Berkeley courts industry as funding declines

UC Berkely, California

The University of California at Berkeley may seem a model for collaboration among industry and academia. Indeed, being part of a state-funded school system, the university's charter is to help boost the California economy and industry with able-minded graduates.

So with great fanfare the university's Reliable, Adaptive and Distributed systems laboratory, or RAD Lab, recently announced alliances with five major corporations.

IBM, Hewlett-Packard, NTT Multimedia Communications Laboratories, Nortel Networks and Oracle each have signed on as affiliate members of the research lab, pledging $170,000 each annually for the next five years

The three add to funding from the lab's three founding corporate financiers, Sun Microsystems, Microsoft and Google, which late last year committed to underwriting the new $7.5-million lab to the tune of $500,000 each annually for five years.

Shorter-term projects

Several academic laboratories, including the Media Lab at the Massachusetts Institute of Technology, long have had corporate sponsors. But the RAD Lab's move was prompted by what its director said is a recent trend of federal support for basic research in computer science away from basic or "pre-competitive" research and towards applied research and shorter-term project financing. That is forcing top research schools like Berkeley, Carnegie Mellow University and even MIT to seek similar industry partnerships to fill in the federal funding gap.

The financing of basic university research by corporations is a trend that likely will keep spreading in the United States, Europe and elsewhere, according to the head of the RAD Lab.

"It's hard to fund an open research project. Until recently, federal grants from the Department of Defense's Defense Advanced Research Projects Agency (DARPA) and the National Science Foundation (NSF) covered research costs for large projects such as this," said David Patterson, the RAD Lab's founding director, who was turned down by both federal funding agencies before going to industry. "Reductions in government funding of information technology research have led us to seek alternative funding sources. We did this because we were desperate, but we do believe in this corporate funding model."

Patterson also is a UC Berkeley professor of electrical engineering and computer sciences, and the key researcher behind the reduced instruction set computer (RISC) architecture made popular by Sun Microsystems and other computer companies as well as the redundant array of independent disks (RAID) disk storage technology popularised by EMC Corp.

Spending declines

In a memorandum to a government subcommittee last year, DARPA said its spending on basic computer science research at universities declined by 5 per cent from 2003 to 2004. DARPA awarded about $207 million in 2002 to universities for computer science research, but that fell to $123 million in 2004, and NSF has incurred both a funding cut and it is funding fewer proposals, down from 30 to 35 per cent in the late 1990s to 21 per cent in 2005.

The RAD Lab now expects to receive as much of 80 per cent of its financial support from industry, compared with 80 per cent from federal government sources from 1980 to 2000. A UC Discovery Grant, a grant from the state of California's Microelectronics Innovation and Computer Research Opportunities programme, and another grant from the National Science Foundation will make up the rest of the lab's current funding.

The RAD Lab was launched in December 2005 to adapt advances in statistical machine learning, a type of artificial intelligence, to the task of maintaining large distributed computing systems that are needed to run Internet businesses, which increasingly need to process more data. Such systems typically need hundreds of engineers to develop, debug and maintain complex computer and network systems. The RAD Lab is designing software that will automate these tasks.

But the lab's work is a large and long-term science project, the type that eventually will require 30 or more graduate students, double the current number. In the past, the US government has funded such "big computer science" projects that made their way into the public domain, including the Internet. Other countries followed, including Japan with its "Fifth Generation" artificial intelligence project in the 1980s.

Following the leader?

"The United States used to be an influential country around the world in that when our government defence agency funded science, other industrialised countries did the same thing," Patterson said. He added that other industrialised countries, such as the United Kingdom and Japan, are likely to follow the US lead in pinching computer research funding, just as they followed the lead earlier when the US pumped up funding.

Patterson said he knows of only one major computer project receiving federal funding recently, the $4.3 million that DARPA granted to the University of Texas at Austin's Department of Computer Sciences, where professors Steve Keckler and Doug Burger have muscled together 30 researchers on the school's Tera-op, Reliable, Intelligently adaptive Processing System, or TRIPS, project. TRIPS aims to develop a new type of scalable, power efficient and high-performance microprocessor architecture.

The DARPA funds run from October 2005 through December 2007, exemplifying what Patterson calls a difficult effort to have to raise more research funds every two years for projects that can last five or more years.

Patterson said he is concerned about the consequences of the dwindling federal support for large-science IT projects. There are other problems as well. Even the largest companies in industry cannot maintain an adequate support level for big science, especially when results may be years away. The disappointment by companies was voiced in a recent report by Lux Research on companies supporting nanotechnology research, which were disappointed by the few commercial successes and smaller return on investment than they had expected.

"Universities are still the best deal in town for access to talent and a fresh perspective, but companies using these collaborations as outsourced research groups can't bend the professors, or students for that matter, to their commercial will," Lux senior analyst David Lackner said in a statement upon releasing the report.

Hands-off

In Patterson's case, the faculty and scientists decide how the money is spent, and the companies have taken a pretty hands-off approach to the basic research. The companies are involved in three-day retreats twice a year with the RAD Lab researchers. The companies can offer advice to the RAD Lab, but they are not able to send representatives to work at the RAD Lab, as has been possible by sponsors of MIT's Media Lab.

In the case of the RAD Lab, companies may not even cash in on their investment: all software and applications emerging from the RAD Lab will be available freely and openly to the public. The benefit the three founding members and the affiliates will get is that they will have access to the technologies at least six months before they are available to the general public.

"I've never filed for a patent at Berkeley," Patterson said. "We put things into the public domain."

"It's a Catch 22 of research funding," he added. "Industry funding research is good for the field and the country, but it might not be good for the company. For the companies, our most valuable product is the students, who are the next generation of leaders for IT."

Another problem is that all but the top 10 research universities may find little left over as they scrounge for corporate funds.

"I don't think this can be done by lots of companies in lots of universities," Patterson said of large science funding. "You need about $100 million to $200 million for scientists working worldwide on a big science project. Companies don't have that kind of funds."

Those days are gone

Gone are the days when Xerox Corp. heaped money into its own future technology research at the Xerox Parc labs and AT&T pumped money into Bell Labs.

Patterson said the best answer long term is for the government to come around and restore funding levels for big projects such as the next generation of parallel processing microprocessors and the software to run them, as well as software for large systems and other advanced information technologies to succeed and feed the economy.

And while Patterson sees the UK and Japan following the current US government trend in cutting funding of long-term computer projects now, he sees a different story emerging in newly industrialising countries like China and India.

China recently announced a new 15-year plan for science and technology, which is expected to produce $30 billion (236 billion yuan) in spending from all sources, including industry, in 2005, and rise to $113 billion (900 billion yuan) in 2020. Basic research is expected to increase from 6 percent of the R&D outlays in 2004 to close to 15 percent of the total in 15 years.

But Patterson and other researchers continue to meet with representatives of DARPA and other funding sources to discuss the funding issue and try to interest the government in once again picking up the tab for the long-term, riskier big science projects that could promise large economic benefit in the future. Otherwise, he fears the candle will burn at both ends.

"The government is cutting back on university lab funding, and industry is relying more on university research," he said. "So you have two different parts of society cutting back on research funding and thinking the other one is picking it up. This is a deadly combination for the future of the United States."