COMPUTER SCIENCE & ENGINEERING AT THE CUTTING EDGE

Mix Of Top Companies And Labs Make The Northwest An R&D Powerhouse

By Ben Raker

Imagine a future where computers are everywhere–not just on every desk or in every lap, but also contained in miniaturized form in almost every object around you. Increasingly powerful networks bring great leaps forward in computation. Computers catalyze other sciences also, as the power to sift and systematize vast quantities of information races ahead.

Now consider that computer systems, networks of computers, and databases will continue to become more complex as they grow smarter, more prevalent, and more powerful. There will be privacy issues as computers become embedded in increasingly everyday items. There will also be new challenges to the security and reliability of networks as connections spring up everywhere. Sophisticated hardware and software will be needed to make sense of the explosion of information and make it possible for people to visualize information.

These revolutions are, of course, already under way. And Northwest researchers are hard at work in corporate, academic, and government labs researching technologies for this coming future–in a dizzying array of research areas.

Evidence of a vital research presence abounds. Recent news tells of spin-off companies and big new players being enticed to the area: Intel created a lab in Seattle in 2001; Google established a Kirkland, Wash. office in 2004; the founder of Linux also set up camp in Portland in 2004. Research by homegrown icons and longtime industry presences, such as Microsoft in Washington and Intel and Tektronix in Oregon, continues to thrive.

Meanwhile, academic departments are winning awards right and left, building new facilities, and adding faculty. They are forging interdisciplinary links to capitalize on other Northwest strong suits, such as life sciences and biotechnology. And the federal government is making its own efforts in research in the region–evidenced by the restructuring of the Pacific Northwest National Laboratory in October 2004 to make computational science one of its five core directorates.

Yet, while the region continues to celebrate its strength in computer science and engineering research, some express caution that the region must not become complacent. Academic and industry leaders warn that state and federal funding for university research lags dangerously, and that universities, which form the backbone of technology centers worldwide, may be eroded by a lack of backing. Most agree that, despite existing momentum, the fast-paced, competitive field of computing will require a continued investment in research and innovation to maintain the region's edge for the future.

Industry research: Microsoft, Intel, and beyond

While corporate computer science is best known for developing new products, the Northwest benefits from local giants that are unusually committed to research beyond the next product cycle.

At the center of this is Microsoft Research (MSR), based in Redmond, Wash. MSR, which was established in 1991, is the branch of Microsoft tasked with exploring the future of computing and inventing new technologies for the long term–five to 10 years beyond current product cycles, according to a company press release.

"We have to constantly innovate in order to give our customers reasons to want to continue to buy things from us," says Kevin Schofield, MSR's general manager based in Redmond. "Microsoft really wants to be a leader in bringing great new technologies to the world. And that's why we invest as much as we do in computer science."

MSR collaborates with the product development groups at Microsoft and can transfer ideas and technologies for new products to those groups. But it is also its own distinct, centrally funded entity. "We talk to all the product groups, but we don't do a particular piece of research saying, this will end up in Windows, this will end up in Office, or Xbox or something like that," says Schofield.

Although product development groups form a far greater portion of the company, the scale of MSR is still significant. With the opening of a new lab in India in January 2005, six labs now make up the MSR network. It has about 700 employees worldwide, with the majority of work taking place in Redmond. MSR engineers work in 50 to 55 areas of computer science at any one time.

"It's a strategic bet from the company that we can build a war chest, if you will, of strategic technology assets and expertise," says Schofield. For example, when Microsoft developers recently sought to ramp up their search technologies, MSR had expertise working in this field for 10 years.

Schofield categorizes the areas of highest interest to MSR in three big "buckets": improving user interfaces, making computers smarter, and making computers more trustworthy.

Within the first category, user interfaces, one area that has seen recent success is the improvement of graphics and animation. Schofield cites work by senior researcher Hugues Hoppe, an alumnus of the University of Washington (UW) Department of Computer Science & Engineering. Hoppe, who won the 2004 Computer Graphics Achievement Award at the prestigious computer graphics conference SIGGRAPH, has done pioneering work in the use of real photographic data to make models for creating new photorealistic images.

