January 2006
During his long career, professor David Patterson led the design and implementation of Reduced Instruction Set Computer (RISC) and, with professor Randy Katz, Redundant Arrays of Inexpensive Disks (RAID) technology, both of which became industry standards. |
In October, five robotic vehicles drove themselves more than 130 miles across the Mojave Desert, making international headlines and marking a milestone in machine learning. Related to artificial intelligence, statistical machine learning refers to methods that enable a computer to improve its performance by analyzing previous results. Now, a group of UC Berkeley researchers hope to bring the same technology into cyberspace. The new Reliable, Adaptive and Distributed systems Laboratory (RAD Lab), funded with $7.5 million from Google, Microsoft, and Sun Microsystems, is developing technology that leverages the power of statistical machine learning so that one person can create the next eBay, Amazon, or even Google, by herself.
"eBay has changed the world but they had to hire hundreds of really smart people to pull it off," says computer science professor David Patterson, founding director of the RAD Lab. "Wouldn't it be great if one person could create a eBay-sized system without an eBay-sized organization?"
To make this vision a reality, the RAD Lab, which falls under the Center for Information Technology Research in the Interest of Society (CITRIS), must weave together innovations in such diverse disciplines as networking, computer architecture, systems theory, and statistics. To that end, Patterson is collaborating with a handful of UC Berkeley professors who are recognized as pioneers in those areas: Randy Katz, Scott Shenker, Ion Stoica, Armando Fox (joining Berkeley next summer), and Michael Jordan, who holds a joint position in computer science and the Department of Statistics.
In an early experiment, the RAD Lab researchers created "operator-friendly" visualizations to cross check results generated by machine learning algorithms. Analyzing real outage data provided by Ebates.com, the researchers determined that many failures could have been predicted hours in advance. (courtesy RAD Lab) [View larger image] |
The RAD Lab calls for a new paradigm in the traditional software design and development process. The current model of software design follows a step-by-step path from the initial idea through development, testing, deployment, to ongoing operation. According to Patterson though, that "waterfall" model is obsolete. It's far too slow and hierarchical for Internet applications where the software behind Web services is constantly being tweaked, improved, and scaled up to accommodate growing numbers of users.
The RAD Lab embraces a very different approach where the operation of the Web applications and services is measured in real-time. Insights gleaned from that data is immediately fed back into the development process to improve the code. The model, Patterson explains, "shortens the 'distance' between a service's users and its developers and allows for faster innovation and bug fixing." And therein lies the rub.
"How can a single human keep track of everything, fix problems quickly, and make sure that the service never goes off the air?" Patterson says.
The answer, he says, is to automatically close the feedback loops using statistical machine learning techniques. Statistical machine learning employs pattern-finding algorithms that compare historical data and find commonalities between them. Those commonalities can then be used to generate a predictive model of what's likely to happen in the future.
"The strength of machine learning is sifting through massive amounts of data and finding insights into what's going on," Patterson says. "So we're trying to come up with statistical methods to measure and observe what's happening within Web services and notice indications of problems far enough in advance that problems can be prevented. Then, techniques taken control theory could be used to automatically help fix the things that are wrong."
Initially, RAD Lab research will be conducted by the faculty co-founders and ten computer science graduate students. The number of collaborators is expected to grow as the research progresses. From day one though, all of the software developed within the RAD Lab will be available to the public through the Berkeley Software Distribution license, a revolutionary "open source" approach created at the University in the 1970s that enables anyone to build upon, change, and use the raw programming code free of charge.
"Our vision is to enable one person to invent and run the next revolutionary Internet service, operationally expressing a new business idea as a multi-million-user service over the course of a long weekend," Patterson says.
Ilan Gur will appear in an episode of Chasing Nature, a new show about engineering and nature that will air on Animal Planet. In 2004, as part of a joint program between the University’s Management of Technology (MOT) program and the United Nations Industrial Development Organization (UNIDO), Gur conducted field research in western China to explore solar-powered lighting solutions for off-grid populations. |
UC Berkeley graduate student Ilan Gur and his colleagues have a profound research goal. "In the long run, we'd like to power the world," he says. A PhD candidate in the Department of Materials Science and Engineering, Gur is leveraging advances in nanoscience to develop ultra-thin and cheap solar cells that someday could be batch produced in bulk, perhaps even roll-to-roll like newspapers are printed. Flexible and durable, the cells might eventually wrap the roofs of buildings or transform a cloth automobile cover into a battery charger.
