Originally appearing Jan. 15, 1998.
IBM labs develop new mark of distinction
YORKTOWN HEIGHTS, N.Y. - Tall, thin and dressed all in black, Mark Lucente looks more like a beat poet than a scientist. Two years ago, he left a research position at the Massachusetts Institute of Technology's Media Lab, joined IBM's research division and, he says, put all his money into IBM stock.
In the cafeteria at the curvacious T.J. Watson Research Center, Lucente talks about his work on imaging technology, saying that someday, every surface - the lunch table, desks, walls - will be a computer display. When will this happen? "In 10 years, but it's getting scary," he says. "It could only be a few years. But you don't want to say that too loudly or a manager will say, 'Send it over to the product group!' "
Lucente is only half kidding. IBM's research labs, one of the crown jewels of American science, has turned itself into an engine of IBM's business growth. It has pushed discoveries into the marketplace like never before, helping IBM create from scratch a $3 billion disk drive unit, a new business in speech-recognition software and a much-heralded way to build faster computer chips by using copper wiring instead of the usual aluminum wiring.
Along the way, the labs have helped IBM become the undisputed world champ of patents. This week, when the annual patent figures came out, IBM was tops for the fifth year in a row, with 1,742, according to IFI/Plenum Data. Canon was second with 1,381, followed by NEC with 1,101 and Motorola with 1,065. IBM has kept its patent lead despite budget and staff cuts. In 1992, the labs employed 3,056; today 2,785.
The lab hasn't always been so productive. It used to be a pure research lab, its output measured more in scientific papers and Nobel prizes than dollars and products. The approach won IBM Research a place at the top tier of science institutions with the likes of Bell Labs and major universities, but it didn't always help IBM's business.
When IBM was adrift in the late 1980s and early 1990s, the lab increasingly worked on products, but not the right ones. Then, in 1993, with IBM in serious trouble, Louis Gerstner was named CEO. He asked the labs - the main one is here in upstate New York, but IBM has seven others around the globe - to become a strategic asset, instead of an appendage.
"We all buckled down and said we will help save the company," says Frances Allen, an IBM scientist since 1957. Adds Caroline Kovac, a vice president at IBM Research, "We decided we have to do more than be smart minds for hire."
So the labs began to change. They aimed for a balance between long-term research and short-term innovation; between isolated deep-thinking and hands-on work with customers. Researchers take their inventions much sooner to IBM's product divisions. While the change irritated some, it invigorated many others.
Take David Nahamoo and Salim Roukos, who work on technology to help computers understand human speech. Their work is built on long-term research, going back to days when they'd need a supercomputer to understand a few basic spoken commands. The commands were always spoken by the researchers. They weren't learning how customers might talk.
By 1995, desktop personal computers were becoming powerful enough to handle speech recognition. The researchers teamed with IBM's health business unit and aimed a product at radiologists, who need to dictate tons of information about every patient - usually into a tape recorder, which means the tape has to be transcribed. The learning curve zoomed for the researchers as they analyzed data from having 800 radiologists from around the country talk into a computer. They learned how to handle accents and vocabulary and found that radiologists speak at a blurring 250 words a minute. "In less than a year, we had a product in the marketplace," Nahamoo says. Now they've got a lot more, most notably a whole line of ViaVoice software, which has made IBM a leader in voice-recognition for PCs.
One of the big successes out of the labs has been IBM's disk drive business, based on breakthroughs in magneto-resistive heads. In 1993, IBM had no PC disk drive business. In 1997, the unit had $2.9 billion in revenue. "Now we have 40%-plus of the high-density disk drive market," says Paul Horn, who runs IBM Research. "We're winning prizes and also having a major impact on our business."
Wall Street has even praised the labs, which is unusual since labs are about the long term and Wall Street is about the immediate. "The company has shown its ability to move new technology out of the lab and into successful products," says a Salomon Smith Barney research report on IBM.
A coming major impact might be IBM Research's much-publicized copper chip discovery. Computer chips always have had to use aluminum wiring because copper doesn't work well with the silicon in chips. But if chips could use copper, the chips would be better and faster. After 15 years of basic research, IBM found a way to make copper work with silicon. "It was a long-term understanding of metals that did it," says Randy Isaac, head of basic research. The discovery will give IBM a head start on next-generation chips.
The culture shift hasn't been easy, but it's imbedded at IBM Research. About 25% of the research division's employees now work directly with customers, Horn says. Rewards are based on contributions to science and business. And the personnel has changed. "We lost some and gained others," Horn says. For IBM Research job candidates in the past, "the decision was IBM, Bell Labs or a university," he says. "Now our competition is little start-ups and McKinsey (consulting firm)."
That difference certainly isn't lost on Lucente. "At MIT, you'd put out an idea and it would go pshhht, and result in more papers," he says. "I'm young and impatient. I want to make stuff and get a million people using it."
E-mail Kevin Maney at kmaney@usatoday.com and include name, address and day phone.
View online extract or purchase article.
