Erico Guizzo, IEEE Spectrum.com
—A U.S. senator has cited three robotics projects as examples of “wasteful” research that lack useful applications and shouldn’t have received government funding.
In a recent report, Senator Tom Coburn of Oklahoma takes aim at the National Science Foundation, the premier source of funding for science and engineering in the United States, raising questions about the agency’s management and priorities. In one section of the report, Coburn criticizes the NSF for squandering “millions of dollars on wasteful projects,” including three that involve robots.
“A dollar lost to mismanagement, fraud, inefficiency, or a dumb project is a dollar that could have advanced scientific discovery,” the report says.
Coburn didn’t give the roboticists a chance to respond, so I reached out to the three groups — from the University of California, Berkeley; University of California, Davis; and Rowan University, in Glassboro, N.J. — to hear their side.
Of course, they aren’t exactly thrilled to see their work “featured” in the report. One scientist quipped that Coburn has just sparked a robot uprising. Picture hordes of bots descending on Washington, D.C. to show the senator who’s wasteful by using him as cookie dough.
The researchers say they welcome scrutiny and agree that there are many improvements the NSF could make. But they argue that the Coburn report evaluated their projects superficially and out of context.
One of the projects deemed questionable involves a PR2 robot, made by Silicon Valley firm Willow Garage and by far one of the most capable robotic systems in existence. Berkeley researchers taught the PR2 how to fold towels, a demonstration that captured people’s imagination.
But apparently Coburn wasn’t impressed. His report notes that the robot cost $1.5 million and complains that it “took nearly 25 minutes to fold each towel.” [UPDATE: The report references the wrong NSF grant; this is the correct one, a $1.2 million award. And the Berkeley researchers got the robot for free.]
Here’s the “exclusive” unveiling of the report on ABC’s “Good Morning America.”
Berkeley computer science professor Pieter Abbeel, one of the researchers behind the project, told me that the towel folding experiment was just a small part of a much broader effort aimed at creating robots that can handle the complexities of real environments. Here’s what he wrote in a rebuttal:
“[I]n order to expand the use of robots beyond manufacturing the machines must be far more sophisticated in terms of their ability to deal with complexity. That’s what our work is all about. Towel folding is just a first, small step towards a new generation of robotic devices that could, for example, significantly increase the independence of elderly and sick people, protect our soldiers during combat, and a host of other applications that would revolutionize our day-to-day lives.”
Coburn also discussed the report with Neil Cavuto from Fox News. After seeing footage of the PR2 folding towels, Cavuto says: “I guess many folks would like that. But how’s the robot doing? Did it indeed fold clothes?” The senator admits he doesn’t know details about the project. “It just caught my eye,” he says.
I asked Coburn’s office for more information on how they selected the projects they thought shouldn’t have been funded. Did researchers or policy experts with relevant scientific backgrounds help Coburn prepare the report? Who are his co-authors?
Coburn “is the author of the report,” John Hart, the senator’s spokesman, told me in an e-mail. He added that the senator, who is a physician, “does have a scientific background,” in addition to a business, accounting, and public policy background. “This is a multi-dimensional discussion.”
I also asked whether Coburn and his staff contacted the researchers prior to the publication of the report to ask for more information or offer them a chance to address the criticism.
“Yes,” Hart said. “Scientists and researchers who are privileged to receive federal funds should welcome and expect questions about their work.” He added: “There are no sacred cows that should avoid examination and, if necessary, dissection.”
But all the researchers I contacted told me they never heard from Coburn’s staff. They said they were puzzled that the report relies so much on press reports rather than material with more scientific content-an approach they found a bit, well, unscientific. One researcher asked if Coburn would judge whether a patient is sick just by looking at the person’s face.
In another project criticized in the report, a UC Davis group is studying how people interact with and control their bicycles. The researchers also want to build a robotic bike.
Mont Hubbard, a professor of mechanical engineering, is working with other faculty and students to develop dynamic models that can accurately describe how people ride bikes. The goal of the project, which received a $300,000 NSF award, is to understand the design parameters that could lead to bikes that are safer and easier to control by different groups of people and for different tasks.
The researchers are using a bike equipped with sensors [photo above] and also building a robotic bicycle to identify the parameters that their models need to take into account. As it turns out, Hubbard says, we know very little about how a bike’s design affects safety, performance, and our ability to control it. In particular, we need to learn more about how the dynamics of the bike and rider affect each other.
“There’s plenty to be discovered,” Hubbard says. “Just because Senator Coburn knows how to ride a bicycle, it doesn’t mean that’s the end of it.”
He adds that increasing bicycle usage would have “health benefits, transportation benefits, environmental benefits.” Surveys show that although Americans don’t bike much, many more would if they felt bikes were safer, he said.
The third project criticized in the report was a “robot rodeo,” a three-day event that took place at a conference for computer science educators in Dallas, Texas, last year. The organizers, Jennifer Kay, a computer scientist at Rowan University, and Tom Lauwers, a robotics entrepreneur, say the goal of the event was to “introduce robot programming to the nearly 1200 educators attending the conference, and to raise awareness amongst participants of how robots could be used in their classrooms.”
They say that despite evidence that robots can be used as educational tools to excite and motivate students, only a tiny fraction of educators have ever programmed a robot or tried them in their classrooms. They told me that the event — which involved months of planning and dozens of volunteers — received only $6,283 from the NSF, a number that the Coburn report doesn’t mention. (Just for reference, that’s one-fifth of what the Senate Hair Care Revolving Fund spent last year.)
And yes, Kay and Lauwers say, the event was designed to be fun:
“Perhaps the Robot Hoedown and Rodeo was singled out because it has an intentionally eye-catching name, and because on the surface it appears ‘fun.’ Indeed in his report Senator Coburn states, ‘Videos of the event posted to YouTube suggest the effort was a source of enjoyment for observers.’ It is precisely this ‘fun’ which our program aims to associate with Computer Science education, so that our current students will choose to become the future researchers that make the kinds of transformative discoveries that improve our society and our economy.”
Coburn acknowledges that NSF grants have supported many scientific breakthroughs, but he insists that the agency could save between $1 billion and $3 billion by eliminating inefficiencies and duplication.
Among other things, he calls for the NSF to defund its social and behavioral sciences division and sharpen its focus on “truly transformative sciences with practical uses outside of academic circles and clear benefits to mankind and the world.” (Full disclosure: IEEE Spectrum has collaborated with the Directorate for Engineering of the National Science Foundation to coproduce “Robots for Real,” an award-winning special report with clear benefits to mankind and the world.)
But picking “winners” is a challenge even for experienced NSF program managers and the scientists who help the agency review its grant applications.
“In many cases, it can be difficult to identify, in advance, what kinds of research proposals might lead to transformative results,” says Dana Topousis, an NSF spokesperson. “For instance, when NSF funded a graduate research fellow in the early 1990s to study digital libraries, we couldn’t predict that that graduate student would co-found Google.”
So who knows? The next Google may very well be a robotics company founded by a pair of NSF-funded researchers. Then again, there’s only one way to find out.
Check out the rest of Spectrum’s robo-coverage here: