Skip to main content

Boston Dynamics' robot dog is good enough to show how terrible home robots are right now

Boston Dynamics' robot dog is good enough to show how terrible home robots are right now

Share this story

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

Boston Dynamics

I'm always amazed by Boston Dynamics. One anecdote I like to share about the company is how the next generation Atlas has 3D printed limbs, with embedded "veins" in the structure to route its hydraulics, which keeps the robot strong while making it vastly lighter and in some sense simpler than the original Atlas. That's just some impressive engineering work on all levels.

Of course, while Boston Dynamics is pretty much at the state of the art when it comes to humanoids, it's probably best known for its terrifying line of dog-inspired, gas-powered robots like BigDog and AlphaDog. It's also famous for kicking those robotic animals, in an amazing technical display of realtime balance on the part of the robot, and an amazing level of this-can't-look-like-anything-other-than-animal-cruelty on the part of the robot's handlers.

Right now the company is owned by Alphabet, but, according to Bloomberg, Alphabet is looking to sell Boston Dynamics because it doesn't have any products that can actually be released in the near term.

Yesterday we got to see Boston Dynamics' latest robot, SpotMini, and if it weren't for the telltale kitten mittens-style gait, you'd think it was from another company. Instead of the battlefield, or the sort of emergency situations the DARPA Robotics Challenge was targeting, SpotMini is clearly designed for the home: it's electric, lightweight, and almost cute.

But there are a bunch of reasons we don't have useful robots in our homes yet. We have Alexa. We have Alexa-like things on wheels. We even have tele-operated video conferencing screens on wheels. But robots that can actually do things in the physical world other than vacuum are scarce.

Here are some of the problems a home poses, most of which SpotMini is engaging with:

1. Homes have doors and stairs and carpets and all sorts of obstacles.

Opening a door is really difficult for robots. Like, just dang hard. You need precise articulation, good balance, and good movement. It's much easier to solve with a stable robot on a wheeled platform, but then that robot becomes useless in almost any home that has steps, differing levels, or even thick carpet — which is most of them.

SpotMini can climb steps and negotiate uneven terrain, so that's basically solved. And while Boston Dynamics doesn't demonstrate any door opening in its video, the arm attached to the top seems strong enough and well enough articulated to manage the feat.

2. Homes have children in them.

Weight isn't just important for robots getting better battery life, or not breaking when they fall over. It's important because of what they might fall on. Like a child, for instance. SpotMini is 65 pounds with an arm — about the weight of a golden retriever. But it’s not exactly soft. That would hurt if it fell on you. Or, and I hate to keep bringing this up, a child. Boston Dynamics will probably have to figure out a way to make an even lighter robot, or to make it so aware of its surroundings that it will have zero chance of harming a human. Both are tall tasks.

3. Homes are full of fragile objects that we want robots to bring to us.

I think the first robot that can reliably get something from the kitchen and bring it to you will be an amazing seller, and basically singlehandedly usher in the era of home robotics. SpotMini is obviously close to achieving this all-important feat. The YouTube video ends with SpotMini's abortive attempt to deliver a Diet Coke. It's an impressive outtake, and props to Boston Dynamics for including it — direct human-robot interaction is super hard, and often kind of dangerous. Things don't end well for the Diet Coke, however, or SpotMini.

But think of how difficult this exchange is. SpotMini has to identify the correct human (not, for instance, the camera man), find a free hand on the human, then put a Diet Coke in this hand and release the Diet Coke at the precise moment it can know the human has a good grip. This is beyond just a mechanical task, it's a how-do-humans-think task. It's something even I'm not very good at, if I'm being honest.

Even just holding something securely without crushing it is still at the cutting edge of robotics. Do you notice in the video how when SpotMini is doing the dishes (kudos BTW, that's amazing), it crinkles the pop can when it picks it up? Most humans know how to not crush things accidentally when they grab them. But there's a lot of subtle intelligence and mechanics behind that process.

4. Homes don't have military budgets.

This, sadly, makes all these other problems much harder. You can't put the best sensors, best actuators, best components, and best intelligence into a home robot. Most of Boston Dynamics' robots cost much more than a luxury car — many of them were developed with multi-million dollar government funds. I have no idea what SpotMini would cost if it was produced at a mass scale, but probably it would still be near a car-level expense.

The hope is that better material science, and better machine intelligence, and better something else will be the breakthrough home robotics need. Until then, Boston Dynamics' tech demos will always be thrilling, but still far from practical.

5. Homes are where cute things live, so we can love them most effectively.

SpotMini is creepy. Not as creepy as, say, BigDog. But still kind of creepy, especially in the way it walks. Can it ever learn how to chill out and be adorable? Let’s hope.

The robot butler of the future