It’s surprisingly transfixing watching robots – great, lumbering, 7ft-tall humanoids – trying to walk and use power tools and drive a car.
But it’s not the only spectator sport on hand at the Darpa Robotic Challenge. There’s also spot-the-billionaire. The first one I clock is Larry Page, co-founder of Google, who walks right in front of me leading his young son by the hand. “Larry!” I say, but then hesitate. I’m not sure how this question will end. “Is Google planning to build a race of superbots that will take over the earth?” is perhaps too bold an opener. “Could I just … ?” I say but he simply smiles and walks away.
Given that Google bought multiple AI and robotics companies 18 months ago on a secret shopping spree, including a Japanese one, Schaft, which won the first round of the Darpa challenge, it’s hardly surprising that Page has turned up for what is the World Cup of cutting edge robotics. Or what Gill Pratt, the programme director at Darpa who designed the competition, calls “the Super Bowl for nerds”. Because what Google is planning to do with its robots is one of the many hot topics of the weekend. The company has “gone dark”, Will Knight of MIT Tech Review says. “Nobody knows. It’s a complete mystery. Though it’s also possible that even Google doesn’t know what to do with them yet.”
But he’s far from the only billionaire in town. Later I spot a face I half recognise: a man photographing his infant son next to a particularly space-age looking humanoid and it’s only later that someone tells me it’s Elon Musk, the Silicon Valley It-Boy, who co-founded PayPal and Tesla and is hoping to launch cut-price space travel with SpaceX. Elsewhere in the crowd there’s Travis Kalanick, founder of Uber, one of the most ambitious – and notorious – companies in Silicon Valley; its ride-sharing app has spread across the globe in a way most taxi drivers compare to a bad case of swine flu.
In fairness, Page and Musk are engineers who probably grew up dreaming of robots and there are similar father-son pairings all over the Los Angeles stadium (and a few daughters), but it’s an indication of what’s at stake here. What the next technological frontier looks like. What big business is eyeing up as the next big commercial opportunity: namely, autonomous robot technology that can operate in a human environment. Or to put it another way: Terminator. Although we’re repeatedly told that the robots are not Terminator; that they’re not going to kill us; or make us their slaves; that there is nothing to fear.
In the press briefing, Gill Pratt says the competition was inspired by the Japanese nuclear disaster at Fukushima and designed around the concept of humanitarian search and rescue. If a robot had been able to access the plant in the immediate aftermath, he says, it’s possible that it could have prevented it from spiralling out of control. And 23 teams have qualified to design that robot: a robot that can operate in our world and use our tools. The winning robot will be the one to complete eight different tasks in the fastest time.
“I know you’re going to ask, ‘Isn’t this Terminator?’” says Pratt. “They’re 200lbs and made of steel and they look scary. But the answer is they can barely stand up on their own. These things are amazingly dumb. They can barely do anything at all. Don’t be fooled by their appearance. They’re mostly puppets.”
Puppets who may one day be able to invade Syria. Because Darpa stands for the Defence Advanced Research Project Agency and it’s as shadowy as it sounds: a highly secretive arm of the US department of defence whose mission is to invent “breakthrough technology”, something it has done time and again, creating everything from GPS to the M16 assault rifle to the Arpanet, the precursor to today’s internet. Noel Sharkey, a British professor, who heads the Stop Killer Robots campaign, says he suspects “it’s an attempt to put a friendly face on the robotisation of warfare”. But any discussion of murderous new ways of killing people is not on the agenda. One of the exhibitors at the expo tells me that the companies were specifically asked not to bring any military hardware with them to what Darpa is billing as a “fun, family day out”.
The problem is that the robots do look like Terminator. And Ron Arkin, a robotics professor from the Georgia Institute of Technology, who is at the challenge because he was a judge in a competition Darpa held for high-school students, tells me about a presentation he held last year for many of the robotocists here called “How NOT to Build the Terminator”.
“Basically, I took all the tasks that Darpa had set – driving a car, smashing a hole in the wall, opening a door – and showed clips from various sci-fi films, mostly Terminator, showing robots doing exactly that.” What was the reaction? “Nervous laughter. But all these technologies are dual use and they need to realise this. There’s just this joy of discovery at the moment – let’s create it. But we have to figure out what we’ve created. And the real question is, what if we succeed? How will it be used?”
