Many academics see research as an end in itself. After all, the purpose of a university is to develop knowledge-not to make a profit. But the researchers who have worked at the MIT Laboratory for Computer Science (LCS) since its inception 35 years ago have shown a keen sense of market potential.
Floating on a secure base of defense funding, the laboratory has over the decades served as a reservoir of technical creativity. LCS researchers have frequently found the Lab a comfortable base from which to launch companies that commercialized their MIT work while maintaining their academic connections.
LCS Director Michael Dertouzos, who started Computek in 1968 to commercialize a graphics device he had invented, sums up the appeal of this dual existence. “I had these two lives,” recalls Dertouzos. “When I got bored [at LCS] I went to the company and got stimulated with real engineering. And when I got bored there with real engineering, which was often,” he came back to MIT. Computek grew to 120 employees before Dertouzos sold the company in 1977 and returned full time to academia-a pattern that was to be followed by many LCS spinoffs through the years.
Infocom: The Legacy of Zork
Many new businesses fail, and those from LCS are no exception. But while one measure of success is financial, another is the intellectual legacy that a company leaves behind. And along this dimension, LCS spinoffs have few equals.
Consider Infocom, the company created in 1979 by Project MAC founder Joseph C. R. Licklider and nearly a dozen other LCS researchers. Infocom sold a peculiar kind of computer game known as “interactive fiction.” Best exemplified by Zork, the games were puzzles: The computer would print a description of the “room” that you were in, and sitting at the computer you would type back instructions on where to move and what actions to take.
Though its product may have seemed frivolous, Infocom was a technological pioneer. To accommodate the many different kinds of computer systems in use at the time, Infocom created a virtual computer called the Z Machine. The Z Machine served as a kind of buffer between the programmers and the outside world of multiple, incompatible computer formats. The first copy of Zork sold in November 1980 and ran on Digital’s PDP-11 minicomputer. A month later, the company was selling Zork for Radio Shack’s new TRS-80 microcomputer. In February 1981, Infocom made a version that ran on the popular Apple II-and proceeded to sell 6,000 copies of the game over the next eight months. Infocom ultimately created 35 different games, and in 1984 had sales of $10 million.
Infocom couldn’t sustain its growth, though, largely because the company was divided against itself. Although games supplied the revenue, Infocom’s management was determined to develop a corporate database tool called Cornerstone. By June of 1985 more than half of the company’s 110 employees were working on Cornerstone. The project became a black hole, sucking up development dollars but never yielding a finished product. In December, Infocom finally shut down its business product division and laid off its staff, but it was too late. In June 1986, with just 40 employees left, the company was sold to California-based video-game maker Activision for $9 million. Three years later, all but five of Infocom’s 26 employees quit or were laid off, and the tattered remnants of the company were absorbed into Activision’s operations. Licklider stayed at LCS, becoming its director. He retired from MIT in 1985 and died in 1990.
While Infocom failed as a business, the company broke important new ground that is still being explored today. Infocom enthusiasts have created Z Machine interpreters for more than 25 different systems-systems as diverse as Windows 95 and the 3Com Palm Pilot. Infocom’s philosophy of making programs that run on any type of computer system stood in dramatic contrast to the prevailing way of doing business at the time. But this idea of a “portable environment” has, in the era of the Internet, taken hold in the form of the Perl and Java programming languages.