These bots aren't yet ready for prime time.
This was the year we were supposed to get hoverboards and flying cars. Instead we got more robots that fall over.
Even though they were all over the news, television, and cinema screens this year, robots have not really entered the mainstream yet. They may build our cars and clean some of our floors, but for the most part, robots remain tucked away in research facilities and space operas.
Quartz compiled some of the biggest robot stories of the year to show why robots aren’t yet ready for prime time, but also hint at a robot future that’s likely just around the corner.
The DARPA Robotics Challenge
Inspired by the Fukushima Daiichi nuclear disaster of 2011, the Pentagon’s research division, DARPA, wanted to host a competition to create robots that could operate in situations generally dangerous to humans. The competition took place over the summer, with robots from universities and institutions from around the world trying to complete tasks like walking through doors, driving cars, turning valves, and a range of other things that needed to be done at the Fukushima nuclear plant.
Although the goal of the competition was a grand one—to spur development of robots that could one day save our lives—the reality was far more comical.
Many of the robots struggled to complete most of the tasks without falling over like drunk humans. They fell out of cars, down stairs, even when trying to turn a knob. And when they did complete a task, they often spent ages considering the job in front of them before acting. If they had been working against a clock, they most likely would’ve doomed any humans they were trying to save.
The winning team, Team Kaist of South Korea, took about 45 minutes to complete a course that probably would’ve taken a human, under normal circumstances, a few minutes. But the competition, with its prat-falling and slow-moving robots, should still inspire some optimism about the future of robotics: In 2004, the agency hosted a competition for self-driving cars in which no entrant completed the course. Eight years later, we have driverless cars on the roads of California, and soon to be many more places.
Who knows—in eight years, maybe we’ll have our own personal robots at home, as more and more money gets pumped into research. Gill Pratt, who coordinated the robotics challenge, has since left to oversee a new billion-dollar venture for Toyota, which has been working on self-driving cars and assertive care robots for years.
Under Pratt’s tenure, it’s possible the Japanese company may soon produce robots to drive us places, care for us in hospitals, or even in our own homes. Hopefully at that point they’ll be able to open a door without falling over.
This is just a Japanese robot that feeds you tomatoes while you run. Not much else to say about this one. Japan is an odd place.
I was very close to not including this on the list, since it’s barely a robot. It’s basically an Internet-connected garbage can that two researchers in Canada built, seemingly to test the limits of our humanity. The bot was left to hitchhike Canada, the Netherlands and Germany, tracking its progress and uploading photos of its travels to its website.
HitchBOT lasted all of two weeks on a similar trip in the U.S., before being decapitated in Philadelphia, the city of brotherly love. The vagabond bot was returned to Canada to meet its maker, having not really advanced robotics in any meaningful way. But it does seem to suggest that not all of us are comfortable with the prospect of having human-shaped robots in our lives.
Slack, the workplace communication tool, is loved by its million-strong user base in part for its ability to plug just about anything businesses use into the software. You can connect project management tools, file sharing services, RSS feeds, and even programmable bots that users can talk to within Slack.
But the company’s built-in bot and de facto mascot, Slackbot, is pretty useless. If you ask it, in very precise language, to remind you to do something at a specific time, it can handle that. It can say hello and will respond if someone references it in a message, but it can’t exactly hold a conversation with you:
Virtual assistants are on the rise—as is Slack—and as natural language processing and voice recognition tools get better, they’re likely to have a bigger impact on our daily lives. Slack wasn’t immediately available to comment on whether it’s planning on building out Slackbot as a more useful assistant, but Quartz may have uncovered its ulterior motive:
Facebook’s virtual assistant, built right into its Messenger app, launched for a small group of users this year. Ask M for a dinner reservation in the Mission district, and it’ll work on getting you one, seemingly like magic.
But unlike Siri or x.ai’s Amy, M is currently comprised partially of humans. Facebook has said it’s currently using humans to help sort through the requests that people are asking M when its AI programming falls short. So while it’s not entirely a robot-controlled assistant, it may well one day be—Facebook is using every interaction with M, every data point and request, to train its AI to better understand what it’s being asked, and how to complete the request.
It’s a slow and methodical approach to building out an AI system, but as Popular Science has argued, could one day produce computers that “can learn well enough to understand the world around them.” Assuming Facebook stays with the research for the long run, that is.
Every other virtual assistant
Apple’s Siri, Google’s Now, Microsoft’s Cortana, and x.ai’s bizarrely named Amy Ingram are all wonderful ideas that don’t quite execute how we’d like. Whether it’s the creepiness in having a piece of software try to hold a conversation with us, or the fact that they can never seem to fully understand us, or even find what we’re after if they do, all of them fall slightly short of the mark.
