Quando spiego il ruolo del #digitallabor nel machine learning, lo script è più o meno questo:
Io: "Dei microlavoratori *addestrano* le intelligenze artificiali in fase di training" Das Kapital: "Beh, certo..."
Io: "Ma non finisce qui: quando le IA sono ormai sul mercato, i micro-lavoratori ne *verificano* i risultati" Das Kapital: "Bof... sì, se proprio vogliamo"
Io: "E non ci scordiamo che quando la macchina non funziona (e succede spesso!), i micro-lavoratori ne *simulano* il funzionamento" Das Kapital: "Complottista! Dietrologo!"
Meno male che il Wall Street Journal, piccolo giornale di estrema sinistra, mi dà una mano a spiegare le cose per bene... ---- The Next Hot Job: Pretending to Be a Robot By Christopher Mims https://www.wsj.com/articles/the-next-hot-job-pretending-to-be-a-robot-11567... As the promise of autonomous machines lags the underlying technology, the growing need for human robot-minders could juice the remote workforce Michael Niedermayer used to fly drones for the U.S. Army and the Central Intelligence Agency, gathering real-time, life-and-death intelligence on battlefields in Iraq. Now he pilots delivery robots for a San Francisco Bay Area startup that wants to disrupt burrito delivery. Postmates, which in mid-August received a permit to operate its Serve delivery robot in San Francisco and is already testing it for food delivery in Los Angeles, employs a growing team of “pilots” to remotely oversee, and at times steer, these four-wheeled food ferries. “We will probably see a drastic increase in our workforce over the next five years,” says Postmates Chief Executive Bastian Lehmann. Michael Niedermayer, a former drone pilot now working for Postmates, said his job remotely steering delivery robots ‘felt like a great fit for my military background.’ PHOTO: TALIA HERMAN FOR THE WALL STREET JOURNAL Disrupting “last-mile” delivery—historically the domain of box trucks, bike couriers and personal vehicles—“felt like a great fit for my military background,” says Mr. Niedermayer. His story is hardly unique. Across industries, engineers are building atop work done a generation ago by designers of military drones. Whether it’s terrestrial delivery robots, flying delivery drones, office-patrolling security robots, inventory-checking robots in grocery stores or remotely piloted cars and trucks, the machines that were supposed to revolutionize everything by operating autonomously turn out to require, at the very least, humans minding them from afar. Until the techno-utopian dream of full automation comes into effect—and frankly, there’s no guarantee that will ever happen—there will be plenty of jobs for humans, just not ones their parents would recognize. Whether the humans in charge are in the same city or thousands of miles away, the proliferation of not-yet-autonomous technologies is driving a tiny but rapidly growing workforce. Real-Life RoboCop Mimi Kongnarinh started on the night shift at Cobalt Robotics, patrolling the offices mostly of other tech and some manufacturing companies. Her work would be familiar to any night watchperson from the last century—chatting up familiar employees, introducing herself to fresh faces and keeping a wary eye out for anything out of the ordinary. What made her job unusual is that she would patrol multiple offices simultaneously, and that for most of her eight- to 12-hour shift, she never left her desk in San Mateo, Calif. By remotely monitoring her largely autonomous robot charges, she was able to add a human touch when necessary—such as politely asking unfamiliar employees to badge in on the robot’s scanner—while also keeping down her employer’s labor costs. “We get a live feed of every event that happens at every site we patrol,” says Ms. Kongnarinh, who now works day shifts. “We can see if it’s a person, motion, anything like that, and we can jump in when we have to.” Nearly all companies using “autonomous” robots have to depend on what the head of Postmates’ technology skunkworks, Ali Kashani, calls the 1-to-N ratio—N being the number of robots a single human can handle. For a company like San Francisco-based Simbe Robotics—which makes five-foot-highrobots that wheel around grocery stores, taking inventory—the N can be fairly high because their robots are almost entirely autonomous. When the robots are introduced in a store, says Lauren Milliken, a field deployment engineer at the company, they must be steered by a human to build a map. Humans also have to take over on rare occasions when, for instance, there’s construction in a store or someone harasses a robot, she adds. For Flytrex, a drone delivery company currently operating in Iceland and participating in a Federal Aviation Administration trial in North Carolina, the ratio is one to one. Eventually a single person will be able to ride herd on a dozen