I can see what he means. From the stands of the Los Angeles County Fairground in Pomona, California, where the event is being held, I sit and watch Hong Kong University’s robot striding down the track. It looks like a man with an oversized insect’s head. Its hands resemble claw hammers. But its movement is human and, from where I’m sitting, it looks as if it’s coming to get me.
There’s a name for this phenomenon in robotics: the Uncanny Valley. The strange, unsettling response that humans have to forms that are almost-but-not-quite human. But then, there’s uncanniness all around. On the night before the challenge begins, I take a tour of the hangar-sized space where the 23 teams are making last-minute adjustments to their robots and the first one I meet is Running Man, a hulking humanoid carrying what looks like a large backpack. It’s one of seven Atlas robots, state-of-the-art droids designed by Boston Dynamics, the cutting-edge company that Google bought, and which Darpa gave as a base model to seven of the teams.
“There’s a lot of teams who don’t have the engineering expertise to build a robot, but they do have the ability to programme one,” Daniel Reyes Duran, a Spaniard on the team from the Institute for Human and Machine Cognition, or IHMC, in Florida, says. “It’s all in the control systems. And we have a really good walking algorithm.”
He looks like a classic knucklehead, I point out. Though a knucklehead that cost $2m. And I wonder aloud what robots would look like if there were a woman on any of the teams. In the far corner of the garage, I find Team NimBro Rescue from Bonn University and one of the few women around, a student called Angeliki Topalidou-Kyniazopoulou. She and her team-mates – all students – show me their robot, which looks less like a quarterback and more like it was fashioned out of an Ikea TV unit.
Their robot didn’t cost $2m. It barely cost $60,000. “We didn’t get any money from Darpa,” says Sven Behnke, the professor leading it. They are, it’s fair to say, the under-robot. The hot favourite is “Chimp”, built by Team Tartan Rescue (all the teams have the kind of names that sound as if they’ve been left over from an episode of The Apprentice) from Carnegie Mellon University in Pittsburgh. It’s a vast red steel Transformer-like droid with caterpillar tracks for feet, a multi-million dollar budget, 18 different commercial sponsors and rather high expectations. On its first run out, however, it spectacularly falls over. But then even more spectacularly, it does what looks like a press-up and then gets up again. No other robot has managed this but the Foxconn logo on its shoulder makes it hard to cheer as enthusiastically as one might.
Another of its sponsors is Amazon but then Amazon has sponsored three of the teams – including Running Man – and the company is making no secret of its robotic ambitions. I bump into a computer vision scientist from the company who talks about a robot “picking” competition that Amazon held last week, expressly designed to find the technology that will replace the workers we call humans. “It’s very a tedious, tough job,” he says. “Human pickers walk up to 11 miles a day.” I know. I did the job for a week for an article and I can’t help but feel a pang for my fellow workers. They thought it was a rubbish job too. But it was still a job.
But then, long before these robots are fighting our wars for us, it’s likely that they’ll be packing our Amazon orders, or delivering them. Several companies have designs in this area and when I interview the head of the Team Tartan Rescue, a professor of robotics called Tony Stentz, I mention that I’d heard a lot of team members had been poached by Uber. “Has that caused a problem?”
He bristles slightly. “The team is fantastic,” he says. “We have no problems there.” But was there some sort of brain drain going on in robotics? “I can’t comment on that.”
It’s only later I learn that Uber not only lured away 40 of Carnegie Mellon’s researchers, leaving one of the top robotics departments in the country in crisis, but that he’s going too.
It’s possible that June 2015 will be seen as the great inflection moment for robotics. When Darpa held its first autonomous vehicle challenge a decade ago, the idea of a self-driving car was pure science fiction. Fast forward to a third challenge in 2007, won by a team from Stanford, the head of which, Sebastian Thrun, was promptly hired away by Google to set up Google X, its highly secretive, advanced technology lab. Three years later, it announced that not only had it created a self-driving car, but it had driven it thousands of miles all over California.
Every major motor manufacturer is now developing self-driving cars. And I spot Thrun in the Google entourage. He’s officially left the company but it’s rumoured he’s advising it on what the next step forward is with robotics and AI.
“It’s super, super exciting,” he says. “It might not be as thrilling as seeing cars zipping around but the complexity of these machines is really really tremendous,” he says. Is it a great leap forward? “Without a doubt.”
And there’s no getting away from the excitement in the teams. They’ve been working flat out on these robots, coming up with completely new codes and strategies and brilliant solutions to what is an incredibly difficult and still very new discipline. “Robots are assholes,” Jerry Pratt, the leader of Team IHMC says. “They’re incredibly frustrating to work with. They hardly ever do what you want.”
