PG NewsPG delivery
Pittsburgh Post-Gazette Home Page
PG News: Nation and World, Region and State, Neighborhoods, Business, Sports, Health and Science, Magazine, Forum
Sports: Headlines, Steelers, Pirates, Penguins, Collegiate, Scholastic
Lifestyle: Columnists, Food, Homes, Restaurants, Gardening, Travel, SEEN, Consumer, Pets
Arts and Entertainment: Movies, TV, Music, Books, Crossword, Lottery
Photo Journal: Post-Gazette photos
AP Wire: News and sports from the Associated Press
Business: Business: Business and Technology News, Personal Business, Consumer, Interact, Stock Quotes, PG Benchmarks, PG on Wheels
Classifieds: Jobs, Real Estate, Automotive, Celebrations and other Post-Gazette Classifieds
Web Extras: Marketplace, Bridal, Headlines by Email, Postcards
Weather: AccuWeather Forecast, Conditions, National Weather, Almanac
Health & Science: Health, Science and Environment
Search: Search post-gazette.com by keyword or date
PG Store: Pittsburgh Post-Gazette merchandise
PG Delivery: Home Delivery, Back Copies, Mail Subscriptions

Headlines by E-mail

Headlines Region & State Neighborhoods Business
Sports Health & Science Magazine Forum

CMU's Raj Reddy fills lives with big questions

Monday, June 15, 1998

By Michael Newman, Post-Gazette Staff Writer

Raj Reddy, who will soon step down as dean of the School of Computer Science at Carnegie Mellon University, is a man of few interests. It's just that his chief interest happens to be, well, the nature and meaning of intelligence.

Raj Reddy, dean of the School of Computer Science at Carnegie Mellon University. (Tony Tye, Post-Gazette)

One of his projects involves placing "the sum of all human knowledge" on the World Wide Web. "It's what I intend to spend the rest of my life working on," says Reddy, 61.

Another project, on which he is working with the federal government, is the creation of a nationwide, superfast computer network. "That will take a lifetime," he says.

He's also working on a computer system to recognize and understand the spoken word as quickly and accurately as humans. Not to be redundant, but - how long will that take? "It's something I'll spend the rest of my life on," he says.

In between these lifetime projects, Reddy has helped found a few companies, advised a few more and inspired countless others. And he has wheedled scores of millions of dollars, from governments and corporations, to fund it all.

Tom Murrin first met Reddy in the late 1970s, when Reddy was raising money for what would become Carnegie Mellon's Robotics Institute. Murrin, now dean of Duquesne University's business school, was then an executive at Westinghouse Electric Corp.

After their meeting, Murrin says, he told Reddy: "You are either one of the most captivating charlatans I've ever met, or you're one of the most knowledgeable minds I've ever been privileged to know."

Apparently he decided in Reddy's favor. In 1980, thanks in part to funding from Westinghouse, the Robotics Institute was established.


Raj's past

Reddy, who has been at Carnegie Mellon for almost 30 years, wasn't always so interested in intelligence. He received his bachelor's degree in civil engineering in 1958 from the University of Madras, in India. He then traveled to Australia as an exchange student.

After receiving his master's degree in 1960 from the University of New South Wales in Sydney, Reddy worked for IBM in Australia.

"He was a tremendous computer salesman," says David Simonds, the chief executive officer of Pittsburgh's RedZone Robotics, who's known Reddy for two decades. "Now he's a tremendous computer-science salesman."

It was at Stanford University, the birthplace of Silicon Valley, where Reddy received his doctorate in computer science. He was one of two graduates in the department's first class, in 1966.

After teaching a while at Stanford, Reddy came to Carnegie Mellon in 1969. Like so many others at the university, he was attracted by the pioneering work of Nobel laureate Herbert Simon in the field of artificial intelligence. (Reddy now holds an endowed chair named for Simon.)

In his three-decade career at Carnegie Mellon, Reddy has held a number of positions within the university and outside it. He was founding director of the Robotics Institute, a position he relinquished in 1991 when he became dean of the School of Computer Science.

"People don't remember what a sad group of people we were six or seven years ago," says Jim Morris, head of Carnegie Mellon's computer-science department. Alan Newell, who along with Simon had pioneered the field of artificial intelligence, had recently died, and several other prominent rsearchers "had gone on to greener pastures."

That's when Reddy stepped in. Since becoming dean, Reddy has attracted several talented professors, helped win funding for dozens of research projects and continued with his own research. In 1994, he received the Turing Award, the highest honor in computer science.

The Turing Award comes with a $25,000 prize and a sterling silver bowl, with which awardees customarily pose before adding it to their trophy cases. Reddy's bowl sits on a desk in his office. He keeps candy in it.


Raj's ideas

The one idea that has animated Reddy through his career is intelligence: what it is, how it's used, whether it can be learned.

His answers, in short, are we're not sure, we're still learning and yes, absolutely.

Reddy's first research interest was in speech recognition - getting a computer to understand and respond to the spoken word. It seems a fairly straightforward problem, but it has led Reddy, and much of Carnegie Mellon's computer science school, to such diverse areas as robotics, computer vision and the World Wide Web.

For robots to be truly useful, for instance, they must learn to "think" like humans - to respond to unexpected situations, adapt to changing environments, and "remember" past experiences. All these anthropomorphic actions require a degree of intelligence, which is called artificial merely to distinguish it from the human kind. (Simon, for one, would argue that there's nothing "artificial" about it.)

But we are just beginning to understand how a human processes information, never mind teaching a machine how to do it.

And teaching a computer to play chess, as it turns out, is considerably easier than teaching it to chat about, say, the weather. IBM 's Deep Blue computer, which learned much of its strategy at Carnegie Mellon, beat world chess champion Garry Kasparov a year ago. But no computer has yet passed the Turing Test - conversed with a human so naturally that it's not even recognized as a computer.