drones, says CEO Yariv Bash, but for the next decade at least, a human must pay attention the entire time each drone is in the air. “You can prepare for 99.9% of situations, but there’s a really long tail of really weird situations, and you need a human for that,” he says. Be it air-traffic emergencies, weather anomalies or volcanic eruptions, the human must always have the ability to override the autonomous system and guide the drone back to earth. While humans might be more capable than artificial intelligence at this stage, these remote-controlled robots can, if handled incorrectly, kill people. When Lives Are at Stake “Remote control of cars is the dumbest and the most insane idea I have ever heard,” says Mary Cummings, a professor of electrical and computer engineering and the director of the Humans and Autonomy Lab at Duke University. “I’ll be the first person they should call as an expert witness when a death inevitably occurs,” she adds. Companies working with remote-controlled robots know there are risks, and try to mitigate themin a few ways. Some choose only to operate slow-moving machines in simple environments—as in Postmates’s sidewalk delivery—so that even the worst disaster isn’t all that bad. More advanced systems require “human supervisory control,” where the robot or vehicle’s onboard AI does the basic piloting but the human gives the machine navigational instructions and other feedback. Prof. Cummings says this technique is safer than actual remote operation, since safety isn’t dependent on a perfect wireless connection or a perfectly alert human operator. For every company currently working on self-driving cars, almost every state mandates they must either have a safety driver present in the vehicle or be able to control it from afar. Guidelines from the National Highway Traffic Safety Administration suggest the same. Most companies in the space have opted for something short of true teleoperation, which is dependent on an absolutely reliable and fast wireless connection as well as the skill of the human remote operator. Alphabet Inc. subsidiary Waymo’s remote operation system and another being tested by Nissan rely on humans to either confirm the vehicle’s choices when it’s unsure what to do or help it navigate around obstacles. Remote-operation software developer Phantom Auto Inc., based in Mountain View, Calif., designs human supervisory-control systems for self-driving vehicles, but the software lets humans assume full remote control if needed. Since autonomous vehicles are still quite limited in their capabilities, factors like inclement weather, construction sites or previously unmapped areas could require operators to step in. Such scenarios worry Prof. Cummings. “If there’s any teleoperation [of a car] going on in a hundred-mile radius, I want to know about it because I want to get off the road,” she says. Shai Magzimof, Phantom Auto’s co-founder and CEO, says that even when a remote driver is “directly” piloting a vehicle, it still has safety components engaged, such as object detection and avoidance. To date, there have been no accidents involving the company’s systems, he adds. Phantom Auto is betting the shift to remote operation might become an important means of employment for people who used to drive for a living. Josh Resnick, a former Israeli army drone operator, trains Phantom Auto’s remote operators and those of its customers. Training new drivers typically requires one to three weeks, he says, and the only qualification is that someone have a driver’s license and a clean driving record. The Secret Technology Helping Driverless Cars: Remote Control Even the best driverless cars struggle to recognize uneven terrain or navigate difficult weather. Startup Phantom Auto is developing technology to serve as a missing link: controlling the vehicles remotely. Photo: Jake Nicol for The Wall Street Journal Many deployments of Phantom Auto’s system involve sidewalk delivery, yard trucks and forklifts, which largely don’t travel on open roadways. In the future, many more of its customers are likely to involve trucks, passenger cars and shuttles on the open road. Those applications are in testing, says Elliot Katz, the company’s co-founder and chief business officer. Our Remote-Controlled Future There might eventually be other requirements, however. One might be a tolerance for working for a lower wage, since remote operation could allow companies to outsource driving, construction and service jobs to call centers in cheaper labor markets. Another might be a youth spent gaming. When Postmates managers interview potential delivery-robot pilots like Diana Villalobos, they ask whether or not they played videogames in their youth. “When I was a kid, my parents always said, ‘Stop playing videogames!’ But it came in handy,” she says.