Starting from scratch
Ruby Lee draws on experience in industry and academia to rethink computer design
Princeton NJ -- Ruby Lee wants a new computer.
She's not thinking about a new machine for her office in the electrical engineering department. She wants to entirely rethink the way computers are designed.
"I am interested in designing the computer processor architecture for the 21st century," said Lee. "What would that processor look like if we could design it from scratch?"
It may sound like a tall order, but in some ways she has done it before. As a chief computer architect at Hewlett-Packard, she was a key figure in a revolution in computer architecture that swept through the industry in the 1980s. With other pioneers, she advocated the use of a vastly simplified set of core instructions that computers use in carrying out all the complex things they can be programmed to do. This simplified system is now at the heart of tens of billions of dollars worth of computer systems sold each year.
"She's really been one of the top instruction-set architects in the world," said Joel Birnbaum, the former senior vice president at Hewlett-Packard who hired Lee in 1981. Before she left the company to come to Princeton in 1998, Lee was leading a team that collaborated with chipmaker Intel Corp. to design a new architecture, which was recently released in the new "Itanium" microprocessor chips and which Birnbaum said will remake the industry once again.
"Ruby has been there for two major revolutions in architecture," he said. "That's quite unusual for a computer architect."
But that's not to mention a third innovation that Lee brought to the industry. She also led the way in creating a set of multimedia instructions that build directly into the core native language of the computer the ability to handle multimedia of all types -- images, voice, animation. Intel later adopted the idea, which consumers saw when the company began advertising that its chips had multimedia-accelerating "MMX" technology. Now essentially all computers use the technique, allowing multimedia to be an everyday part of computing.
Princeton professor of computer science Kai Li called Lee's work in this area a "seminal contribution" that reflected her ability to predict very early on how important graphics and multimedia would become.
It also is impressive, said Princeton electrical engineer Sun-Yuan Kung, that Lee's solution was compatible with all the existing technology. "That was really a smart move and was a great contribution," he said.
Having reached this point, however, Lee is ready to start over.
Moving from industry to academia allows her to "start from a clean slate," she said. "In industry you can rarely start from a clean slate," she said, noting that successfully entrenched products must always build on the previous versions for compatibility reasons. Now that she is here, she said, "I find that freedom exhilarating."
In her re-examination, one of the first things she found lacking is the way computers handle security. She noted that in the early days of computers, security amounted to putting the machines in a locked room. Now that computers are connected all the time to networks, including the global network of the Internet, security is a much more difficult and crucial issue.
Current security tools, such as virus-checkers, computer "firewalls" and encryption techniques, are a patchwork that can be difficult to use and are far from comprehensive. Lee wants to make security part of the basic architecture of the computer, just as she did with multimedia.
"We need security that is implicit and invisible," she said, "not security that is a negative impact on our quality of life -- like having to line up for a long time at the airport to go through checks. The real security is security that is built in, not added on as an after-thought. For example, the core architecture should protect people from misuse of their interconnected information appliances by third parties. Security encompasses a broad range of societal needs like privacy, anonymity, user authentication, service availability, confidential transactions and intellectual property protection."
On the multimedia front, Lee is going back to the drawing board to develop what she calls the "canonical, minimalist instruction-set architecture for multimedia processing that is fully general purpose, high-performance, low-cost, low-power, so it can be implemented in the smallest information appliance, like a Dick Tracy watch."
In both of these pursuits, Lee's approach is to develop technology that does more than just run fast. "I think about how you can make a processor more effective, rather than just more efficient," she said.
Rising above the fray
The goals may sound lofty, but it would be a mistake to assume that Lee is no longer interested in whether her ideas will win over the market. "I can't get away completely from the idea of my research being important to the future of the computing industry," she said.
Indeed Lee already has taken cuts at the problems and has produced specific technical advances. In her latest innovation, she has discovered a way to reduce by 100-fold the number of operations a computer must execute to scramble a series of bits, the smallest units of information in a computer.
Current computers are best at processing word-sized chunks of information, where each word is typically 64 bits. This design, however, is inefficient at manipulating bits within a word, which can be valuable for encrypting data such as credit card information sent over the Internet. A typical computer might have to execute hundreds of instructions to arrive at a specific arrangement, or permutation, of a 64-bit unit.
Lee and her students recently published papers showing how computers of the future could be modified to perform 64-bit permutations in, at most, six instructions, while keeping all the benefits of word-oriented processors. And since submitting that paper, she has further reduced it to just one step, she said.
While Lee is looking for collaborators at Princeton to further the theory and use of permutation, she also is keeping an eye on the bigger picture. She is trying to understand the philosophical and sociological aspects of security, including issues such as balancing copyright protection with fair use and academic freedom. "What I would like to do is start a conversation, first at Princeton, then nationally and internationally, to answer the very basic questions of who should be allowed to access what type of information and when. Then the technologists can respond with what can or cannot be achieved, or what else can be achieved."
This ability to balance practical solutions and visionary thinking is uncommon in the field, according to Lee's colleagues.
"Ruby was able to rise above the fray and make more pervasive contributions," said Birnbaum, noting that Lee always was able to deliver in the "blood and guts, schedule-driven product development" of Silicon Valley.
Returning to academics
After earning her 1980 Ph.D. from Stanford University, Lee stayed on for 15 months as an assistant professor until Birnbaum recruited her to Hewlett-Packard. During her years in industry, Lee kept in touch with her academic counterparts and maintained a consulting professor relationship with Stanford.
Nonetheless, Lee encountered some surprises when she returned full time to academia. One was learning about the extent of the inequities that women still face in rising through the science and engineering fields. Shortly after she arrived, former president Harold Shapiro invited her to a conference on the subject at the Massachusetts Institute of Technology. Now she is a member of a task force on gender equity in the natural sciences and engineering established by President Tilghman.
"I thought there wasn't much of a problem in Silicon Valley -- at least I never felt much of an impact," she said. "But I have seen enough data to convince me that there is a real problem with the biases women face, and hence in attracting and keeping women in academic science and engineering fields. This is a problem that seems incongruous with the 21st century -- we must find creative and equitable solutions."
Lee is excited about another aspect of being at the University: the opportunity to interact with students.
"The greatest challenge in teaching is to give these very bright students the most important concepts that we have learned over the years, and yet train them to think out of the box, because we don't want the next generation of computer architects designing more of the same," she said.
She also has enjoyed recruiting top graduate and undergraduate research students to her quest for new and better ways of designing computers.
"The students are fearless because they don't know what can't be done," she said.
Perhaps one of them will build her a new computer.
Deadline. In general, the copy deadline for each issue is the Friday 10 days in advance of the Monday cover date. The deadline for the Bulletin that covers April 15&endash;21 is Friday, April 5. A complete publication schedule is available at deadlines or by calling (609) 258-3601.
Editor: Ruth Stevens