5 May 1999

SCS's Goldstein and Thrun Receive NSF Career Development Awards

Seth Copen Goldstein, assistant professor of computer science, and Sebastian Thrun, assistant professor of computer science and automated learning and discovery, are recipients of 1999 National Science Foundation (NSF) Faculty Early Career Development awards. The $200,000, four-year grants, which can be matched with industrial funding and additional NSF money, are intended to foster the work of promising young scientists early in their careers when they are in greatest need of financial support.

Goldstein received the award for his work in reconfigurable computing. He has been developing chips and compilers that are not constrained to specific tasks, but can do many jobs as they're needed. "My goal is to take take general-purpose programs written in standard languages, create a circuit description like a custom chip and use it to reconfigure a piece of silicon at run time,"Goldstein says. "In this way, you get an application-specific processor that will run up to hundreds of times faster than general purpose chips.

"We want to change the very nature of reality," he continued. "For example, we want to use reconfigurable computing to reduce the cost of getting things to market while improving fidelity, handling more information, and producing better video and clearer sound."

Goldstein was born and raised in New York City. He attended the Bronx High School of Science and went on to Princeton University where he received a bachelor's degree in electrical engineering and computer science in 1985. After graduation, he worked for several years and started a company, which marketed an object-oriented programming environment. He sold the company in 1991. "I always knew I would go back into academia," he says. "I got a lot more pleasure out of people telling me how they used my product than putting money in the bank."

Goldstein earned his doctorate at the University of California at Berkeley and, imbued with the entrepreneurial spirit, started another company in California. Shortly after that, he switched his focus from developing compilers for parallel computing to reconfigurable computing. He joined the Carnegie Mellon faculty in 1997. "It costs ever more money to design new processors," he observes. "And the way technology moves, we add more transistors to chips all the time. Eventually, we'll reach a point where the expense is too great for the payback we get." He's looking for ways to use the transistors profitably.

Goldstein feels that solving the issues around reconfigurable computing will also contribute to enhancing the nascent field of nanotechnology. "I think nanotechnology is it," Goldstein says, "and one thing we can do to maximize that technology is build reconfigurable computers. I believe the technology we use to develop reconfigurable computing will be applicable to computers using nanotechnology."

Goldstein believes reconfigurable computing also has the potential to enhance space exploration. Because it offers high performance, easy design, fault tolerance and low power demands, it will make the computation parts of a space mission easier to do. Once he gets these ideas in hand, and if he remains true to form, Goldstein should be starting another company in about three years.

Sebastian Thrun is working on integrating machine learning into programming languages. He uses robots to demonstrate the results of his work. "Programmers take an overly narrow approach to instructing computers," he says. "It would be great to instruct them in the way we instruct people, rather than having to program them by hand. Our goal is to give people a programming language to support this concept. By introducing teaching methods we can become much more effective in developing software."

Thrun grew up in Hildesheim in northwestern Germany. He chose to study computer science because he had been earning extra money programming for several years. His enthusiasm for the field grew as he learned, but he was constantly looking for a way to tie the science in with people.

Thrun earned a bachelor's degree in computer science, economics and medicine from the University of Hildesheim in 1988. He went on to the University of Bonn, where he received master's and doctors degrees in 1993 and '95 respectively. He spent time at Carnegie Mellon as a visitor and a researcher before joining the faculty in 1995.

In his first semester at Bonn, Thrun decided to focus on artificial intelligence "because of the human component." He joined a research group working on robot manipulation and started trying machine learning ideas on the manipulator.

He also spent time at GMD, the German research center for computer science, where he was exposed to scientific research and publishing papers. He submitted a paper to a NIPS (Neural Information Processing Systems) conference and was amazed when his paper was accepted. "I went to Denver and gave a talk before 700 people about the fusion of machine learning and planning," he recalls. A chance meetings with SCS's Alex Waibel at that event eventually led to a talk at Carnegie Mellon and an invitation to spend a year as a visiting scientist.

Back in Bonn, and armed with concepts he had developed working with Carnegie Mellon roboticists, Thrun and his colleagues gained recognition with "Rhino," the world's first robotic museum tourguide, which did a stint at the city's Deutches Museum. Returning to Carnegie Mellon, he has continued the international collaboration and created Minerva, a museum tour guide that spent two weeks last summer at the Smithsonian's National Museum of American History.

"I believe robot systems should learn," Thrun says. "The next major advance will come when they are able to improve their performance based on experience. We are just at the beginning of fulfilling this potential."

Return to: SCS News Releases
SCS-Today
School of Computer Science homepage

This page maintained by copetas@cs.cmu.edu.