Symbolics Paces ‘Smart’ Computer Field
Early last year, the Securities and Exchange Commission was looking for a way to computerize the job of sifting through millions of pages of documents to keep track of companies taking steps to thwart takeovers.
It was a tough assignment because few computers can digest legal language and recognize its subtleties. Moreover, identifying anti-takeover measures is especially complicated because many companies shroud their actions in vague phrases such as “fair-price amendments” and “shareholder-rights clauses.”
The task, however, was accomplished by an accounting firm that used computer equipment built by Symbolics, a fast-growing company based in Concord, Mass., that does all its manufacturing in Chatsworth. Symbolics officials say the system has had almost 100% success in detecting anti-takeover measures in SEC filings.
Dominant in Field
Symbolics computers are built to handle just that sort of job. The company has emerged as the dominant equipment maker in the burgeoning artificial intelligence business by combining the brainpower of computer designers educated at the Massachusetts Institute of Technology with the manufacturing skills of some San Fernando Valley computer executives.
The 6-year-old company is so dominant in its field that one trade publication recently called it the IBM of the artificial intelligence computers. An industry newsletter, Artificial Intelligence Markets of Natick, Mass., estimates that Symbolics last year captured 64% of the $156-million market for so-called symbolic-processing computers, machines that can process symbols rather than the strings of numbers that most computers handle.
Russell Noftsker, Symbolics’ chairman, boasted: “We started the market, and the market has been growing as fast as we’ve been able to grow.”
Artificial intelligence is the catch-all term used to describe computers designed to emulate human reasoning.
Science-Fiction Images
The idea often conjures up science-fiction-inspired images of human-like computers, such as Hal in the Stanley Kubrick film “2001.” In reality, it usually involves getting a computer to perform relatively mundane chores. The machine “learns” the rules of thumb so that it can function somewhat like a human expert.
The computer might, for example, evaluate a bank customer’s loan request based on what it knows about credit history and the amount of income needed to keep up with payments. Or it might help a petroleum engineer find an oil field by interpreting basic geological data.
Programming Left to Others
Symbolics stands apart from many others in the artificial intelligence industry in that it does not write the programs that help people such as petroleum engineers or loan officers solve their everyday problems. That task is left to programmers who tailor the company’s equipment to specific uses.
An advantage of the Symbolics machines is that they run LISP (List Processing Programming Language), a complex computer language but one that is popular among artificial intelligence programmers in the United States.
Unlike most computers, machines capable of running LISP can manipulate symbols, which requires a powerful memory but gives the computers great flexibility.
Founded in 1980 by 21 people, 12 of whom were artificial intelligence researchers from MIT, Symbolics shipped its first product in 1982. As the commercial potential for artificial intelligence systems has become recognized, the company has grown rapidly. Symbolics now has a payroll of 880, with 250 employees in Chatsworth.
In the six months ended Dec. 29, its earnings more than quadrupled, to $5.5 million, while revenue rose 73%, to $51.6 million. To accommodate its growth, the company plans to move its four Chatsworth facilities by October into a nearby 228,000-square-foot complex with more than twice the company’s existing manufacturing space.
Still, the company maintains some traditions that began with its founding, including an occasional “Symbolics Frolics” beer bash at its headquarters.
That headquarters outside of Boston notwithstanding, the company has strong roots in the San Fernando Valley. Noftsker, 44, who has a home in Woodland Hills, was the co-founder of Pertron Controls, a Chatsworth firm that makes computerized welding controls. Minoru Tonai, another of the 21 founders and Symbolics’ vice president of western operations, was an executive with PertecPeripherals, a Chatsworth maker of computer memory devices.
Branching Out Recommended
Before starting the company, Noftsker, the former director of MIT’s Artificial Intelligence Laboratory, and some of the other founders tried unsuccessfully to persuade computer companies to branch out into the artificial intelligence field.
“All of them thought there was no market for the technology,” he said. “The only way to get it working was to prove there was a market ourselves.”
Fortunately, Symbolics’ founders had a head start from their days at MIT.
“They had built the prototypes at MIT. The machines did exist, so it’s not like they started from scratch,” said David Bendel Hertz, director of the University of Miami’s Intelligent Computer Systems Research Institute.
Symbolics has only three direct competitors--Xerox in Stamford, Conn., Dallas-based Texas Instruments and LISP Machine Inc. of Andover, Mass. Symbolics’ market share, according to the Artificial Intelligence Markets newsletter, is five times that of its largest competitor, Xerox. The newsletter estimates that the industry will grow to $1 billion by 1990.
Symbolics officials said they believe that they have established a “de facto standard” for artificial intelligence equipment. “You know you are the leader when everyone compares themselves to you,” said Howard Cannon, a Symbolics founder who is director of marketing.
But some artificial intelligence authorities believe that the field is becoming increasingly competitive. Several large companies, such as Digital Equipment, are building computer work stations that, although not as powerful as the Symbolics machine, can run LISP.
‘There Are Alternatives’
“It’s not clear that we are going to go down the road with a Symbolics-type machine in the future. There are alternatives out there that are staring Symbolics in the face, and Symbolics is smart enough to know it,” Hertz said.
Carol Weiszmann, editor of Artificial Intelligence Markets, sees potential competition for Symbolics from the so-called “LISP on a chip” that Texas Instruments is developing. If developed, she said, that technology could put artificial intelligence into automobiles, medical instruments and military weapons.
In response to its competition, Symbolics is trying to broaden its market. Last month the company announced that this year it will introduce a small, powerful computer designed for end users such as loan officers and geologists. That marks a departure from Symbolics’ practice of building computers for software writers.
At the same time, the company, which has had the reputation of selling expensive equipment, cut its prices 4% to 25%. Its machines now range in price from $44,900 to $250,000.
Problem Finding Skills
Symbolics executives said they believe they can maintain their technological edge. They say their biggest problem in managing the firm’s rapid growth is finding enough highly skilled people to design, make and test its complex equipment.
They also believe that the market for artificial intelligence equipment will grow, arguing that such markets as computer-aided design, factory automation and financial services are virtually untapped.
“We don’t see a leveling off in the near term. Artificial intelligence is adding a whole new element to the data-processing industry. There almost is an unlimited range of opportunities,” Noftsker said.