"The reason you don't see solar cells on rooftops everywhere is because they're incredibly expensive to produce," says Gur, who is a student in the research group of nanotechnology pioneer Paul Alivisatos. "Traditional solar cells require high-purity silicon that must be processed in high-temperature, high-vacuum conditions and clean-room environments like those in microchip fabrication facilities."
This scanning electron microscope image depicts a film of nanocrystal solar cells on top of a silicon substrate. (courtesy the researchers) |
Gur spearheaded the development of the first solar cells ever to be made entirely from inorganic nanocrystals, chemically-pure clusters of anywhere from 100 to 100,000 atoms. Unlike the silicon used in today's solar cells, the rod-shaped nanocrystals of cadmium-selenide and cadmium-telluride are deposited in a chemical solution process at fairly low temperatures. Essentially, they're made in a beaker instead of a clean room. The resulting film of photovoltaic material is just 200 nanometers thick, or 500 times thinner than a human hair.
The inspiration for the new approach came from previous breakthroughs in the Alivisatos Group in the creation of prototype solar cells made from a combination of nanocrystals and an organic polymer, or plastic. The researchers showed that the plastic solar cells could be inexpensively manufactured in bulk quantities. Nanosys, a start-up company co-founded by Alivisatos, licensed the technology to further develop it for commercialization.
A close-up image of one of the prototype solar cells. (courtesy the researchers) |
"With that technology moving to the next stage of development, we started thinking about the ultimate limitations of plastic solar cells," says Gur, who is also a researcher in Lawrence Berkeley National Laboratory's Materials Sciences Division. "One problem is that the polymers are organic so they degrade in air. The cells must be encapsulated to protect them, and that's a pretty serious limitation when you're trying to make something that's dirt cheap."
The researchers realized that an all-nanocrystal cell might offer the best of both worlds. The nanocrystals are inorganic, so they aren't sensitive to air. And while they can be processed far more easily than the silicon in traditional photovoltaic cells, the nanocrystals share the semiconductor's desirable sunlight absorption and electron transport characteristics.
"If you can make this technology practical, you could imagine that people would be able to make solar cells where they don't have the tremendous capital needed to build plants, like in developing nations," Gur says.
In October, Gur, Alivisatos, and their collaborators published a paper in Science outlining their novel approach. While the mechanism by which the current flows in nanocrystal cells is similar to the plastic devices, the performance doesn't seem to deteriorate as the material ages. Still, the first prototype cells, layered on a conductive glass substrate, can only convert 3 percent of energy from light into electricity, far less than most of today's commercial solar cells that are more than 10 percent efficient. However, the nanocrystal cells are comparable in efficiency to the state-of-the-art in plastic solar cells and can likely be optimized enough to make the technology marketable.
"If we can produce this for as cheap as we hope, the efficiency doesn't have to be so high because you could just install more of the material," Gur says. "There are certainly places where people would trade space for energy."
The researchers are now working to boost the efficiency by altering the structures of the nanocrystals and tweaking the chemistry of the production process. Meanwhile, they're exploring other kinds of nanocrystals made from elements more common and environmentally-benign than the cadmium-based nanocrystals used in the first experiments.
"The world is facing a massive energy problem," Gur says. "It's not clear that this is the solution, but it has potential. And I know that we have to be working on many potential solutions if we ever hope to solve the problem."
Josephine Chang with the sensor chamber apparatus used to test the electronic nose. In previous work, Chang wove an electric textile that incorporated organic transistors. |
Imagine that the milk in your refrigerator could sniff itself, changing the color of the carton if its contents were spoiled, or tiny sensors throughout an airport unobtrusively screened for explosives, replacing the bulky, expensive machines that contribute to security checkpoint traffic jams. UC Berkeley graduate student Josephine Chang is building just such an "electronic nose." Fabricated with a modified inkjet printer squirting organic electronic inks, the e-nose could potentially beat the sensitivity of today's commercial gas analyzers while costing tens of dollars instead of tens of thousands.
"Right now, electronic noses are so expensive that they're mainly used by the military or sometimes in industry," Chang says. "But if we can significantly lower the cost, they could be everywhere, inside toasters, medicine cabinets, possibly even food packaging."
Chang is a graduate student in the research group of Vivek Subramanian, a professor of electrical engineering and computer sciences. Subramanian and his students are pioneers in organic electronics, a form of electronics that uses conductive polymers, or plastics, in lieu of inorganic materials such as the copper or silicon found in traditional circuits. Previously, the researchers have demonstrated an inkjet printer and family of electronic inks that can pattern circuits onto paper, plastic, or cloth without damaging the material.
