Monday, October 14, 1991 - Page updated at 12:00 AM

E-mail article     Print

Knowbots: Thinking Computers On Horizon

It's easy for crew members on "Star Trek."

When they have questions, they merely ask the ship's computer, which responds in a soothing female voice. For today's computer users, drowning in a sea of disorganized data, it's not that simple.

But help is on the way. Imagine a thinking electronic librarian, a Max Headroom who not only compiles data but analyzes it. A computer that understands spoken commands and presents the information in whatever form the user desires - text, sound, images, or any combination of the three. This computer even has a "face" to go along with its "voice."

Scientists are doing just that, imagining - and developing - what's called a "knowbot." The user-friendly device could assist children with homework, medical researchers finding cures for diseases and businesses seeking market trends.

It's the Holy Grail of computing, says Grant Fjermedal, Seattle author of "The Tomorrow Makers," a book about computer scientists.

Today's computer can tap into oceans of information. CompuServe, for example, has 1,500 databases and offers access to more than 100,000 publications, credit profiles of 13 million organizations, movie reviews, an encyclopedia with 10 million words - and more. And that's just one of a growing number of electronic services.

US West Communications announced last week it plans to start its Community Link information service in Seattle next year.

But sorting through all this information is overwhelming, if not impossible. People are lost in a "magnificent data wilderness," says Fjermedal.

Many computer scientists are working to tame that wilderness, focusing on different pieces of the puzzle.

Microsoft is pursuing Chairman Bill Gates' vision: Information at Your Fingertips. The company is working on ways to make it easier to record and summon information. It is close to releasing software for notebook-size computers that recognize handwriting. Its popular Windows program allows users to control a computer by clicking images rather than typing arcane commands.

AT&T's Bell Laboratories is researching ways that computers can understand commands spoken in plain English. One experiment involves speaking a simple question to the computer, "What is the cheapest flight to San Francisco?" Advanced computers can understand some speech, but the trick is getting them to understand, say, a Boston accent or "Valley Girl" talk ("Like, wow, what is the cheapest way to the Bay?")

"By the end of the decade, machines will be routinely talking to people and people will be talking back to machines," said A.G. Fraser, executive director of Information Sciences at Bell Labs, in a company publication.

Today, a system costing a few thousand dollars can recognize speech spoken normally by anyone. The system, though, is restricted to about 200 words.

But with new microprocessors, said Fraser, "we are steadily increasing the vocabulary size our machines can handle. By the year 2000, we will probably be able to provide an affordable 20,000-word vocabulary speech recognizer." What's more, the machine will talk and listen to various languages, such as English, Japanese or Italian, he said.

Test models of a knowbot already are found in some research centers, says Greg Blonder, director of materials and technology integration research at Bell Laboratories. He says knowbots will reach consumers in five to 10 years.

Apple Computer is involved in several knowbot research projects. One, called the WAIS Project, is an effort to develop software that searches databases, returns with a list of articles and asks the person which ones are important. The program then goes off to find other articles similar to those selected. WAIS stands for Wide Area Information Server.

"Corporations are starting to gag on gigabytes of word-processing files, memos, reports, articles and E-mail archives," Brewster Kahle, WAIS project manager told Macweek, a trade publication. "Corporate memory is stored in this form, but executives have no way to get at it."

Another Apple experiment, called Rosebud, produces a daily newspaper. A person tells the program which subjects are interesting and the program each day gathers certain articles that are packaged like an electronic newspaper.

In 1988, Apple released a video demonstrating a possible future product called the "knowledge navigator." The video showed a young professor pushing a button on his desk. A notebook-size screen comes on, a "person" appears dressed in a bow tie, tells him about phone calls and appointments, and makes calls to arrange new appointments. The man tells the face to, say, look for facts about logging in the rain forest. The information is then displayed in sound, text and graphics.

But several challenges lie ahead.