But the crowd loves the falls. All weekend, humanoids and robots that look like transformers or monkeys (Nasa’s is called “RoboSimian”) teeter and then topple. There’s a moment of extraordinary pathos early on when Team Aero’s multi-million-dollar wheeled droid gets stuck in the sand. And when the Running Man has a successful round, scoring the full eight points, there’s an all too human moment of hubris, when he does a victory wave and then collapses on top of the barrier.
A blooper reel of the most spectacular falls is compulsive viewing for all the teams. And though Gill Pratt has warned us that the action will be “not quite like watching paint dry” but possibly worse — he compares it to a game of golf — it’s actually bizarrely gripping. At one point, I watch a robot fail to open a door for a good 20 minutes. And when it succeeds, I find myself cheering. When it falls, I gasp and groan. It’s a visceral reaction: humanoid disaster and humanoid triumph seems to tap into some ancient centre in my brain that I have no control of. And I’m not the only one. During the last heat, the stadium is in uproar. There are huge cheers, massive groans.
I’m rooting for Team NimBro Rescue’s Ikea television unit, but it’s a hard robot to love. Instead of a head, it has what looks like a CCTV camera ripped off the wall outside the lab. It does astonishingly well on the first day. It drops only one point and finishes in a record time. When I catch up with the German team later, they can’t quite believe it either, though ultimately it’s a single-minded Korean team, Kaist, who triumph with a robot that has the best of both worlds: legs and wheels.
At one point in the proceedings, John Markoff, a veteran technology correspondent for the New York Times, who has a book on robots coming out this summer, Machines of Loving Grace, says: “It’s not about the machines. It’s about us.” What did he mean? “We can impute human qualities on to almost anything. Look at how people treat their dogs.” And this perhaps, is what the contest is ultimately about: human-robot interaction, the next frontier.
Out in the expo area, I come across Pepper, a “social robot”, that is about to be launched as a “domestic companion” in Japan next month. It’s about 4ft high with huge eyes, programmed to track your own and the voice of an enthusiastic but possibly not very bright female cartoon character. “Can you make me a cup of tea?” I ask it. “I could try,” it says. “But I would rather be a friend.” The kind of friend who doesn’t listen to a word you say, it transpires, and busts out the disco moves at the slightest opportunity. The kids clustering around it give it hugs on command, however, and are utterly transfixed. A few stalls along, I find Nasa showcasing its Valkyrie robot, which flunked out of an earlier round of the challenge. It looks like a super-sized version of the Stig in an all-white body suit with a helmet head. But then I notice it has swellings in the chest area. Does your robot have breasts?
“It has some feminine features,” Ron Diftler, manager of the robotics lab at the Johnson Space Centre in Houston, says, “So that it’s functional but also appealing. It’s like with people. We wouldn’t be as attractive if we had neutered features. Her sibling robot has male features.”
I can’t help thinking that a robot with enormous tits is exactly the kind of robot only a male robot designer would come up with. She’s about 7ft tall with shoulders as wide as a table and what looks like a laser gun coming out of her mouth.
“We are entering an age of robots,” John Markoff says. “These machines will live among us.” But whose robots will they be? They’ll reflect the biases and values of whoever codes them and my piffling concerns about sexism are just the start of it: there’s not even any real debate about whether we should allow robots to start wars or not.
Autonomous weapon systems are inevitable, says Ron Arkin, the how-not-to-build-the Terminator professor. Because they already exist. And recently the voices warning about the potential consequences of AI have grown: Stephen Hawking, Bill Gates and Elon Musk, have publicly warned about the consequences – and dangers – of artificial intelligence.
“In Terminator, they didn’t create AI to have some sort of Terminator-like outcome,” Musk has said. “ It’s sort of like the Monty Python thing. Nobody expects the Spanish inquisition.” Nobody does expect the Spanish inquisition. Least of all, roboticists. “They’re the most recalcitrant people I know,” says Arkin. “Maybe it dampens this thrilling joy of discovery. But there’s no doubt that they’re sticking their heads in the sand.”
What Darpa’s 2015 robotic challenge has made clear is that it’s a Brave New Robot World. And since Darpa and Google and Amazon and Foxconn aren’t going to, we should probably be talking about it, human to human. Before they steal our jobs, or rescue us from nuclear meltdown, or start firing laser missiles from their enormous robot breasts.
guardian.co.uk © Guardian News and Media Limited 2010