Graphics work is important to user interfaces because of the importance of vision to understanding. "Our eyes are our highest bandwidth way of getting data into our brains," says Schofield.

Meanwhile, another MSR group, called VIBE (for "visualization and interaction for business and entertainment"), is exploring what is happening to the visual interface as people work with multiple or differently sized displays. Exploring small screens is important, as the popularity of cell phones and handheld devices continues to increase. At the same time, there is also potential for consumers to begin using bigger screens and more of them, as display technologies are becoming less expensive. Mary Czerwinski, who heads the VIBE group and has a background in cognitive psychology, recently found that adding multiple displays immediately increases productivity among users. MSR is exploring the implications of changing screen sizes and whether changes to user interfaces could enhance productivity.

In the area of making computers smarter, MSR is exploring areas such as data mining, machine learning, and handwriting and speech recognition. Data mining is basically the ability to sift through huge, complex quantities of data and make sense of them. This has potential for many scientific fields since massive databases can conceal useful information that a smart computer might pick out. In one project with the University of Washington and Australia's Royal Perth Hospital, Microsoft researchers are using data mining and machine learning to find genetic patterns useful in developing vaccines against the HIV virus–a challenging problem because the virus is constantly mutating. The first vaccine designs are now undergoing laboratory testing.

Finally, in the category of "trustworthy" computing, MSR is exploring advances in the security of large, distributed systems, and making systems more reliable (less crash-prone). Within this area, MSR is looking at smoother ways to have networked machines pick up the load when one machine fails so that users never notice the failure. They are also developing tools to test new code written by Microsoft engineers to get bugs out early.

Trustworthy computing, says Schofield, is "in the end, all about how do we get people to really trust their computing devices as they become more ubiquitous in our lives."

This theme of ubiquitous computing is the major focus of another corporate research lab: Intel Research Seattle. Intel has had a huge presence in Northwest computing for many years, mostly through its giant chip manufacturing and research industry in Oregon. Approximately 10,000 people are employed by this industry in Oregon today. Now, in addition to hardware research, Intel has also opened several labs around the country to explore the future technologies that will use its chips.

"If we can come up with good reasons why people would have more computing in their future homes or future work lives, in the end that would be good for Intel," says James Landay, director of Intel Research Seattle.

The lab in Seattle uses an open intellectual property arrangement to encourage collaboration with UW researchers and to leverage the power of its 18 researchers in the ubiquitous computing arena.

"Collaboration is the whole point," says Ed Lazowska, who holds the Bill & Melinda Gates Chair in Computer Science and Engineering at the UW. "It's a real departure for corporate research."

Using this open model, the lab has springboarded to a top position–possibly the top position–in the field of ubiquitous computing. At the main conferences on this topic in 2004, UW and Intel researchers together had more high-ranked papers than any other research center.

The idea behind ubiquitous computing, or "ubicomp," is that computing in the future will take all forms and sizes. Computers will be embedded in everyday objects that can track users and assist them wherever they are. One nagging concern with these technologies is that privacy can be compromised by ubiquitous, smart gadgets, creating a sort of "Big Brother" scenario. To be useful, computers would need to be aware of where the user is and what he or she is doing.

To tackle the issue of defining a user's position without compromising that user's privacy, Intel Seattle has been working on a project called Place Lab. This makes use of the fact that transmitters of radio waves and WiFi stations each have a unique signal. With devices that listen for "beacons" with mapped locations, the user can determine position through a process similar to triangulation. The process is passive: "You don't have to communicate back with them at all; your device just listens," says Landay. "It's built from the ground up to be privacy aware."

Intel is addressing the question of what a subject is doing through a project called SHARP. This project makes use of the idea that all objects in the near future may carry a radio frequency identification (RFID) tag. Such tags are already being required of vendors by retailers like Wal-Mart because of their usefulness in tracking cartons of goods. RFID tags are similar to barcodes, but each tag has a unique ID.

