Bring up the U.S. Defense Advanced Research Projects Agency (DARPA) in technology circles, and most people will think of the blue-sky research that the agency funds, like the work that spawned the Internet (see “DARPA’s Disruptive Technologies,”). Bring up Intel, and a different image comes to mind: a not very imaginative research-and-development program that cranks out one Pentium processor after another. Great stuff, but hardly research capable of producing tomorrow’s technological breakthroughs.
That could all change, as Intel’s research director David Tennenhouse is engineering a sweeping overhaul of his organization, modeled largely on DARPA. Tennenhouse, who directed DARPA’s Information Technology Office for three years before joining Intel in late 1999, says the problem is straightforward: although Intel will shell out more than $4 billion this year for R&D-ranking it among industry’s top spenders-the company rarely ventures off the familiar semiconductor road map into emerging areas like ubiquitous computing, wireless networking and biological computing. But such “disruptive research,” Tennenhouse says, is “the research that’s going to lead to new business for Intel or open up areas that are going to jar the road map.”
Tennenhouse bills Intel’s new research structure as a DARPA-like “virtual laboratory.” The company will follow the agency’s lead by using a small cadre of program managers to identify and fund projects-inside the company and out-that fit Intel’s long-term strategy but are beyond the scope of its existing business lines and research. At the same time, Tennenhouse plans to open six to eight small “lablets” near top universities; the first three will be running by this fall.
Intel’s modeling of a significant portion of its research-which might eventually total more than $100 million per year-after a government agency appears unique in business, says Harvard Business School’s Henry Chesbrough, an expert on industrial R&D. The changes, he notes, illustrate the need for companies to balance the pressure to improve existing products with the desire to hit a few home runs. “Every company has to learn how to access the wealth of ideas that are distributed outside its own four walls,” as well as those inside, says Chesbrough.
Tennenhouse spent more than a year studying Intel’s R&D structure before he began implementing the new plan last February. The company employs about 6,000 R&D people, almost all in business-division labs. Intel also sponsors some 360 university projects, including several disruptive studies. Tennenhouse didn’t want to upset these efforts; he wanted to enhance them and, especially in the case of the disruptive projects, make them part of a more formal long-term strategy. What he didn’t want to do was create a separate central research lab like those at IBM, say, or Microsoft.
The answer was to create a small group-fewer than 20 people-to evaluate, fund and oversee the additional disruptive studies he felt would be vital to long-term growth. These efforts can take place in Intel’s existing labs or in universities and nonprofit research organizations, in close conjunction with Intel scientists. If and when they mature, the efforts will be brought into the main R&D pipeline.
Tennenhouse identified five “sectors” for Intel to explore: microelectromechanical systems (MEMS), distributed systems, biotechnology, statistics and machine vision. Sector directors were charged with developing strategic plans in their areas and working with researchers to develop projects that fit those plans. Projects that make it through an approval process led by Tennenhouse will receive $2 million to $3 million a year for two to four years. In contrast to the vast scale of Intel’s conventional semiconductor research, which can involve hundreds of people on a single effort, the ideal disruptive-project size is probably five or six people, says Tennenhouse. “Most good research gets done at that size.”
Another principle guiding Tennenhouse’s vision is that some of the sponsored projects originate at Intel. Big firms tend to expect disruptive ideas to come from outside the box-and outside their walls. Tennenhouse, though, thinks the opportunity to work on disruptive projects will be a creative spark for current employees-and could even become a great recruiting and retention tool.
Many of the initial efforts funded, in fact, are taking place in-house. One is Roy Want’s “Ubiquity” project. The idea is that in the future people will carry “personal servers” through which they issue commands or make requests. But rather than harbor displays and do all the computing themselves, the devices will tap into local computing infrastructure. Say you want to review a PowerPoint presentation while on the road, Want says. Your device would relay the request wirelessly to the local network, and the page would be shown on the nearest display-a hotel-room television or office monitor. Before Intel, Want was at Xerox’s Palo Alto Research Center, which supports many such disruptive projects. But he says the Intel program is unlike anything at PARC in that his work is now done in close association with a business unit. At PARC, he says, far-out efforts “ran free,” with no connection to Xerox businesses.
In parallel with in-house efforts, Intel will step up its funding of disruptive projects in universities. But Tennenhouse is worried that the focus of university computer science researchers has become too short term-so he hopes the new lablets will become a vehicle for encouraging longer-term efforts. “We really do want them [looking] farther ahead,” he says.
Each lablet, which will house 20 to 30 researchers, will help Intel link up with a professor whose work fits with the firm’s strategic plans. The researcher will take a leave of absence, maybe two years, to get the lab started. “It’s not unusual for companies to establish research labs adjacent to major universities,” says Ed Lazowska, chair of the department of computer science and engineering at the University of Washington, near where the first lablet started this July. “What’s special, though, is that intimate collaboration with the neighboring university. We’re going to have several dozen new researchers located adjacent to our campus, whose mission is to collaborate with us.”
A lot of ifs surround Intel’s new structure. Can the lablets, for instance, build enough critical mass to stand on their own in a large organization? And at Intel, admits Tennenhouse, the idea of starting disruptive research in business-unit labs has met with resistance, because it means taking top researchers off vital road map work-or possibly diluting the company’s focus on its core business. Tennenhouse figures it will take at least five years to determine if the new model is working-and probably more. And then, even if some great projects make it “downstream” into the main R&D fold, he’ll have another worry: “The problem is, [if] the really good or great people take their projects downstream, that can leave you with people that are pretty good but not great.” That, he says, would be a sure way for the new effort to wither.
The dark secret behind those cute AI-generated animal images
Google Brain has revealed its own image-making AI, called Imagen. But don't expect to see anything that isn't wholesome.
The hype around DeepMind’s new AI model misses what’s actually cool about it
Some worry that the chatter about these tools is doing the whole field a disservice.
The walls are closing in on Clearview AI
The controversial face recognition company was just fined $10 million for scraping UK faces from the web. That might not be the end of it.
This horse-riding astronaut is a milestone in AI’s journey to make sense of the world
OpenAI’s latest picture-making AI is amazing—but raises questions about what we mean by intelligence.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.