The Rochester Review, University of Rochester, Rochester, New York, USA
By Scott Hauser
When IBM's chess-playing computer, Deep Blue, beat world grand champion Garry Kasparov in 1997, even those who didn't know the difference between a byte and a bishop talked nicely to their computers the next day.
"IBM Chess Machine Beats Humanity's Champ," was the ominous headline in The New York Times.
Would a sci-fi future where humans are slaves to their robotic masters be far behind. . . ?
Not quite, says Paul Horn '73 (PhD), senior vice president and director of research at IBM.
Deep Blue was an important development, but not because it was "smarter" than humans, says Horn, who oversees the lab where the "inscrutable conqueror" (as another headline described it) was developed.
"When we look back at Deep Blue, I do think people will see it as a watershed," Horn says. "I think people will see the battles between Deep Blue and Garry Kasparov as the true start of the information age.
"The real lesson of Deep Blue, I think, is that it showed how important it is to work on really tough problems, on the 'grand challenge problems,'" Horn says. "And it showed how 'deep computing' can help solve those problems."
Since 1995, Horn's role at IBM has been to direct the work of nearly 3,000 scientists as they research, understand, and develop technologies that can solve those grand--and sometimes not so grand--challenges of the information age.
He joined the company in 1979 after a six-year stint as a professor of physics at the University of Chicago. Finding that he liked the opportunities to apply research as well as conduct basic research, Horn worked his way up through several corporate positions.
In 1993, he was named director of the Almaden Research Center in San Jose, California, the heart of Silicon Valley. As director of all IBM research, he currently oversees eight laboratories from his office at the division's headquarters in Yorktown Heights, New York.
Horn's ascent is no surprise to at least one Rochester professor who knew him in his grad school days. Physics professor David Douglass remembers Horn as one of the brightest students to pass through his department in the past 30 years and speaks of him in the same breath as such other luminaries as Nobel laureate Steven Chu '70 and Jack Crow '68 (PhD), director of the National High Magnetic Field Laboratory in Tallahassee, Florida.
"Very talented and ambitious," Douglass sums up.
Twenty-five years later, that talent and ambition would appear to have found a most compatible home at IBM.
Beyond all the Deep Blue stories, the company's successes seem to have been tumbling into the pages of the world's technology and financial press on a daily basis for the past few years, thanks in large part to its research crew.
In 1998, the company was granted a record 2,658 U.S. patents, 700 more than the number two contender. It was the sixth straight year that the number of IBM patents had increased.
And there have been some other head-turning announcements:
The company's new disk drive for storing images and data--the Microdrive--is scheduled to be introduced to the photo-taking public as a component of new cameras by the end of 1999. The drive can hold up to 340 megabytes of information--about 5 billion bits a square inch--and is expected to become part of an industry standard for storage in high-tech devices like digital cameras.
With the development of a copper-wired computer chip, IBM researchers are exploring new materials that will help keep Moore's Law on track. Named after Intel founder Gordon Moore, the axiom predicts that the speed and capacity of computer chips will double every 18 months. But many in the field think the physical limitations of traditional semiconductor technology, which uses silicon transistors connected by aluminum, is reaching its end. In late 1997, IBM unveiled a computer chip that uses copper to connect individual transistors on a single chip, a technology that will allow further advances in speed while also allowing for increased miniaturization.
IBM researchers also are extending the excitement of Moore's Law to an area of computer technology that has generally been considered a boring backwater: computer storage devices. Advances in the field are now coming faster than Moore's Law predicts for semiconductors, and at the front of the race is IBM. Early this year, the company announced that the latest version of its innovative computer disk drive--known as the giant magnetoresistive, or GMR, head--can hold a record-setting 20 billion bits of information--or about 2.5 gigabytes--in each square inch. (That's about half what a state-of-the-art desktop computer can store in its entire drive.) The device topped the 1997 record of 11.6 billion bits per square inch held by . . . IBM.
On another front, the company has signed a $650 million partnership with seed-developer Monsanto Corporation. The goal is to create a genetic sequencing program that could cut development time for new "smart" hybrids in half.
In other words, Big Blue is bubbling beyond the boundaries when it comes to the stereotypical research organization.
"We have outstanding people doing outstanding things" is how Horn explains it.
He emphasizes that a big part of the company's research and development success lies in providing an environment that allows its scientists to be creative. The other big part is that it gives them access to a breadth of other experts who can evaluate the resulting new ideas--and whenever possible channel them into new products or technologies.
Says Horn: "It's both managed and it's bubble-up."
At the heart of top-flight research and development are a couple paradoxes, he notes.
Paradox one: To get short-term impact on the company, you have to do long-term research.
Paradox two: In order to develop ideas that will later find an application in a specific area, you have to allow the ideas to flow freely across disciplines and company units.
"We fuzz the boundaries between very basic research, applied research, and development, and that allows us to have a conduit so that we can move products very rapidly into development," Horn says.