Intel uses sensors to detect movements of RFIDs in order for computers to guess what a user is doing. For example, if a sensor catches a bowl moving, then a cereal box, then a milk carton, the system might infer the user is eating breakfast.

One application Intel is testing could one day help seniors remember to take their pills or allow a family member to check that an elder relative ate her meals for the day. This project is still in its early phases and both the technology and privacy concerns haven't all been sorted out yet, Landay says.

While Intel draws on strong faculty expertise at the UW to advance its projects, UW researchers also recognize the advantage of having industry labs like Intel's and Microsoft's near campus, allowing for different levels of collaboration.

These two companies, "are almost unique in spending any appreciable amount of money looking more than one product cycle out," says Lazowska, who, besides being a UW professor, has also been on MSR's technical advisory board since its inception. "What's incredibly cool is that Intel and Microsoft both have research labs here in Seattle."

But Microsoft and Intel aren't the only industry players. Other companies, though more proprietary, are also advancing computer science in the Northwest. Amazon.com is unmatched in the online retail industry, and develops original algorithms for its large-scale distributed systems, searches, and recommendations. It bases much of this work on studies of user interaction. "We just have a tremendous amount of volume that we deal with and we find ourselves very often on the cutting edge of what technology is able to do," says Larry Tesler, vice president of shopping experience. "Standard vendor solutions don't do it, so we have to come up with our own."

In Seattle, there is also Cray, Inc., one of the world leaders in designing supercomputers. There is Adobe, which is well known for graphics software. There are various spin-offs, like the Intel-UW startup Impinj, which is developing silicon RFID tags that be updated with information. Boeing, though best known for airplanes, has a computer science component and played a role in the early development of the Seattle computer industry. And, just arrived in 2004, there is a Google office in Kirkland, Wash., that will employ search engineers.

While Seattle dominates the region's software research, it isn't the only place in the Northwest where companies are interested in conducting computer science R&D.

"When you look at industry in Oregon, it tends to be more hardware and chip oriented, which is why you see Intel, Tektronix, that whole set of companies down there versus a little more of the software spin up here," says MSR's Schofield. The area hosting this industry in Oregon is often referred to as the Silicon Forest, an obvious pun on California's Silicon Valley.

One of the reasons the chip industry took off there, says Schofield, is the presence of strong academic programs. "You look at University of Oregon and Oregon State University and some of the programs they have there, and it's no wonder."

Schofield speaks for many when he says that computer science research at universities play a key role in stirring up tech economies. "Strong tech centers always grow and thrive around strong research universities," he says. "That's a belief that's not just sort of my own personal belief, but that's a belief of Microsoft Research and Microsoft as a whole."

Others voice the danger of putting too big a burden on corporate research, however great its contributions. Corporations must be largely focused on their business interests and, therefore, be concerned with product development.

"Microsoft and Intel are making major investments in fundamental research," says Lazowska, "but they are unusual among information technology companies, and even at Microsoft and Intel, research is a very small proportion, really just a few percent, of overall R&D." Lazowska and others argue that universities complement industry by focusing on diverse, unrestricted research projects that bring about long-term fundamental changes in the field.

Most in the field would agree that some mix of corporate and federally funded research is appropriate to building and maintaining critical mass for innovation.

"It's a contact sport," says Lazowska. "You have to be there bumping into each other."

Academic and government research: strength in variety, interdisciplinary applications

Throughout the region, one of the major trends in computer science research at universities is that projects vary widely. Another characteristic is that the projects form interdisciplinary links to reveal exciting applications for computer science.

The region's largest, best-funded, and most varied department is the UW's Computer Science & Engineering. It consistently ranks among the top computer science departments in the nation and turns out graduates that work at major companies and universities nationwide. It also creates spin-off companies and technologies used the world over. In autumn 2003, the department moved into its spacious new Paul G. Allen Center for Computer Science & Engineering, a facility to match the top-notch research underway inside [NWS&T, Autumn 2003].

