Boston is not home to many entertainment studios, but it can lay claim to one of the most successful makers of YouTube videos: Boston Dynamics, which has racked up more than 225 million views on the site. The company’s biggest hit so far shows a humanoid robot named Atlas, a kind of 21st Century Tin Man, traipsing across a snowy landscape, picking up cardboard boxes, and being knocked down by a human — but determinedly getting back up again.
Boston Dynamics was founded in 1992 to build some of the world’s most advanced walking robots, and at that, it has succeeded. But far more people have seen the company’s bots on YouTube — or on the HBO show “Silicon Valley,” where a dog-like robot was hit by a car — than in person. That’s because for all of its history, Boston Dynamics has been a research-and-development shop building tiny quantities of extremely sophisticated robots, mostly as part of Pentagon research projects.
Everything has been crafted by hand at the company’s Waltham facility, and affordability, given the military customer base, generally hasn’t been a factor. When I asked founder and chief executive Marc Raibert about the cost of an earlier version of Atlas, all he would divulge is that it was north of $1 million, but not quite the $2 million reported by some media outlets. Raibert had become like Antonio Stradivari, but for ambulatory robots instead of violins.
But now, Raibert’s company is under new ownership — Japanese tech conglomerate SoftBank — and suddenly he has a mass-market vision. Boston Dynamics, he said, wants to build “the Android” of the industry — a robot that other software and hardware companies will build add-ons for, as they do for Android mobile phones. At a trade show in Germany last month, Raibert told the audience that after 26 years of “working on the future, the advanced stuff, now we’re trying to make some practical products.”
Two questions that haven’t yet been answered: What useful work will they do, and what will they cost?
Boston Dynamics has gone through three phases since its founding. First was life as a military contractor, often working for DARPA, the Defense Advanced Research Projects Agency. Then it was acquired in 2013 by Google. The Silicon Valley firm was buying up robotics companies as part of a new strategy overseen by Andy Rubin, a Google executive who helped created the company’s Android operating system.
“At a high level, Google wanted to make a robot that could do anything,” said Alex Broadbent, a former Boston Dynamics project manager. “You open the box, turn on the robot, and tell it to cut the grass, clean the house, or take care of grandma. If you look at Google’s products, they are easy to use and extremely versatile.”
Andrew String, another former Boston Dynamics employee, said Google wanted a robot that could sell for under $1,000. “Of course, that’s a pie-in-the-sky dream.” (String’s current employer, iRobot Corp. of Burlington, manages to do it, with its lowest-end Roomba floor cleaner selling for $300.)
Under Google’s ownership, Boston Dynamics decided not to pursue further contracts with the military. That eliminated the company’s major source of revenue. When Rubin left Google in 2014, the whole robotics initiative seemed to lose momentum.
The third phase began in 2017, when SoftBank acquired Boston Dynamics from Google. The Japanese company already sells a humanoid robot called Pepper, which functions largely as a greeter in retail environments (Pepper’s price over three years, which includes data connectivity and insurance, is about $14,000, according to the Robot Report, a trade publication.)
The first product Boston Dynamics will try to produce in significant quantities is its battery-powered SpotMini bot. SpotMini is a 60-pound, four-legged animal-like robot with a box-shaped body that moves around with surprising spryness. It can carry a payload of about 30 pounds, which could be cargo, or an add-on like a robotic arm or special sensor, and run about 90 minutes on a charge. At the CEBIT trade show in Germany, Raibert said the company plans to build about 100 SpotMinis by the end of 2018, and 1,000 a year thereafter. He posited a few possible uses for them: security, emergency response, and parcel delivery (yes, SpotMini can easily climb your front steps to get to the porch.)
SpotMini got a few million bucks worth of free publicity in March when Amazon CEO Jeff Bezos tweeted a photo of himself walking next to the bot at a California conference: “Taking my new dog for a walk,” Bezos wrote.
But while the demos and YouTube videos dazzle the rest of us, many in the robot business are deeply skeptical that there is a market for a general purpose walking robot.
“The videos are jaw-dropping, but when I start thinking about what is the business model, that’s where it gets shaky,” said Dan Kara, a longtime robotics analyst at WTWH Media.
“What do legged robot systems give you that wheeled systems don’t — other than it uses more power and it’s less robust?” Kara said. “These things fail more often because it’s complex stuff.”
Kara said Boston Dynamics’ plan may be to sell the first batches of SpotMinis to university or corporate research labs “to let them play around with it, and see what they come up with.”
“My guess is that it will be very hard for Boston Dynamics to make the leap from cool research demonstration to cost-effective product,” added Joe Jones, one of the creators of the iRobot Roomba, and now cofounder of Franklin Robotics, a Lowell startup developing a robot to weed gardens. “Research is all about the biggest bang, but the imperative for products is bang-for-the-buck.”
Raibert declined to be interviewed, but he did suggest the company will have more to say in the fall — perhaps more details about the launch of SpotMini.
The first time I ever interviewed Raibert, way back in 2001, he had already identified the challenge his company would face selling products.
Talking about the work he did when he was a professor running a lab at the Massachusetts Institute of Technology, he said, “When you work in a place like MIT, you try to do the most advanced thing — layer after layer of complex stuff. With a product, you don’t want it to be that complex. You want it to be something you can ship around in a cardboard box and [customers] turn it on and it works.”
Boston Dynamics has been successful at many things in its first quarter-century. But success in the “cardboard box” phase will be the biggest hill for SpotMini, Atlas, and the rest of the company’s bots to climb.