A Robot Vice Squad Is Something to Fear
Singapore tests out a device that spots smokers and sidewalk skateboarders, and yells at them
For a long time police robots were simple, understandable, limited in their use and, most importantly, familiar. The bomb squad machine defusing an explosive is something we've all seen in movies and television shows. And what is familiar can't be a scary robot. It's just a handy tool that we perceive doing its work in the background of life, not holding our interest.
However, robots have been getting steadily cheaper and smarter, which means (a) they get assigned new, unfamiliar jobs and (b) more police departments can acquire them. So there's a steady spate of stories about new police robots, a lot of which mention fear and anxiety.
I have mixed feelings about these. On the one hand, it is wise to be wary and vigilant about powerful technologies for police forces --- which can be brutal and oppressive, even when they don't work for dictators. On the other hand, it is not wise to be wary and vigilant about the wrong things.
As I wrote a few weeks back, I think some of the alarm about police use of Boston Dynamics' Spot Mini, a four-legged robot, was largely misplaced. Now, this week, came stories about Xavier, a camera-equipped, loudspeaker-wielding robot that seems less menacing than a cop-Spot at first glance. At second glance, though, Xavier is by far the scarier robot.
Xavier is a mechanical spy and scold. In a test last month in Singapore, two Xavier devices wandered about a neighborhood, looking for "undesirable social behaviors." As Aimee Chanthadavong reported here, those include "smoking in prohibited areas, illegal hawking, improperly parked bicycles, congregation of more than five people in line with existing social distancing measures, and motorized active mobility devices and motorcycles on footpaths." When its 360-degree-spanning cameras spotted what its AI decided were signs of such mischief, Xavier would send an alert back to its human handlers at their desk. It would then speak a message telling the scofflaw humans to cut it out. AFP reporter Catherine Lai, for instance, witnessed the machine maneuver itself up to a clutch of senior citizens watching a chess game. Then, she writes:
"Please keep one-meter distancing, please keep to five persons per group," a robotic voice blared out, as a camera on top of the machine trained its gaze on them.
This is pretty different from the way the New York City Police Department used its four-legged Spot. That device was intended to be used in special circumstances (for instance, in hostage standoffs) where human officers might be endangered and a wheeled robot couldn't make it down the stairs. When used, they were remote controlled by an officer --- which means Spot couldn't make decisions and it didn't replace a cop. Spot was also conspicuous and expensive. For all those reasons, it was ridiculous to complain (as some NYC pols did) that the robot was a tool for spying on people, or oppressing them. (A better case against Spot was that it cost a lot in a time when the city wasn't spending money on other needs.)
For all that Xavier lacks the creep factor of a robot that walks like a dog, it's much closer to robot dystopia than police-Spot ever was.
For one thing, it appears to have a good deal of autonomy. In fact, its ability to work without a human was one reason it was developed, according to Ong Ka Hing, deputy director of Robotics, Automation & Unmanned Systems at Singapore's Home Team Science and Technology Agency, as quoted in Lai's story. “The workforce is actually shrinking,” he said.) Not only does Xavier make its way around pedestrians and street furniture and cars without a human driver, but it also seems to decide which people to watch. (It's not clear to me at what point in this loop a human official takes a role --- I would guess a person, not an algorithm, decides whether to yell at people --- but it appears the robot is on its own before then.)
This means Singapore can field a bunch of Xaviers without hiring a bunch of officers to drive them. And Xavier, a conventional wheeled robot, likely costs a lot less than Spot. So it'd be much easier for a government to flood the streets with Xaviers. Then too, the robots wouldn’t stand out that much, given that a culture of aggressive surveillance and correction is already in place. Singapore is unapologetically a city of 90,000 surveillance cameras for 5.5 million people, where laws about personal conduct are enforced that elsewhere are ignored or aren't even on the books.
Xavier robots look to be a far cry from being able to shoot at people or arrest them. But in their mission and their tech, they are unexotic to the public. They merely add mobility to surveillance tools that people are already used to living with. That's just the sort of incremental change that makes a robot feel like a normal part of life.
Plenty of other robots are already on the market with more brutal uses. Consider this remote-controlled "riot system" from a division of Way Industries, a Slovak firm that makes military and other heavy equipment. Adding autonomy to this "riot system" would be scary. (Perhaps not as scary as the thought of it remote-controlled by a guy with 6 weeks of training and a grudge in his heart, but, still.) In the long run, though, I think the robots to fear are the ones like Xavier, which extend the state's power without drama or flash.
News
That's one small step toward dogginess: Speaking of Spot, good journalists have been careful to note that it's not a "robot dog," because it does not have a "mind" of its own. Spots are remote controlled. The AI savvy that it uses is in the software that makes the robot able to walk (and sit and dance) as directed.
So I was interested to see here on Space.com that researchers at NASA and the JPL are working on a version of Spot that will have a bit of autonomy. Their Spot will make some decisions (for example, what kind of terrain to avoid, and what objects in its path might interest scientists) on its own. It's a NASA project because this version, called "Au-Spot," is being developed to explore Mars. That far away, remote-control isn't an option, because it takes several minutes for a radio signal to arrive from Earth. The rovers that wander the planet now have some autonomous systems too. But Spot, being a legged machine, can go where wheeled rovers cannot.
U.S. Target-picking AI Goes Live: This week two GBU-32 bombs were dropped on a U.S. Army firing range. It was the first time the U.S. Army dropped live ordnance on a target selected by artificial intelligence, as Pat Tucker reports here. The AI involved used satellite photos and other data to choose targets from a vast area, and a vast stream of information. In an earlier exercise without live ammo, the AI proved so good at crunching data to reveal targets that the team feeding its choices to older computers ended up crashing those older machines. (That AI-driven weapons could end up fighting wars too fast for humans to follow is, in fact, one reason to be anxious about the near-future of military robots and computers.)
Robot Nail Painting Also Goes Live: Last Tuesday saw the debut of an autonomous nail-painting robot in New York City's Rockefeller Center. As Tom Maxwell reported at Input, a company called Clockwork says its device can paint a complete set of human nails in 10 minutes, thanks to 3D cameras and a powerful processor. They charge $10.