Among active research areas at UW are: embedded systems, computer architecture, networks, operating systems and distributed systems, data management, graphics, artificial intelligence, and robotics.

UW researchers have worked with Microsoft and other industry partners to rise to the top of the field in graphics and animation [NWS&T, Spring 2004]. "The best nucleus of computer graphics research in the world is between UW and Microsoft Research," says Lazowska.

Another project with links to Microsoft Research is a program called Classroom Presenter, developed by UW professor Richard Anderson. The program, which Anderson began developing while on sabbatical at MSR, uses Tablet PCs to make class presentations more active. The typical PowerPoint lecture is static, as Anderson explains. By using the Tablet PC, teachers can add notes during the presentation by writing them on the device screen, which is immediately projected for the class. Anderson is now exploring what happens when students each have their own tablets with which to participate.

UW CSE, the program that launched the world's first successful full-text web search engine (Webcrawler) and the world's first successful web meta-search engine (MetaCrawler), pursues more traditional nuts-and-bolts computer science alongside applied and cross-disciplinary work. In one ongoing study, researchers seek to reduce failures causes by device drivers. As professor Hank Levy explains it, each time you plug in a new piece of hardware, it brings its own code to your system–which causes many of the system crashes we see today.

Levy, along with professor Brian Bershad and graduate student Mike Swift (who is writing his doctoral thesis on the topic), looked at ways first to isolate new drivers from the operating system, and secondly, to create pieces of code that could step in when a driver failed and effectively prevent the system from failing until the driver was restored. Previous research by Levy, along with faculty colleague Susan Eggers and graduate student Dean Tullsen (now on the faculty at UC San Diego) invented Simultaneous MultiThreading, a new computer architecture now commercialized by Intel as HyperThreading.

On the more interdisciplinary end of the spectrum, professor Rajesh Rao and researchers in the UW's Neural Systems Lab, along with neuroscientists and doctors in the UW Medical Center, are working to simulate parts of the human brain with computer algorithms. They are also working with robotics to see if machines can learn by imitating sequences of behavior. Finally, this group is attempting to decode brain signals to control devices. In December 2004, these researchers collaborated with a neurosurgeon at Seattle's Harborview Medical Center to help a disabled patient move a cursor on a screen just by thinking about it moving. The researchers used electrodes to map the patient's brain signals and linked these to the device control.

Computer scientists and neuroscientists are also collaborating at the University of Oregon (UO) in Eugene. They have collaborated on a Neuroinformatics Center–using the power of computers to sort out data sets about how the brain works.

The UO program, which emphasizes cross-disciplinary work, also works on assistive technologies for disabled populations. One project, called EyeDraw, assists severely disabled children who have no muscle group with enough control to move a mouse except their eyes. The researchers are working on ways for these children to perform the challenging movements associated with drawing on a computer.

Another UO project works on detailed functional and structural models of dinosaurs in a field they call cyberpaleontology. Bioinformatics is another focus at UO, as displayed in projects that range from gene evolution research to genomics. Yet another uses UO's strength in network-distributed computing to pull together a Web-based database of biologists studying zebrafish around the world. Distributed systems bring many computers together to work on one task.

"I think distributed and network computing throughout the state is big," says Michal Young, chair of UO's computer science department.

At Oregon State University (OSU) in Corvallis, collaborative research using distributed computers is also important. Networks help studies in various disciplines, bringing together engineers to simulate earthquakes and study tsunamis, for example. One program links the world to a large wave tank owned by OSU so that researchers can run experiments on tsunamis from afar [NWS&T,, Autumn 2003].

OSU has strengths in machine learning, artificial intelligence, system usability research, end-user software, animation graphics, and visualization. "We have a very significant industry in Oregon in what I would call image-oriented technology," says Terri Fiez, director of the school of electrical engineering and computer science.