Today's computers are limited in their ability to collect and manage data in all its forms, especially disk-hogging images. Personal computers need more powerful microprocessors to handle programs that use artificial intelligence and "fuzzy logic" to guess at what people want. Access charges to databases need to be cheaper. Copper telephone lines are inefficient for transmitting data compared with fiber-optic cable.

But Dottie Smith, a University of Washington librarian specializing in information stored on computers, says technology is not the biggest problem. The task is getting those who operate databases to agree on standards for accessing information. Giant commercial databases such as CompuServe have customers that may not want to change, she says.

Then there is the interface - the method by which humans and machine converse.

Bright Star Technologies in Bellevue, founded in 1987, has developed a patented technology of placing talking heads inside the computer. Often called an anthropomorphic agent, the "head" can be anyone - you, the boss, a friend, Zsa Zsa Gabor, a cartoon image. The image is created by videotaping speech patterns of a person and matching the mouth movements to recorded sounds. Bright Star has been used at information kiosks at the UW's annual Computer Fair.

"Within every human being is a deep-seated need to hear and see a human being talking to them," says Elon Gasper, founder and chairman of Bright Star.

The technique is far from perfect. The eyes jerk around and the 120 facial movements are more entertaining than realistic. Still, the result is powerful: The computer has a personality. That lump of silicon grins, frowns and smiles.

This talking-head approach makes it easier for users to work with what will be a powerful and complex machine, says Gasper.

"I don't see what else to use. The alternative is the disembodied voice," he says. "People don't want to talk to a voice. That's the reason why they used a voice in (the movie) `2001.' It makes it scary."

Every advance in the control of computers, from punch cards to the mouse, has been a step toward a more friendly, natural form of communication, he says. The next step is the talking head.

Gasper recently returned from a conference in Japan, where electronic giants NEC and Panasonic also are working on talking-head technology.

With 16 employees, Bright Star has released interFACE, which allows a product like Microsoft Word to be customized so a talking head pops up to give help and "hear" certain spoken commands.

Another Bright Star product, At Your Service, acts as a personal assistant who monitors schedules, delivers electronic mail and does other tasks. At Your Service can be customized so the assistant can be an English butler, a grouchy cartoon boss, a woman with a French accent or Phil, an absolutely inoffensive man.

Gasper said Apple Computer, a division of Motorola and Boeing have used Bright Star's Technology. Boeing used the talking head in a training program, he said. Bright Star is now developing products for IBM compatible computers running Multimedia PC, a standard pushed by Microsoft and others.

Bright Star's talking head also has been tried in a program developed by Interactive Design of Seattle for the Franklin Pierce School District in Tacoma. A friendly robot helps the students, monitors their progress and tests their knowledge.

Dave Cupp, director of special services for the district, says the talking head has been intriguing and entertaining. The students like it. After awhile, though, older students don't need help and switch off the head, he says.

Cupp is reluctant to predict the role of talking heads in the future, but he's "relatively positive" about them.

Others are less positive. They see a blurring of the line between people and machine.

Craig Ragland, director of development for Interactive Design, says the talking head can be useful in guiding people through information. But he worries that people might start thinking of their pleasant-faced machines as substitutes for genuine friends.

"If you can substitute pleasant interactions with fake people for depressing interactions with real people, that's attractive," says Ragland.

At a computer conference at Stanford University, several panelists objected to talking heads. One called for laws against their use.

Still, Gasper believes people will grow comfortable with talking heads. This is the flip side of controversy years ago, when there was fear that computers would dehumanize people, he says.

"I don't know if people want faces talking to them," said Blonder of Bell Laboratories. "There are some who will enjoy talking to an (anthropomorphic) agent and others who will be annoyed as hell."

With or without a face, the knowbot is coming.

Librarians need not worry about losing their jobs, says Dottie Smith.

"There are competing programs. Who's going to tell you which to use?" she asks.

Copyright (c) 1991 Seattle Times Company, All Rights Reserved.


Get home delivery today!