We often need to contort our voices just to get them to understand us. As voice recognition and language understanding software gets better—many companies, like Nuance, SoundHound, and IBM are working on it, as are many universities—it’s likely that virtual assistants will become smarter, and might even become the default way that we search for information on the web. But they’re still a galaxy away from the voice-activated computers in our science fiction.
While I’ll reserve judgement on the actual new droid from Star Wars: The Force Awakens until the film is released next week, the robotic toy version that was released to intense media fanfare back in September was not that great. For $150, Star Wars fans could get a BB-8-sylizied version of robot toy manufacturer Sphero’s rolling ball robot toy. It’s controlled by an app, much like any other, cheaper, remote-controlled toy would be.
I may have completely lost the ability to control him, but he seems to be having fun pic.twitter.com/xblZXkK9Lr— Mike Murphy (@mcwm) September 3, 2015
But it’s supposed to have additional modes that let owners program routes for the droid to follow, as well as understand voice commands and be able to patrol rooms. Some owners doubt its ability to do those things, and, in testing, I found it to be erratic and difficult to control. It also kept wanting to decapitate itself, which is never ideal.
For everyone who dreamt of sitting in a giant fighting robot as a kid, it is exactly as awesome as expected.Posted by Quartz on Sunday, July 26, 2015
A group of engineers in Oakland are working on a giant robot that will fight another giant robot from Japan. While the entertainment value of this endeavor may end up being enormous, the robot can’t really do much on its own right now, and its creators previously told Quartz they’re still working on figuring out “how not to die” while fighting.
But ridiculousness aside, MegaBots has raised over $550,000 to create a sport out of its robots, and wants to encourage a generation of kids to get interested in engineering and robotics. The team has even been hiring for jobs with titles like “Giant Fighting Robot Designer” right now.
The drone with a gun on it
A teenager in Connecticut built a drone that could fire a handgun earlier this year, and has now built one with a flamethrower on it. These serve no useful purpose, but considering one U.S. state has already legalized the ability for police to fly drones with crowd suppressants onboard, it’s worrying to think that all our policing could be done remotely via flying robots. Like Amazon Prime Air, but for justice.
Robot World Cup
Some of the best examples in recent years of robots failing spectacularly at the tasks before them have been the participants in the robot world cup, or RoboCup. These are teams of robots of various sizes and abilities that are supposed to play small games of soccer against each other.
This year’s championship was held in Hefei, China, with the goal of the competition being to promote robotics and AI research. Organizers want to be able to field a robot team to take on a group of humans by the year 2050. The competition has been running since 1997, and it seems that goal hasn’t gotten any closer—these robots are pretty terrible at being able to pass, move, and score goals—the things you generally need to be able to do to play soccer.
The breakfast bot
Simone Giertz built a robot for Motherboard that was supposed to feed her breakfast.
She built and programmed the bot herself, and the results ended up pretty similar to one of the DARPA robots, namely: This is what breakfast would look like were it served by the heavily inebriated. But her work points to something far more impressive and exciting. She bought the parts over the web, programmed it on her own, and got a robot to do something for her—albeit terribly.
Will 2016 be any better?
Today’s robots, like computers, are only as smart as the people operating them. They’re empty shells that, for the most part, just follow our orders in a rote fashion until something breaks. But many smart people are working on making robots that can solve problems on their own, that can overcome obstacles, and actually understand what we want them to do.
We’re getting to the point where robotic technology—from sensors, to computer vision, to movement actuators—is sophisticated enough to build robots that can find their own way in the world. Alphabet’s Boston Dynamics has built a robot that can run through the forest on its own, and researchers are working on robots that can learn how to walk in much the same way children do.
The things we take for granted—fine motor skills, perceiving colors, being able to understand others—are exceedingly difficult to show a robot how to do. And as Facebook’s artificial intelligence research director Yann LeCun told Popular Science, there are no eureka moments in AI research.
“This is one of the biggest, most complicated scientific challenges of our time, and not any single entity, even a big company, can solve it by itself,” LeCun said. “It has to be a collaborative effort between the entire research and development community.”
So it’s unlikely that we’re going to see fewer robots doing silly things next year—2014 was just as bad as 2015—but the future is nevertheless bright. It might be a few more years before we get our own personal C-3POs, but robots, like humans, learn from their mistakes. And as long as they keep getting up when they fall over, we may one day have robots to complete the tasks that we aren’t able to do, or don’t want to.
NEXT STORY Why Driverless Cars Are Like Elevators