So Forrester submitted his thesis and prepared to move on.
But Luis de Florez ‘11, who headed the navy’s Special Devices Section and would soon be named rear admiral, had been talking with his alma mater about designing a precision flight simulator for the navy. He envisioned a general-purpose machine that could be used to test proposed designs of new airplanes. Intrigued, Forrester stayed at MIT to lead the project. In late 1944, he began conceptualizing an analog computer that would run the Airplane Stability and Control Analyzer. (Digital computers were then virtually nonexistent.) He worked on it for a year, but “it didn’t seem feasible to build an analog computer to do what was required,” he says.
In the fall of 1945, Forrester bumped into Perry Crawford ‘39, SM ‘42, on the steps of 77 Mass. Ave. He told Crawford, who was about to leave MIT to work for de Florez, of the problems his lab was having with the flight simulator. Crawford suggested considering digital computing: it was more precise and could enable parallel processing. Being, as Forrester puts it, “completely uninhibited by organizational charts,” Crawford would prove an effective champion of the unproven technology. In March 1946, the navy agreed to fund development of a digital computer called Whirlwind to run a general-purpose flight simulator.
Forrester and Robert Everett, SM ‘43, led a team that got Whirlwind running–at least intermittently–in 1949. It used 4,500 vacuum tubes, occupying the space of five or six offices. But vacuum-tube memory proved wildly expensive and unreliable. Forrester recalls that Whirlwind “always worked when the admirals were there, but it quit pretty much when they walked out of the building.”
That year, Forrester saw an ad for a magnetic material called Deltamax, and it got him thinking about the possibility of storing digital data with magnetic fields rather than electrical charges. He needed a material the polarity of whose magnetic field was hard enough to change that it wouldn’t accidentally flip, causing data to be lost. Metals were probably too slow, but what about magnetic ceramics? If magnetic ceramic cores with the right properties were mounted in an array of wires, each intersection could function as an independent switch, storing one binary digit. Although ferrites experts at Philips Research in the Netherlands said it couldn’t be done, Forrester’s team found a German ceramicist who was building television tubes and asked him to try. Occasionally, he produced a satisfactory magnetic core, proving it was possible. “Often the art runs ahead of the science,” Forrester says. “You find out you can do something and then find out why.”
“We spent the better part of $1 million figuring out what was going on,” he recalls. “There were probably 50 variables,” including which elements to use, how finely to grind them, at what temperature to fire them and for how long. By 1952, Forrester’s lab had perfected the material that made magnetic-core memory viable. The result, Forrester says, was memory that “was completely reliable and relatively inexpensive.”
The development of Whirlwind itself ultimately eclipsed the flight-simulator project. Forrester’s team had stopped working on the simulator in June 1948, and navy support for Whirlwind waned. “We probably would’ve been pretty well budgeted out of existence, except for Russia exploding the atomic bomb” in 1949, says Forrester. “That brought into sharp relief the weakness–or almost nonexistence–of the American air defense system.” The air force stepped in to fund Whirlwind, and Forrester led the division at MIT’s Lincoln Lab that designed the Whirlwind-based SAGE air defense system, which operated until 1983.
Whirlwind’s greatest legacy, however, was magnetic-core memory, which Forrester says made the first two-plus decades of digital computing possible. “We spent about seven years trying to convince the industry that it was a good idea,” he says. “Then we spent about seven years in patent courts trying to convince them that they hadn’t all thought of it first.”