Take, for example, the GMR head, the ultra-small computer drive. The product's origins go back to the late 1980s, when two French academic scientists discovered a way to measure minute amounts of magnetism.
Neither the discoverers nor virtually any other scientist thought the new understanding of magnetism would be very useful.
Even though computers use magnetism to keep electronic 1's and 0's where they are supposed to be, the French discovery required very low temperatures and very high magnetic fields--and, as Horn says, "You can't freeze your computer every time you want to use it."
But a lone IBM researcher was intrigued by the idea and wouldn't let it go. After four years of tinkering--and a number of meetings with the French scientists--the IBM researcher had developed the first giant magnetoresistive drive. That early model has since been improved upon until the 1999 version can hold the equivalent of the text of about 2,500 contemporary novels spinning in the space of a quarter.
For Horn, the story exemplifies a successful research environment. "If we hadn't had a scientist working in that
particular field, we never would have known about the new development," he says. "And if our researcher hadn't had the freedom to let the idea bubble up over time, he never would have been able to carry it further."
As for Horn, he says he likes to be "on the boundaries" as a manager. He spends a lot of time in touch with academic institutions to understand what their scientists are working on and whether their ideas strike a chord with his own researchers. He also is on the lookout for talent. (The core team of scientists working on Deep Blue, for example, were graduate students at Carnegie Mellon University when they first began tinkering with the computer chess program.)
"We believe very strongly that it's important for industry to participate in the field at all levels, beginning with the academic," Horn says. "And 'participate' is an important word. It's not that we're sitting back and reading journal articles. That doesn't work very well."
One of the company's key priorities in the coming years will be taking advantage of Deep Blue's success in what IBM calls "deep computing." In contrast to personal computing, deep computing (named in homage to the chess champ) is the use of computer technology to solve the data-laden, information-saturated problems that only recently have been within the grasp of even the most advanced computers.
Who Is Deep Blue?Lost in all the hubbub of the "human versus machine" battle of Deep Blue and Garry Kasparov in 1997 was the fact that IBM never did say the chess-playing computer was "smarter" than its carbon-based counterpart.
"Computers and humans 'think' very differently," says Paul Horn '73 (PhD), senior vice president and director of research at IBM.
Indeed. Deep Blue, powered by an IBM RS/6000 SP computer, was capable of looking at an average of 200 million positions per second. The machine could easily look ahead six moves (or 12 plies, as a player's move and the opponent's countermove are known in chess), and thanks to some fine-tuning in its software, had the ability, in certain exchanges, to evaluate up to 30 moves ahead.
In short, the machine imitated what the very best human grandmasters can do, and could do it considerably faster.
Kasparov was able to stay competitive (he eventually lost 3.5 games to 2.5 games) because he was "outhinking" the computer in other ways, such as looking for weaknesses in its style of play, or taking advantage of spontaneity.
"In the right environment, computers can do things that are kind of like intelligence, but they don't operate the same way human intelligence does," Horn emphasizes.
The distinction often disappeared.
"People anthropomorphized the computer," he says. "They talked about how tall it was and how wide it was and that it was six years old.
"They talked about it as a 'he' or a 'she' rather than an 'it.' It's an 'it.'"
In its matches against Kasparov, Deep Blue was capable of looking at an average of 200 million positions per second, or about 20 billion moves in the three minutes allotted during a tournament. While the computer also has some chess- specific enhancements as part of its software, its speed and processing power--which easily outgun what a human could do--are on the order needed for future generations of computer problems: brain- busting tasks like forecasting weather to the point of specifying whether it's going to rain on Minneapolis but not St. Paul, designing and flying a new airplane without leaving a workstation, testing nuclear weapons without a single explosion, mining data from the overwhelming amount of marketing information available in the digital age, designing new drugs, and sequencing genes.
"These are the very big problems that you could not think of solving with a computer even a few years ago," Horn says.
But that is just what IBM is now thinking of: This spring the company established the Deep Computing Institute, a $29 million research effort to bring experts in academia and industry together to try to tackle some of the world's most challenging business and scientific problems.
Deep computing is at the heart of IBM's partnership with Monsanto.
It was also used by the designers who developed the new Boeing 777 on computers. Airplane manufacturers no longer have to build a prototype, see if it works, tinker with it, build another prototype. And then repeat the process.
The computer can be the drawing board, and then both plane and pilot on the test flight.
"One of the real benefits of deep computing is that you can do things on a computer that you had to do physically in the past," Horn says.
The other lesson from Deep Blue, he says, is that scientists can't be afraid to try to solve the problems that are conventionally described as unsolvable.
"It's important to work on the really tough problems," he says. "There's a huge value in working on the grand challenges.
"It's good for IBM, but it's also good for society as a whole."
Scott Hauser wrote the profile of film actor Robert Forster '64 in the last issue of Rochester Review.
Copyright 1999, University of Rochester