There's a Simple Reason We Don't Already Have Robots in Every Home
Human lives are messy. Robots still need things to be neat. Amazon's entry into the "household robot" arena won't change that.
One reason this newsletter exists is that labs around the world are doing so much ingenious research around friendly, articulate, human-aware robots. Swept up in it all, I tend to forget that regular folks (including me) have yet to see, or want, such a machine in their homes. People love autonomous vacuum cleaners and seem open to a look at other one-use devices, like robot lawn mowers and robot weed-pullers. But these devices are robots in the same sense that a dishwasher is a robot. They follow instructions for doing a specific task that once required a person.
I don’t meant to diss the achievements involved in these devices. It is no trivial task to make a household machine that can move around humans and their stuff, and whack only weeds and not flowers or the cat. But once people get used to the mobility of handy gadgets that do one thing, I don't think they will even call them robots.
"Robot,” for most people, is a word for something that can interact with people to assist with daily-life tasks — ie, assignments that are various, overlapping, not totally predictable and highly dependent on what's going on among humans at the moment. It's a great idea for me to log on to a website to pay the mortgage today, unless my son yells out in pain from another room, in which case it's a terrible idea to keep tapping at a keyboard, unless it turns out he's just mad about missing a shot in his video game, in which case it's a good idea to get back my keyboard. This is called a dynamic and unpredictable environment. Robots have a hard time with those.
Humans, on the other hand, can’t help creating them. And they want a machine that can handle it. As would an assistant. Or even a dog.
Such a machine will have to be able to respond to spoken instructions on the fly, like "don't answer the door! Oh, wait, it's grandma, go ahead, answer it!" It will need to know that its map of the living room floor will have to accommodate shoes and backpacks and a tennis racket in the afternoon, after school lets out, but not in the morning. In other words, a robot that can do many things as needed will have to be able to understand people, both in receiving direct commands and in fitting in with the household. It will need to be sociable.
Robot-makers have known this and thus have come up with a number of fairly companionable, fairly friendly, fairly versatile robots for the home. As you’ve probably noticed, none has broken out.
Some were delayed (like Blue Frog Robotics' Buddy), some had their corporate creators die (only to be resurrected more recently), as is the case for Jibo and for Anki's robots, Cozmo and Vector. In any event, the robot equivalent of an iPhone — a machine that answers needs people didn't know they had — has yet to appear.
This week, though, Amazon decided to enter the arena, with Astro, a 2-foot-tall, 20-pound robot that leverages Amazon's already robust home-surveillance technologies.
It's appliance-white, has a screen for a face (which displays just huge circles that every neurotypical human will read as eyes), runs around on wheels, has a camera (which can periscope up to look around) and uses AI to respond to instructions and get to know the household. Which means it is a lot like Buddy and a number of other companion/assistant robots, like the machines mentioned in this roundup.
Unlike other home-robot ventures startups, Amazon has a lot of presence already in many homes, and thus experience with other devices (Alexa, Echo, its Ring spying system, its new drone that flies inside your home to spy even more thoroughly on everything) and vast troves of data on people's behavior. Maybe that will make a difference to people, so that many shell out $1,000 for an Astro (current price) or $1,500 (soon-to-be price).
But maybe not.
The not-so well-kept secret of social robots is that they aren’t that capable. They haven’t cracked the incompatibility between human fecklessness and robots’ need for neatness. Astro seems no exception. Amazon's supposed advantage here — the devices it has already produced that take spoken orders and learn your habits — might turn out to hurt Astro sales more than they help. After all, as Steve Crowe wrote earlier this week at The Robot Report, much of what Astro can do can be done by an Echo device. For the thousand bucks it would cost him to get an Astro to follow him around from room to room, offering to play music, call his grandmother or check the weather, Crowe could instead buy 32 Echo Show units and blanket his home with those.
I read that and thought, yes, but … embodiment! A thing that can move about in the room — to play with a child, or gently suggest to an older person that he ought to get up and take a walk. That's the robot promise, isn't it? An intelligence that isn't screen-bound, that occupies the same spaces we do?
Trouble is, Astro and his little predecessors from other companies don't have bodies that can do that. Not much, anyway. "Until Astro can climb stairs and open doors, it’s a (very) limited toy that solves few new problems," as one of Crowe's readers put it. Crowe noted other limitations: "It doesn’t have arms to pick up and transport items. It has limited storage capacity. It can’t go outside. It won’t deter an intruder who’s already in a house. And its functionality is limited for users who don’t subscribe to other Amazon smart home products such as Ring."
Still, Astro's appearance is significant, and not just because it makes life hard for robots that look much like Astro but aren't backed by the global Goliath that is Amazon. I think Washington Post reporters Heather Kelly, Chris Velazco, Jay Greene and Tatum Hunter got to the heart of it when they described Amazon's fall product announcement show as "an opportunity to slowly increase customers’ tolerance for what’s normal." Whether or not we all go out and get little white wheeled big-eyed companion robots, their adoption by Amazon ensures that people won't find them to be as exotic or eccentric as they seemed last week.
After all, it's not that long ago that a small flying spy robot, designed to fly inside your house, would have seemed dystopian. In 2021 Amazon will sell them (to a "lucky" few) for $249.
You know where you really do need a robot? A flooded, burning, collapsing tunnel filled with toxic gas
In last week's newsletter I mentioned a DARPA contest for autonomous robots designed to help soldiers or first responders in very difficult settings, like a flooded tunnel or a burning subway station or a deep cave full of people who need rescuing.
The final round of the multi-year competition took place Friday in a cave in Kentucky — a place that "blended aspects of cave, tunnel and urban subways and was often unlit, filled with smoke or on a steep slope," according to DARPA. The contest, in which robots had one hour to find various objects and identify them in the murk, was a thriller: The two leading contenders ended up in a tie.
However, the team from Oxford University's Robotics Institute finished sooner, and so they won the $2 million top prize. Their solution, appropriately for a hellish environment, was called CERBERUS (which stands for CollaborativE walking & flying RoBots for autonomous ExploRation in Underground Settings, because I guess in robotics you can pick any letter from any word to form your acronym).
Their approach used four ANYmal four-legged robots, plus a flying drone, to find the objects. The other high-scoring team, from Australia's national science agency, thus took second place, and $1 million. Full results, both for physical robots and a parallel contest for simulated ones, are here.