"When we say, 'Hi, how are you,' we don't think of that as intelligence," says Reddy. "But that's actually the hard part." Such intelligence involves the simple yet complex task of perception, which is a difficult concept to define, much less emulate.

Part of how Reddy has dealt with this issue is through what he calls an "80/20 rule."

Take the example of an "intelligent" computer calling to buy theater tickets. It already knows its owner's preferences (musicals, aisle seats, matinees) and it can easily convert this goal -- buying a ticket -- into smaller, more easily attainable goals: looking up the number, dialing it, requesting information from the theater, transmitting a credit-card number to buy.

But what happens if aisle seats are sold out? Or if the show itself is sold out, and the ticket agent suggests a Shakespearean tragedy? Or if the theater burned down and its number now belongs to a phony religious cult that pretends it's selling theater tickets but actually uses callers' credit cards to buy real estate in the Caribbean?

All of these situations would create problems for a computer, and probably will for the foreseeable future. Enter Reddy's 80/20 rule: Make the computer perform 80 percent of the task -- calling the theater, finding out what seats and times are available and getting prices -- and leave the other 20 percent to the human.

The rule also has the advantage of speeding things up. If a problem is going to take 20 years to solve, Reddy says, probably about 80 percent of it can be solved in 10 years. The remaining 20 percent would take another 10 years to solve. Reddy figures it's better to spend that time solving 80 percent of some other problem.

"The idea is that man and machine work together interactively," he says.


Raj's influence

Perhaps even more than his ideas, Reddy is known for his influence. He has been enormously successful at attracting talented scientists and large research projects to Carnegie Mellon.

At a campus symposium 10 days ago to honor his 60th birthday, dozens of professors, researchers and businessmen, from Carnegie Mellon and elsewhere, testified to his importance. Among his most effective tools of persuasion is something called the "full Raj" -- not to be confused with the mere "half Raj," a tactic used for less critical tasks.

"A 'half Raj' is just an arm on your shoulder," Morris says. "In a 'full Raj,' he brings you into his neck. He's not shy about asking people to do things."

Stories of Reddy's many enthusiasms abound. Chuck Thorpe, a principal research scientist and one of the founders of K2T, a robotics firm in Duquesne, tells of how Reddy would often wander into his office and suggest some new path for his research.

"It was always a great idea," Thorpe says. "But then you say to yourself, 'I already have a Ph.D thesis."'

David McKeown, another research scientist, came to Carnegie Mellon in 1975 to work in the field of computer vision -- programming a computer to "see," to recognize and understand its environment.

"Computer-vision research at CMU was very different from (research) elsewhere," he says. It was interdisciplinary, with researchers from other university departments, and it was investigating areas of the field that others were not. ("Optimal edge extraction" was one of them, if anyone asks.)

"Raj's instincts have been impressive," McKeown says.

Among the projects Reddy is either involved with or has inspired are Informedia, which uses technology to search video, speech and text; the Universal Library, which aims to place and coordinate "all the world's knowledge" on the Web; and the intelligent-vehicle project, which allows humans to rest or read while a computer operates their car.

"My entire project is about a Raj Reddy vision," says Michael Shamos, director of the Universal Library. "He told me it was the project he was going to spend the rest of his life on. Of course, I soon learned that there are five or six other such projects."

And Reddy was one of the founding directors of the Robotics Institute, which has so outgrown its quarters that it is almost doubling in size, adding two stories to its building on campus. The spacious new quarters have been dubbed "Raj Mahal."

Beyond Carnegie Mellon, Reddy's impact is felt at the companies he helped found or consults for.

Ravi Koka, president and chief executive officer of Seec, the Findlay-based maker of software for mainframe computers, first met Reddy when Reddy was on sabbatical in India in 1975. He would come into the offices of Koka's company in Bangalore and spend hours talking and helping with research.

A dozen years later, when Koka was ready to start Seec, Reddy introduced him to investors and invested some money himself. He's now chairman of the board.

Reddy has proved helpful in negotiations with the federal government, one of Seec's big customers. He's also available to speak at technical conferences. But most important is his strategic vision, Koka says. "If we're having a serious technical discussion about the future, we need his input," he says.


Raj's future

Next to intelligence, the future is Reddy's favorite topic. When he announced last winter he was stepping down as dean, he said it was so he could spend more time teaching and doing research, which he hopes to be able to do by the middle of next year.

He has few hobbies -- he sometimes takes walks and he reads the occasional book -- and his two grown daughters now live on the West Coast, leaving his Oakland home to himself and his wife. His seven brothers and sister live near Bangalore, where he returns about once a year.

Of his many projects, none of which he realistically expects to finish, he is philosophical. "They're all interrelated," he says. Like many problems in artificial intelligence, he says, "in order to solve any one of them, you have to solve all of them."

Will a computer be able one day soon to understand when you order a "hoagie with everything, hold the fries"? Will it respond when you say, "Let me hear Mahler's Fourth Symphony"? Will it drive you safely to your job every morning, and home every night?

It depends. "There are many things, while they may not be themselves worthwhile, you learn so much getting there," Reddy says. "Society has to find a way to fund crazy, way-out ideas. We know that many discoveries have happened through accidents."

Just ask James Gosling, one of the inventors of the Java programming language, who credits Reddy with inspiring him while he was a student at Carnegie Mellon.

"Anyone who thinks they can predict the future is just nuts," he says. "Because these things just rattle around like ping-pong balls with deranged mice inside them."

Reddy may not put it quite that way. But he certainly agrees with the sentiments behind it.



bottom navigation bar Terms of Use  Privacy Policy