EE Times: Print: Front Page
Computer science R&D goes begging for funds

 
Recomputing the Future: Second of three parts

Government funding for long-term computer science research at U.S. universities has plummeted, leading many to fear the country will lose its leadership in the field that engendered the PC and the Internet. Industry-sponsored research can't make up the shortfall, and it's unclear whether government will reverse the decline.

Computer science "essentially 'outgrew' the ability of the Department of Defense to be its primary source of directional influence, let alone funding," and the DOD never developed a transition strategy, according to a report the Defense Science Board issued earlier this year.

By its own estimates, the Defense Advanced Research Projects Agency — the government's lead funding agency for long-term R&D — has cut its spending on university research in computer science almost in half over the past four years. Darpa gave universities $214 million for computer science research in fiscal year 2001 and just $123 million last fiscal year.

Darpa director Anthony Tether has suggested that the raw numbers can be deceiving. At Congressional hearings in May, Tether attributed the decline to a combination of factors: An information assurance program became classified, shifting work from universities to private groups; Congress canceled a separate, $12 million program in 2004; and two major programs — one in supercomputing and one in software agents — have progressed from the university to the industry research phase.

Overall Darpa funding for computer science and for universities in general has remained "more or less constant," Tether told Congress.

The raw numbers may also be deceiving — and troubling — at the National Science Foundation, the other major gov-ernment source of support for long-term computer science research. NSF funding for such projects has in fact risen since fiscal 2001, from $478 million that year to an estimated $613 million this fiscal year — but the rise hasn't kept pace with the growth of the field. Thus, whereas NSF once accepted a third of all proposals, last year it accepted just 14 percent of research proposals in computing.

"We are looking at a situation where perhaps 40 percent of the good proposals we get, we don't have the money to fund," said Peter A. Freeman, who heads the computer directorate at NSF.

"It's a difficult situation; computing is a very important issue, but it's not urgent," Freeman said. "There are so many urgent issues — war in Iraq and on terrorism and natural disasters — that soak up resources. It doesn't leave us money for things that, in the long term, history may judge as more important."

"We have a broad concern, not just with Darpa but with NSF, too — and not just in computer science but in the whole IT area generally," said Dick Lampman, director of HP Labs.

"Research communities are literally falling apart," said Justin Rattner, director of research at Intel Corp. "I get e-mails every month from top researchers saying, 'NSF just cut my funding; can Intel support my program?' We do what we can, but it's hard to replace NSF funding."

Darpa, meanwhile, "has abandoned the community of researchers that has been responsible for so many important advances. It will be hard to rebuild this community at a reasonable cost once it disperses," Rattner said.

Universities are feeling the pinch. At Stanford, the amount of research funded by Darpa has collapsed from about 80 percent of the university's CS budget in the late 1990s to just 15 percent and "dropping rapidly" today, said Bill Dally, chairman of the university's CS department. Darpa's shift to industry partnerships "pushes universities into short-term [research] projects," he added.

At the Massachusetts Institute of Technology, Darpa funding in 1999 accounted for 62 percent of the CS budget, with the money earmarked for three- to five-year projects. Today Darpa supplies just 24 percent of MIT's CS budget, mainly for 12- to 18-month projects, said chairman Rodney Brooks, chairman of the university's CS department.

"It's getting harder to develop fundamentally new ideas," Brooks said. "We are doing more work on things we already know and turning it into flashy demos. That doesn't make for a PhD thesis."

"I don't envy the younger generation of faculty today. It has a major effect on morale when you can't get a grant for three years," said Jitendra Malik, chairman of the CS department at the University of California at Berkeley. "This is sad because it comes at a time when CS problems are still varied and vibrant."

Edward D. Lazowska, a past chairman of the Computer Research Association (www.cra.org) and a CS professor at the University of Washington, blames the current Bush administration for running up the national debt while cutting back on computer R&D, a field with a track record for fueling economic productivity and growth.

"In 10 years there will be a double whammy. These debts will come due, and we will have downshifted the productivity engine that helps pay them," Lazowska said. "The nation has an intellectual as well as a physical infrastructure. If you don't make adequate investments in it, it won't bite you for perhaps five years, but when it does the cost of recovery is enormous."