Cynthia Brown, chair of the computer science department at Portland State University (PSU) in Portland, names network computing as a strength of her department. Other strengths include multimedia networks, network security, and adaptive systems (systems that learn). With 12 new faculty and new facilities for the engineering school of which it is a part, computer science at PSU is growing. It collaborates with hardware companies in the area such as Intel and Tektronix. Brown notes that companies are big on embedded computers in Oregon, and security-related work is another strength of the region.

Security is also a focus at Idaho institutions. At Idaho State University (ISU) in Pocatello, there are around 45 students actively involved in maintaining Web sites and materials for the State Department, Department of Homeland Security, and National Security Agency. ISU also has two researchers working on using computing and mathematical models for various types of pattern recognition.

The University of Idaho (UI) in Moscow partners with ISU in a national center of excellence in cybersecurity. The UI department also works in high-performance computing and applying computers to large biological data sets–bioinformatics [NWS&T, Spring 2000]. "The security area seems to have visibility here and I also think the bioinformatics group seems to be pretty well regarded around the nation," says Bob Hiromoto, chair of computer science at UI.

Cybersecurity is also a focus of the Pacific Northwest National Laboratory (PNNL) in Richland, Wash. Deb Frincke, chief scientist for the cybersecurity group at PNNL explains that one of the programs they have developed, called enhanced anomaly detection, defends a network by not only scanning its perimeter for invasion, but also by raising an alarm if it detects anything going on in the system that's at all unusual. Frincke suggests that as computers become ubiquitous, and inputs to networks increasingly come from new devices and sources, the need for security will only increase.

PNNL is also working heavily in the field of visual analytics, developing tools to assist people in analyzing data through visual representations. "For certain kinds of problems where the information is very complex, people can understand a great deal more information visually than they can through other means," says Kris Cook, project manager in PNNL's computation science and mathematics division.

The support of broad areas of science has always been the focus of PNNL's computational program. And, in October 2004, computational and information sciences became a distinct research directorate–one of five defining PNNL's main research directions. "By bringing and creating this new directorate in the laboratory it's brought computation up to the level of a research organization within the lab," says deputy division director Deb Gracio. In 2004, PNNL also connected by fiber-optic cable to the Puget Sound region to enhance networking capabilities and to make its renowned supercomputer more available to the rest of the region. All of this helps PNNL play a role in supporting computing needs in a variety of fields at the lab and throughout the region.

While there is a broad diversity of projects under way at Northwest institutions, there are a few core strengths of the region. One of these is cybersecurity. Another is network computing. Another, most notable in departments in Washington and Oregon, is collaboration with strong health sciences departments.

"There are strong medical research communities both in Oregon and at the University of Washington," says MSR's Schofield.

"If you ask Bill Gates what are the areas that are going to have the most impact over the next twenty to thirty years, he'll say, really there are two fields," says Schofield. "And they're life sciences and computing. And by the way, the thing that's changing life sciences is computing."

Schofield cautions that, despite strengths in these areas, the region will need to work to maintain its edge. "We're well positioned today, but it's a place where we could get complacent and we could fall behind. It's a place where we absolutely need to continue to invest very heavily."

Images:

Top: Machine learning and robotics are strengths of the University of Washington's Department of Computer Science and Engineering. In this image, a robot kicks a ball. Photo: UW Neural Systems Lab

Middle: Patrick Baudisch, a researcher in Microsoft's Visualization and Interaction for Business and Entertainment (VIBE) group, demonstrates moving information across different-sized displays to a group of attendees at a Microsoft Research Road Show. Photo: Susan Ragan

Bottom: INSPIRE, software developed by researchers at Pacific Northwest National Laboratory (PNNL), allows users to quickly and easily discover trends, key issues, and hidden information relationships in large volumes of text. Another software program developed by PNNL called OmniViz can analyze and graphically display genomic sequence data for life and chemical sciences. PNNL researchers received an R&D award for this technology. Photo: PNNL

Ben Raker is a Seattle writer and editor who studied science writing at the University of Washington.