November 2, 1998
‘Smart computer’ researcher at UW wins prestigious Packard Fellowship
A University of Washington professor researching ways to build computers with the intelligence and adaptability of living creatures has been awarded a highly competitive fellowship from the David and Lucile Packard Foundation. Chris Diorio, assistant professor in the Department of Computer Science & Engineering, is in a select group of 24 promising scientists and engineers nationwide to receive one of the five-year, $625,000 fellowships.
“I’m very excited to receive this level of recognition considering where I am in my career,” Diorio says. “It says a lot about the Packard Foundation that they are willing to support some pretty far-out research with no immediate applications on the horizon. It also says something about the growing realization of the benefits of biologists and computer scientists working together more closely.”
The Packard Fellowships were established in 1988 to develop future scientific leaders, further the work of promising young scientists and engineers, encourage networking among these researchers and support efforts to attract talented graduate students into university research in the United States. Over the past 10 years, the foundation has awarded fellowships worth $125 million to 224 faculty members at 44 universities in the United States. The UW has had two previous winners: Assistant Professor of Biochemistry David Baker and Associate Professor of Astronomy Christopher Stubbs.
“The unrestricted nature of the fellowship, along with the commitment of funding for five years, will enable the fellows to pursue lines of inquiry that might be too risky for standard funding mechanisms,” said Lynn Orr, dean of earth sciences at Stanford University and chairman of the advisory panel that selects fellowship recipients. “This approach reflects David Packard’s style of operation – find talented people and give them the resources they need to do top quality work, along with the responsibility to determine how best to use those resources.”
Diorio plans to use the fellowship award to fund undergraduate and graduate students in his laboratory and to underwrite integrated circuit fabrication for student design projects – an expensive but important step in training the next generation of computer engineers.
With 13 years of experience designing chips for industry, Diorio joined the UW faculty last year with an interest in developing a new generation of circuits that can learn and adapt to their environment in the same way biological organisms do. Software companies such as Microsoft already are developing programs that try to anticipate computer users’ needs and questions based on past inputs, but these efforts are still limited by the underlying digital hardware. Diorio’s approach is to develop integrated circuits, modeled after neurobiology, that autonomously evolve to improve performance.
In his proposal to the Packard Foundation, Diorio described a set of single-transistor devices he developed that mimic the sort of adaptation and competition exhibited in nature. In his laboratory, Diorio assembled several of these “synaptic transistors” on a circuit and operated them in a random pattern. Over time, the transistors used most heavily began delivering more current output with the same voltage input while neighboring transistors delivered less output. This was the first working demonstration of a functional computer circuit adapting its performance based on usage. Diorio’s next goal is to build larger and more complex synaptic transistors that adapt to usage patterns as well as systems that recognize and adapt to changes or coincidences in time.
In August, Diorio was co-organizer of the second annual University of Washington/Microsoft Research Summer Institute, which focused this year on the exploration of intelligent systems from biological and computational perspectives. The institute brought together 45 top biologists and computer scientists to discuss ways in which their respective fields can work together to better understand the manifestation of intelligence in nature and how it can be incorporated into technology.
“The bottom line is that even though our brains operate very differently from computers, we believe we can learn from biology to build machines that exhibit intelligence,” Diorio says. “The Packard Fellowship will help us reach that goal.”
###
For more information, contact Diorio at (206) 543-7165 and diorio@cs.washington.edu, or Annedore Kushner of the Packard Foundation at (650) 948-7658.
