CALABASAS — Paul Griffith got his start in engineering as a boy helping his dad fix broken televisions and radios in the rural town of Johnstown, Pa. But when Paul was 10, his dad lost his business because the TV and radio tubes he repaired were suddenly replaced with transistors.
"He could understand when there were tubes in the sets, but then, when transistors came along, it was a different type of technology, and he fell behind because he had nothing to teach him the new way of doing things," Griffith said.
Twenty-three years after his father gave up and went to work in a steel mill, Paul Griffith is president of Micro Expert Systems, a software company developing several computer products, including a system that helps tutor workers to adjust to newer and more complicated technologies.
Micro Expert specializes in artificial intelligence, which creates computers that will think and act like people and help them solve complex problems.
But Micro Expert is still a company in the idea stage and has only eight employees. Its revenue for the year ended March 31 was only $116,000, and it made a $2,400 after-tax profit.
"Everything we do is research. None of it is out in the field yet," said Griffith, 33, although the company has landed some promising research contracts with the National Aeronautics and Space Administration and the Army.
Indeed, for all its technical wizardry, the one thing the artificial-intelligence industry hasn't figured out yet is how to make a profit. Today, there about 450 artificial-intelligence companies with an aggregate $1 billion in revenue, but very few of them are making any money, according to Harvey Newquist, editor of AI Trends, a newsletter in Scottsdale, Ariz.
Twenty-one years after artificial intelligence was developed, it is still too impractical and expensive for almost anybody to use except the military and federal government. Many of the companies set up to sell artificial intelligence to corporations and small businesses are suffering. "In 1987, a lot of these companies had layoffs, losses and nasty annual reports," Newquist said.
Despite the long odds, Micro Expert was founded two years ago by Griffith and former UCLA professor Phil Borden because they, too, thought that artificial intelligence had commercial applications. The company's largest research contract is a two-year, $500,000 deal with NASA.
Micro Expert is developing software that will be used on the space shuttle to help it dock with a satellite in need of repair, or dock with the space station that is set for completion in 1996, if all goes well.
The idea behind this project is to teach computers to visually recognize objects, in this case, the space station or a satellite. Scientists call it machine vision. "The ability to recognize images is something humans can do very well," said Borden, 45. "But computers are lousy at it."
Micro Expert's software will not only let the shuttle see its target, but will realign the shuttle so the two can meet. "Right now, the astronaut looks out the window and sort of guesses the distance," said Kent Dekome, an electronics engineer at NASA's Lyndon B. Johnson Space Center in Houston.
Micro Expert also is developing tutoring systems. It has a $400,000, two-year contract with the Army to use artificial intelligence to teach recruits when to fire Hawk surface-to-air missiles and how to repair them. The system acts much like a flight simulator and enables students to practice figuring out whether an oncoming plane is a friend or enemy. Micro Expert's software simulates various tactical defenses used by enemy fighter pilots, including jamming radar and flying close to friendly craft. It also simulates electrical failures of the radar system and guides the students in repairs.
"You teach skills differently from how you teach knowledge," Borden said. "You can describe in great detail how to swing a bat, how to spot a curve ball, but when you're done with the description, the operator still may not be able to hit a baseball. You have to have ways to simulate the swing of the bat, how fast the ball is coming in, etc."
Micro Expert Systems also has an $8,000 pilot contract with USC to help scientists decipher the earliest known translation of the New Testament Gospels, which were written on animal skins about 1,200 years ago, then erased and covered over with other writing.
Micro Expert has developed some computer software that helps a computer memorize every twist and curve of the visible letters in the indentations of the animal skins. The computer study is so detailed that it divides every letter into dots smaller than grains of sand, then stores those letters in its memory to help identify other letters that are difficult to discern.
Met at Consulting Firm
Borden and Griffith met at a Culver City engineering consulting firm. Borden, especially, thought that computer engineering could stand improvement.