Researchers in Subramanian's laboratory and elsewhere have experimented with printable organic electronics to create ultra-low cost "smart tags" that promise to replace the ubiquitous UPC bar code on products. Subramanian has also made headway on inexpensive plastic displays that could be rolled up. Still, a great many hurdles remain before organic electronics can provide the fast switching and high flow of current necessary for those applications. According to Chang though, the electronic nose technology is "very forgiving of organics' weaknesses."
"One challenge with organic electronics is they're very sensitive to their environments," she says. "It's hard to keep them stable in air. But using organics for sensors takes advantage of that sensitivity. It flips around a weakness and turns it into a strength."
A microscope image of an organic thin film transistor sensor that the researchers printed using conductive ink. [view larger image] |
The basic component in the electronic nose is an organic thin film transistor (OTFT). These are similar to transistors on a computer chip, only much larger and slower. Because the OTFT's electrical properties change in the presence of certain compounds, the devices are well-suited for sensing applications. Furthermore, the organic transistors can be chemically altered to make them respond to different substances, alcohols or acids, for example. As a result, the researchers were able to print an array of sensors, each tuned to measure a particular kind of gas, onto a single piece of silicon. The "all-purpose" electronic nose can then be "trained" to identify a particular odor based on the signature response of the sensors.
"It's pattern recognition," says Chang, who worked on the project with Subramanian, undergraduate student Vincent Liu, and a team led by College of Chemistry professor Jean Fréchet. "Apple pies may smell differently, but you can know the general pattern of the odor."
Last December, the electronic nose research won Chang and graduate students Brian Mattis and Steve Molesa a top prize in the Berkeley Technology Breakthrough Competition, sponsored by the Center for Entrepreneurship and Technology. The aim of the competition is to "showcase high-impact science research and discoveries with the potential to be commercialized within the next five years."
Right now though, Chang isn't planning to take the electronic nose to market. After earning her PhD later this year, she'll move to the East Coast to join IBM's famed Watson Research Center.
Cool Alumni: Inspiring engineers, one student at a time
by Patti Meagher
Kevin Kornegay (M.S.’90, Ph.D.’92 EECS) with a vehicle built by students on his Autonomous Underwater Vehicle Team, will move in January to Georgia Institute of Technology, where he will be the Motorola Professor in the School of Electrical and Computer Engineering.
DENIS DEFIBAUGH PHOTO |
Growing up in Queens, New York, Kevin Kornegay was "a nerd and proud of it," he says. "I was always building radios or oscilloscopes and tearing apart electric motors." His natural tendency to tinker was fueled not only by the electronic gadgets his mother bought him, but also by techno-wizard Barney Collier, the character played by black actor Greg Morris on CBS television's 1966–73 hit spy show, Mission Impossible.
"Collier was a technologist, and that had a significant impact on me," Kornegay says. "When African Americans are portrayed positively in the media, kids looking for examples to emulate can find important role models."
Kornegay (M.S.'90, Ph.D.'92 EECS) began his professional career as a researcher at IBM's T.J. Watson Research Center in Yorktown Heights, New York. In 1998 he joined Cornell, where he is now associate professor of electrical and computer engineering, focusing on mixed-signal integrated circuit design for broadband communications. Best known for founding Cornell's Broadband Communications Research Lab and his work on high-performance circuit design, he has recently mentored a number of award-winning student teams building autonomous, or unmanned, submarines.
"Autonomous systems require integration of sensors, computation, mechanical and electrical systems, and artificial intelligence," Kornegay says. "This is systems engineering at its highest level." Watching his students put it all together, he adds, is his favorite part of the job.
"These projects culminate in designs that bring theory and practice together," he says. "It's thrilling to watch the students solve challenging real-world problems in real time."
Kornegay's honors are too numerous to list in their entirety. They include the 2005 Janice Lumpkin Educator of the Year award from the National Society of Black Engineers; recognition as one of Science Spectrum magazine's 2005 "Trailblazers," a list of the nation's top minority scientists; and a National Science Foundation CAREER Award.
"When I graduated from Berkeley," he says, "there were only two African American electrical engineering Ph.D. graduates in the country, and I was one of them. For the most part, those numbers haven't changed." Recent figures from the National Science Foundation confirm that blacks, Latinos, and Native Americans account for 23 percent of the U.S. population, but only 6 percent of its science and engineering labor force.
Today, Kornegay is doing his part to inspire more women and underrepresented minorities to excel in science and engineering. In his seven years at Cornell, he has mentored 14 Ph.D. graduates, three of them African Americans, one Latino, and one woman.
"Through my professional accomplishments, I try to be an example," Kornegay says. "If I can do it, they can do it."
