Monday, November 17, 1997
IBM Plans a Magic Show For the Folks at Comdex
Staff Reporter of THE WALL STREET JOURNAL
LAS VEGAS -- It has taken Mark Lucente two years of coaxing, but the researcher from International Business Machines Corp. has finally gotten a computer to be at his beck and call.
So this week, in keeping with this city's showtown quality, IBM and Mr. Lucente plan to show off a little spectacle of their own at the annual Comdex computer show. Billed as "human computing," it is a prototype system that responds to visual and voice cues.
It works like this: Standing in front of a wall-sized screen that essentially acts as a giant computer monitor, Mr. Lucente says "give me the world." Up on the screen pops a globe. "Make it this big," he commands, bringing his hands, palms outstretched, closer. The globe shrinks by half. "Move it over there," he says, pointing to the left corner of the screen. The globe moves there.
For several more minutes, Mr. Lucente commands all sorts of colorful shapes to magically appear, move, shrink, spin, change colors and disappear. There is no keyboard or mouse in sight, nor are there wires. Mr. Lucente simply sits on a sofa or walks about, talking and gesturing to the wall.
Visibility Is the Thing
By giving such a "gee whiz" demonstration, IBM is clearly hoping to grab
the attention of the more than 225,000 attendees here who will be assailed
by everything from a kangaroo and a talking, Internet-surfing Chevy Blazer
to a simulated hunt for a T-Rex and pop singer Chaka Khan.
Yet while the demonstration might seem out of "Star Trek," what's unusual about it is that it relies almost entirely on widely available technology -- not all of which IBM developed.
Mr. Lucente is using a version of IBM's $59 ViaVoice software that lets users talk to their personal computers. A wireless microphone picks up his voice and a video camera tracks his movements. This information is sent to "machine vision" software that is commonly used in factories to visually identify defective products coming off a conveyor belt. A powerful IBM NetFinity PC server crunches the data and a rear-projection system displays the images on the screen. The total cost: under $20,000, and that could be reduced if Mr. Lucente passed on using the expensive, giant screen.
Mr. Lucente's work, which IBM Chairman Louis V. Gerstner praised as "astounding" and "terrific" on a recent visit to IBM's research labs, is also a reflection of Big Blue's growing desire to be a "solutions" provider, a middleman developing applications that don't necessarily have to all end up in one IBM box. The Armonk, N.Y., computer company is betting that Mr. Lucente's work will eventually find its way into everything from interactive store windows and video arcades to virtual-theme parks and classrooms.
Ripening Technologies
"A whole bunch of different technologies have ripened," says Mr. Lucente,
who did some of the work at Massachusetts Institute of Technology's
Media Lab before joining IBM. "The real issue is who will be the first to
integrate all of this and get it into the market."
To be sure, other companies are promoting some of the same technologies that Mr. Lucente is talking about, especially speech recognition. Microsoft Corp. Chairman Bill Gates often speaks of talking to computers and the software giant plans to include speech recognition in future Windows operating systems.
Yet IBM has added some of its own twists. The key to its human computing setup are the complex algorithms -- or mathematical formulas -- that Mr. Lucente wrote that combine voice and gesture. The software takes his image and digitizes it, assigning a different color code to parts of his body. That way, the computer knows it is Mr. Lucente's hand (coded brown) that is pointing to a corner, where he wants to put a shape, and not his head or leg. The process operates somewhat on the same principle that enables a computer to recognize where a mouse is pointing to on a computer screen.
"The hard part is combining information in an intelligent way," says the 33-year-old researcher, who likes to reserve his fingers for playing his acoustic guitar rather than typing at a computer keyboard. "When I point to a corner it is a vague spot for the computer, not like a mouse pointing to a pixel."
Despite two years of work, his computer can only perform rudimentary tasks and is easily confused. When a reporter wearing a brown sports jacket tried to move objects around by hand gestures, the system didn't respond well because it kept confusing the brown color-coded arms with the rest of the body. Mr. Lucente concedes it might take several years to refine the technology.
But eventually, he says, human computing will make computers accessible to everyone, by eliminating bulky user manuals or the need to understand how Windows software works. "If computers are so great, they should be adapting to humans and not the other way around," he says. "After all, real life doesn't come with a 'minimize Windows' button."