With Google's Robot-Buying Binge, A Hat Tip To The Future
by: Steve Henn, NPR, March 17, 2014 3:03:00 pm
In less than a year, Google has bought more than a half-dozen robotics companies, setting the industry abuzz. But when I ask Google what it's up to with all these robots, the company won't say a thing.
"They are very careful — they haven't disclosed what they are doing," says Richard Mahoney, the director of the robotics program at SRI International, a nonprofit technology accelerator in Menlo Park, Calif. Mahoney also served on the board of Redwood Robotics, one of the companies Google bought.
"If I had information that wasn't proprietary, I would share it," he says. "But right now they are being pretty careful about what they are telling people."
Mahoney, like dozens of others in the industry, had to sign a nondisclosure agreement to do business with Google. As I poke around, talking to folks even at companies that hadn't been bought by Google, these nondisclosure agreements keep popping up.
Brian Gerkey, CEO of the nonprofit Open Source Robotics Foundation, says he is under such an agreement with Google. "But even under an NDA, [Google] won't tell us anything," he says.
When I ask Melonee Wise, CEO of Unbounded Robotics, if Google had made a bid to buy her company, she says, laughing, "If they had, we couldn't disclose that."
The secrecy around Google's robotic ambitions has fueled some wild speculation. Theories of what Google is working on range from the wildly ambitious — picture a fully autonomous, self-aware C-3PO — to the mundane, like factory automation.
But as executives at Google seem to be cashing in on several new trends in robotics, it's also fostering a sense in the industry that, after decades of false starts and unfulfilled promise, robots may be on the verge of becoming ubiquitous.
A Better, Cheaper Robot
Elaborate sensors, chips and lightweight batteries built into smartphones have helped to dramatically reduce the price and increase the availability of many of the parts needed to build a robot.
Paola Santana, co-founder of Matternet, describes her company's lightweight delivery drones as little more than smartphones with wings.
Even sophisticated, multipurpose industrial robots are becoming much more affordable. Unbounded Robotics' flagship bot, the URB-1, is a friendly little orange-and-white robot on wheels. When it's cruising around the office, it's about the size of R2-D2.
UBR-1 has one arm that can reach around and grab things, and it gets taller when it stops — its spine extends.
"People are always surprised by that," said Wise, one of UBR-1's creators. "They'll say, 'Oh, it gets bigger, like ET!' "
UBR-1 is designed to work right next to people in warehouses and small business, doing tasks such as sorting packages. Previous generation of robots like this cost up to $400,000. UBR-1 and its competitors cost just a tenth of that.
Open-Source Software Platforms
A decade ago, building even the simplest robot was pretty tough, says Gerkey, at the Open Source Robotics Foundation. "To get a robot to do something useful, you need to be an expert in many, many different areas," he says.
Getting a bot to move around your house without crushing your dog was a multifaceted challenge. You'd need a mechanical engineer, computer vision expert and a wiz at motion planning all working together — just to get started.
But today, Gerkey says, open-source robotics software has many of these solutions baked right in. His foundation curates and distributes software that does the basics of robotics and hands it out to developers for free.
This has allowed researchers like Wise, at Unbounded Robotics, to tackle harder problems such as trying to teach bots to plug themselves in.
"Just to recognize one type of outlet in different lighting conditions was a very difficult problem," Wise says.
Roboticists approach problems like this by feeding their machines reams of data. They show the robots thousands of pictures of different electrical outlets in different lighting conditions and create software to help the machines recognize the patterns.
It turns out this is also how Google teaches its search engine to anticipate your needs and offer you results before you've finished typing. It is exactly this kind of data-driven statistical analysis that is one of Google's core strengths.
And this skill set may well be why Google is suddenly feeling so much love for so many robots. If so, Wise understands. "I feel affectionate toward all robots," she says. "There is this growing series of pictures of me basically spazzing out and hugging robots."
Apparently, Google executives seem to know the feeling.