SAN JOSE, Calif. Researchers in computer science are taking on new models and fresh subjects as they grapple with new and old challenges, according to a panel discussion that gathered four senior R&D executives at ResearchFest, an event sponsored by eBay Inc.
"One of the differences about research at Google is the extent to which researchers are embedded--everyone is mixed together," said Peter Norvig, director of research at Google, noting that many researchers do their work while part of product development teams.
"While engineers are told by their managers what to work on and then they make suggestions, the researcher tells the manager what he will work on and the manager makes suggestions," Norvig said.
'"The question is why doesn't everyone do this? Why do they isolate research as the building on the hill?" he asked.
Norvig answered his own questions, noting that most companies are concerned researchers will get sucked into short-term product development roles and lose focus on novel and long-term projects. That's one reason why Xerox purposely created its Palo Alto Research Center (PARC) on the opposite coast from its head office, said Mark Bernstein, president of PARC.
Indeed, getting senior researchers to work on fresh topics is one of the ongoing challenges in R&D, said Jim Spohrer, director of services research at IBM's Almaden Research center in San Jose.
"We don't get anywhere near the one third of projects based on out-of-the-box thinking we would like to see for long term impact," said Spohrer. "The most mature senior researchers do conservative things to help the business along while the new researchers come up with the wacky ideas we like," he said.
"It's not that we are not willing to back risky ideas. It's getting the best and brightest minds to focus on them. The measurement systems keep people thinking inside the box," Spohrer added.
Among its other novel approaches, Google sometimes wraps an engineering team around a research effort, said Norvig. For instance, its speech recognition researchers work with the Google 411 product team and a machine translation research effort is working with a separate product group.
"The researchers keep working on making the underlying technology better while the products roll," Norvig said.
Google has been able to maintain the close ties between research and product teams, in part because its software products generally require relatively small teams and they often work under flexible release deadlines. The company has also been able to maintain its engineering-led culture despite its ten years of rapid growth, he said.
"In most companies there comes a time when the guys in suits come in and push engineers out. That hasn't happened at Goggle yet," he said.
Panelists agreed Web services have become a hot new focus area in computer research. "The ability to provide information as a service will be the next big opportunity for the next 20 years," said Bernstein of PARC.
IBM aims to hire many service scientists, just as it pioneered hiring computer scientists more than a generation ago, said Spohrer. Long term, the company aims to turn a business modeling methodology it acquired with consulting firm PriceWaterhouseCoopers into a CAD simulation tool, predicting the likely size of business a new services concept could spawn.
"We would not build a chip without doing a simulation first so why embark on a multimillion dollar business without simulating it first?" he asked. "We think we can get very scientific about this."
Panelists disagreed about the extent to which a rising business in patents has impacted R&D departments. "This has virtually no impact on the day-to-day life of a researcher," said Ashok Chandra, general manager of search labs at Microsoft.
However, Norvig recounted a recent email asking if he could provide one or more researchers for a sabbatical to study prior art on a patent being asserted against the Internet company. For his part, Bernstein said PARC had just closed a $30 million licensing deal.
The patent system is effectively broken, said Spohrer of IBM, and if it did not exist already could not be brought into being today.
"This is an area ripe for some brilliant innovation," said Spohrer. "We all want the benefits of our innovation but the current patent system is not the best way of doing it."
The ResearchFest event was timed with news that the company received the 2007 National Medal of Technology and Innovation, the first Internet company to get the award. At a White House ceremony Monday (Sept. 29), President George W. Bush cited eBay's online trade technology as a factor in the growth of the Internet worldwide.
The company was founded over Labor Day weekend in 1995, when computer programmer Pierre Omidyar wrote the code for an auction website on his home computer. Omidyar sold a broken laser pointer in the company's first online auction, fetching a surprising $14.98, said Laurie Norrington, president of eBay's Marketplaces division.
"Last year we sold $60 billion in goods on eBay, making us comparable to WalMart or target," she said.
Researchers at eBay are focused on refining the company's dynamic search algorithms and exploring issues in so-called cloud computing among other topics, according to lab director Neel Sundaresan, who hosted a separate panel at the event.