A new technology transfer office will help researchers assess whether work can be transferred to internal business units or licensed for external use. That could boost projects like Richter, an ultra-sensitive accelerometer developed by the Information and Quantum Systems Lab under veteran nanotechnology researcher Stan Williams.
"In the past, when we got to this advanced development stage, it was a crap shoot getting [a technology] into the market," Williams said.
"We have a lot of structure now to focus on the big problems we can solve to have an impact," said Qian Lin, director of the Multimedia Interaction Lab, which is working on face recognition, among other projects.
"The proposal process is a new thing, and we will have to make bigger bets," said Rob Schreiber, assistant director of HP's Exascale Computing Lab. But "If you don't make bigger bets, big things don't happen," he added.
IBM Research also tries to focus about a third of its work on fundamental science, and it maintains close ties to its internal business units as well as universities and venture capitalists. Like HP, IBM has defined a handful of broad research themes for which it sets "grand challenge" tasks, such as creating a petaflops computer.
Shifting to software
But "everybody does [things] a little differently," said Dally of Stanford.
"The big player these days is Microsoft Research. They are a force to be reckoned with," with big external-funding programs and researchers who publish regularly and show up at technical conferences, Dally said. By contrast, Intel has funded many university projects but has not has as high a profile at conferences or in peer-reviewed journals. As for IBM, "it's a shadow of its former self," Dally said. "Their work has become much more applied research, and you see less publishing from them."
Apple, meanwhile, has carved out a unique space with its iPods and iPhones, but the company largely popularizes the good R&D done elsewhere in multitouch displays and basic wireless and digital media technologies."It's very rare now that any company in the computer industry can control something end to end, so research really requires collaboration," said Joel West, an associate professor of innovation and entrepreneurship at San Jose State University and co-author of a book on collaborative innovation.
West noted that HP inadvertently gave birth to the PC industry when one of its employees, Steve Wozniak, left the company to help found Apple because his inventions had no place inside HP.
Today, Google is a poster child for open innovation, especially inside the organization. Though the search-engine giant has only 120 full-time researchers, it encourages all employees to spend time on pet projects of interest.
"Most companies have a mix of open and closed innovation, the latter one being their secret sauce," said West.
Just what HP's secret sauce will be remains to be seen. At its March 6 event, Shane Robison, chief strategy and technology officer, suggested in a visionary talk that the future is all about software and services, not desktops and printers.
"Today we are at a stage where the Internet is pervasive as a platform and software developers can target it for the broadest distribution of their work," Robison said. "Increasingly, we are presenting our technologies as services."
Many of the lab projects HP showcased at its event were software innovations. But the company still has plenty of active hardware projects.
"If we can develop integrated CMOS photonics, their low-power and high-bandwidth capabilities could be applied to many-core processors that could remove some of the communications and synchronization bottlenecks that are a big problem in parallel computing," said HP's Schreiber. His lab is doing such work with colleagues in the lab of nanotechnology researcher Williams.
A year ago, HP jumped into the search for new and better parallel-programming techniques, Schreiber said. The company is now working on transaction processing--new ways to take advantage of processors with different kinds of cores and new, low-level software constructs.