China calling
Dally of Stanford is among many researchers who blame the crisis on a shift by Darpa to more short-term research with industry as well as the rising costs of the war in Iraq and national emergencies such as Hurricane Katrina.

"All the money is getting sucked up, and research is easy to cut," said Dally. "This is a wake-up call. There is a crisis brewing in this country. Some other country could come in and take away leadership in computing.

"By abdicating our research leadership position we make it easy for a South Korea, India or China to step up to the research plate," Dally added. "If one country put $500 million or $1 billion a year into building world-class computer science research, they could take leadership within a decade — 15 years at the outside."

That's already happening, according to Microsoft Research director Rick Rashid, who recently returned from a tour of China, where the company has a strong research presence. Rashid rattled off a list of new construction projects at universities in Beijing, Shanghai and other cities — including a new, 30,000-student university built in the last three years.

"The Chinese have decided investing in universities is critical to their future, and they are putting an enormous amount of resources into the physical infrastructure. It's eye-opening," said Rashid. "There's no doubt it will put them in a good position in five, 10 or 15 years."

About half of the basic-research component of the 700-person Microsoft Research group is in labs outside the United States. The company recently opened a lab in Bangalore, India, and in December will launch a partnership in Europe to create an institute for computational biology, Rashid said.

The computer industry's other main research organization, Intel's 900-person Corporate Technology Group, has likewise gone international.

"Five years ago, we had virtually nothing in research outside the U.S. We were just setting up our Beijing center," said Intel's Rattner. "Now we have labs in China, India, several in Russia. And we are setting up another, in Guadalajara, Mexico."

U.S. universities graduate about 1,000 CS PhDs a year, compared with about 2,000 in China and just 30 in India. While Rattner asserted that the "U.S.-trained researcher is still dramatically better than a researcher from just about anywhere else on the planet," he warned that U.S. superiority is not a given. "China and India are turning out astonishing quantities of [technically trained] people, and they will soon have both quantity and quality," he said. "U.S. competitiveness is at risk."

"From my discussions with governments outside the U.S., they would love to have more research in their backyards, whether it's at universities or by companies moving in," said Prabhakar Raghavan, a former IBM Corp. researcher who is expanding research operations for Yahoo Inc. from its current corporate R&D base of roughly 40 people.

Proposed U.S. government policies that would exert security-related restrictions on foreign-born researchers could exacerbate the problem. The Department of Commerce proposes that foreign-born researchers be licensed to operate some technical equipment in university labs. The DOD is proposing separate security clearances for foreigners working in export-controlled areas. The proposals could go into effect next year.

"We are creating a hostile environment for the intellectual elite," said the University of Washington's CS professor Lazowska. "Either we let these bright people into our country, or our country will become uncompetitive as they move offshore."

The computer industry is making small attempts to bolster university research, but given the size of the government losses and the consolidating nature of the PC-driven industry, such efforts aren't nearly enough to breach the funding gap.

Sizing up industry research in CS, Malik of Berkeley recalled that Bell Labs, Xerox Parc, HP Labs and IBM Research have declined or narrowed their focus in the past 20 years, while Intel, Microsoft and Yahoo labs have grown. "If I were to construct a balance sheet, on the whole the balance is negative," he said.

"It's pretty clear there are a lot fewer companies — just a handful — making investments in research," said Lampman of HP Labs, which employs about 600 researchers.

"We plan to double our university research budget over the next three years [to 2008]. I just got that approved," said Intel's Rattner. "We've tried to pick up some of the slack. But no one is going to re-create a Bell Labs."

Nokia announced in late October that it will spend $2.5 million a year for three years as part of an open-ended deal for a mobile research lab conducting long-term chip, software and services research at MIT.

Microsoft fellowships
Last year, Microsoft started a $1 million annual program that hands out five two-year university research fellowships. "I've been lobbying people to pony up more, but companies can't replace what the government does in the level of investment," said Rashid, whose Microsoft Research budget is estimated to total about $200 million.

There are limits to what shareholders will let companies spend on university labs, Washington's Lazowska observed.

The government may be ready to revisit the issue. In the Pentagon, on Capitol Hill and in the administration, policy-makers are reassessing requirements and the government's role in helping to fill them (see story, page 16).

But Lazowska was skeptical: "Will this administration adequately prioritize engineering, science, advanced education and research? I haven't seen signs of that."