Robots: Viral Success, but Practical Challenges Remain

Turns out, Skynet might have been an under achieving bunch of pikers. At least that is the impression, if the developments in robot technology from Boston Dynamic and others are to be believed in their latest batch of viral videos.

CNBC:

Boston Dynamics, a Softbank-owned robotics firm, released new videos on Thursday showing off some new robot tricks.

The company’s Atlas bipedal robot can now run like Forrest Gump and even jump over a log in its way. Last year we saw Atlas do a backflip.

The infamous door-opening robot, SpotMini, is getting even creepier. In the latest video, SpotMini is autonomously navigating around a lab, avoiding obstacles and can walk up and down stairs.

The videos have become very popular:

[youtube https://www.youtube.com/watch?v=vjSohj-Iclc&w=560&h=315]

[youtube https://www.youtube.com/watch?v=Ve9kWX_KXus&w=560&h=315]

But do the video clips, impressive as they are, reflect the whole truth?

Sort of, according to Wired UK:

Boston Dynamics causes a viral sensation every time it posts a new video of one of its robots moving around the lab. It’s the first sign of the inevitable robo-pocalypse, guys!. But with a little advice from some experts, you can begin to separate the hype from the facts and appreciate Boston Dynamics’s work (and your own humanity) better.

If you want to act like a robotics expert when viewing one of these videos, one of the first things you should do is be critical about how Boston Dynamics, a private company rather than an academic entity, doesn’t publish enough of its findings. This makes it hard to know what’s really going on inside the robots. “We have an idea about what approaches they are using” says Ioannis Havoutis, a researcher in robotics focusing on leg locomotion at the Oxford Robotics Institute, “but apart from a few papers, we can only guess what they are doing.”
Know the margins of error

Once the complaining’s out of the way, start by understanding the calculations and margins involved in the robots’ activities. The knack to the antics of Boston Dynamics’ robots is that they have a larger margin of error than most robots are given.

“Boston Dynamics do not worry about sub-millimetre accuracy, they worry about the functional accuracy,” says Thrishantha Nanayakkara, reader in design engineering and robotics at Imperial College London. “[Atlas] is metastable, so it’s stable most of the time. There is a probability that it can go wrong, and they take that chance. Most robots we know in the industry don’t take that chance.” Being metastable means Atlas has to balance itself to stay upright, just like a human.
But even Atlas’ backflip only requires “a very crude calculation to make the jump” he continues. “Then when it lands, it makes the corrections. It doesn’t have to be perfect, just good enough.”

Meanwhile in the real world, integrating robots into high stakes situations is still a frustrating challenge, such as this example from TechCrunch:

As the Asahi Shimbun reported yesterday, members of Japan’s Nuclear Regulation Authority are now urging plant operators Tokyo Electric Power Company to find new technology and methods to aid in the cleanup. Robots keep getting fried on their missions, literally from radiation damage, or stranded on-site wasting precious money and time.

The implication is that, perhaps, the clean up will move faster if Tepco’s energy and the government’s money is redirected to chemistry, biology, and so-called “safe containment,” building some sort of structure around Fukushima Daiichi like the “sarcophagus” around Chernobyl. Or perhaps humans need to trust AI to move robots through some of their tasks. All of the robots deployed in the cleanup effort have been remote-operated by humans, so far. The government watchdog’s critical comments followed the latest robo-fail revealed by Tepco.

On March 23 the company said it had attempted to send a survey robot into a containment vessel to find fuel debris, information it needs to decommission the plant. But the PMORPH survey robot, developed by Hitachi-GE Nuclear Energy and the International Research Institute for Nuclear Decommissioning (IRID), couldn’t get its cameras to the predetermined location. As a result, it only sent back a partial report.

Just one month earlier, Tepco aborted a mission using a Toshiba “scorpion” robot that was built to scramble over rubble, capture images and data inside the plant’s facilities. The robot could tolerate up to 1,000 sieverts of radiation. And yet, it had trouble within the hostile environs of the number 2 reactor where it was dispatched.

These followed a string of earlier robot losses at the plant going back to the Quince 1, the first robot to enter the facility after the disaster. Developed by the Chiba Institute of Technology, the International Rescue System Institute, and Tohoku University in Japan, Quince went into the power plant’s reactor 2 building where it measured radiation levels, collected dust samples and video footage. It ran several missions but eventually disconnected from its communications cable and got stranded within the building.

It’s not like anyone thought it would be easy to make robots capable of finding and retrieving molten nuclear fuel, or decommissioning and decontaminating a nuclear power plant. Japanese researchers have been trying to create robots with these capabilities since the 80s, as Timothy Hornyak wrote in the journal Science last year. Robots remain incredibly tantalizing technology.

So while impressive, it is fair to view these viral clips as more marketing and tease than scientific revolution. And your fears of a Skynet-like robo-pocalyse? From Popular Science:

When we put aside fantasies like foom, digital megalomania, instant omniscience, and perfect control of every molecule in the universe, artificial intelligence is like any other technology. It is developed incrementally, designed to satisfy multiple conditions, tested before it is implemented, and constantly tweaked for efficacy and safety. As AI expert Stuart Russell puts it: “No one in civil engineering talks about ‘building bridges that don’t fall down.’ They just call it ‘building bridges.’” Likewise, he notes, AI that is beneficial rather than ­dangerous is simply AI.

Artificial intelligence, to be sure, poses the more mundane ­challenge of what to do about the people whose jobs are eliminated by automation. But the jobs won’t be eliminated that quickly. The observation of a 1965 report from NASA still holds: “Man is the lowest-cost, 150-pound, nonlinear, all-purpose computer system that can be mass-produced by unskilled labor.” Driving a car is an easier engineering problem than unloading a dishwasher, running an errand, or changing a diaper, and at the time of this writing, we’re still not ready to loose self-driving cars on city streets.

Until the day battalions of robots are inoculating children and building schools in the developing world, or for that matter, building infrastructure and caring for the aged in ours, there will be plenty of work to be done.

Welcome to the brave new world, which as it turns out is still a work in progress.

What say you? Login and Comment.
New to Ordinary Times? Welcome! Start here.