Three tech giants to finance research

University of California computer scientists plan to announce on Thursday that the companies--Google, Microsoft and Sun Microsystems--will underwrite a $7.5 million laboratory on the Berkeley campus. The new research center, called the Reliable, Adaptive and Distributed Systems Laboratory, will focus on the design of more dependable computing systems.

The Berkeley researchers say that under the terms of their agreement with the three companies, the fruits of the research will be nonproprietary and freely licensed. Each company has agreed to support the project with $500,000 annually for five years. Although the companies are frequently rivals and only occasionally allies, they have concluded that they can operate most effectively by bringing technology innovations to market quickly.

Computer scientists have grown increasingly alarmed that federal support for basic or "pre-competitive" research is being eroded by shifts toward applied research and shorter-term financing.

Earlier this year, MIT researchers announced several similar corporate-backed basic research efforts, and Carnegie Mellon University officials said they were working on similar arrangements.

The Berkeley lab's founding director, David A. Patterson, is a veteran computer scientist who has led a variety of academic research projects that have had a significant influence on the computing industry since the 1980s.

Patterson, currently the president of the Association for Computing Machinery, a national technical organization, has recently been a vocal critic of the shift of basic research funds away from universities and toward military contractors.

"We're trying to sustain the broad vision, high-risk and high-reward research model," Patterson said of the new Berkeley effort.

The Berkeley researchers began looking for industry support last year when they realized that the Pentagon Defense Advanced Research Projects Agency, known as Darpa, was withdrawing support for basic research at the university, he said.

In a memorandum submitted to a Congressional committee earlier this year, Darpa officials disclosed that its spending on basic computer science research at universities had declined by 5 percent between 2003 and 2004. Government officials and corporate research executives noted the indirect effects of the changes in federal research support over the last five years.

"When funding gets tight, both researchers and funders become increasingly risk-averse," said William Wulf, president of the National Academy of Engineering.

During the 1980s and early 1990s, Patterson directed research efforts at Berkeley into processor and computer storage design. The projects helped lead to reduced instruction set computing, or RISC, and redundant array of independent disks, or RAID. Both start-ups and existing computing companies capitalized on the ideas and adopted the technologies. Patterson was also involved in a project at Berkeley that led directly to Inktomi, an early Web search engine.

A focus on machine learning
Sun in particular has relied on Berkeley research, incorporating it into the design of its Solaris operating system and its Sparc processor, the latter of which was based on Patterson's RISC project.

While welcoming corporate funds for research on information technology, a number of industry executives and academic officials cautioned that such financing would not fill the gap left by the federal shift in priorities.

"I think it's terrific that a number of leading IT companies are stepping up to the plate to fund this research," said Thomas A. Kalil, a technology adviser to the Clinton administration who is now special assistant to the chancellor for science and technology at Berkeley. "But I don't think it's realistic that companies will be able to fill the void left by federal retrenchment."

The new Berkeley laboratory will consist of six faculty members and as many as 30 graduate students.

The research focus of the new center will be to apply advances in the use of statistical techniques in machine learning to Web services--from maps to e-mail to online calendars--which have become an increasingly important part of the commercial Internet.

Machine learning is the study and application of computer algorithms that improve automatically through experience.

Daniel T. Ling, Microsoft's corporate vice president for research, said the Berkeley lab "fits perfectly into our recently announced interest in Web services."

The power of machine-learning software was demonstrated earlier this year during the Darpa Grand Challenge, which pitted dozens of research teams as their robotic vehicles independently navigated a desert course. The contest featured a $2 million prize.

The Berkeley researchers are hoping to use these software techniques to make computing systems that provide the growing array of Web services more stable and easier to support.

"There is something quietly happening in machine learning that is going to result in a real revolution," said Greg Papadopoulos, Sun's chief technology officer.

The research plays well to Sun's businesses, Papadopoulos said, but just as important, it is vital to create a supply of skilled researchers who can develop system-level projects.

"We have to concentrate on finding the people with these skills," he said. "Because we aren't going to be able to out-code the rest of the world."