In our economy, many of the jobs most resistant to automation are those with the least economic value. Just consider the diversity of tasks, unpredictable terrains, and specialized tools that a landscaper confronts in a single day. No robot is intelligent enough to perform this $8-an-hour work.
But what about a robot remotely controlled by a low-wage foreign worker?
Hollywood has been imagining the technologies we would need. Jake Sully, the wheelchair-using protagonist in James Cameron’s Avatar, goes to work saving a distant planet via a wireless connection to a remote body. He interacts with others, learns new skills, and even gets married—all while his “real” body is lying on a slab, miles away.
Several elements of this scenario are no longer science fiction. Companies now produce and sell robots (including the VGo, iRobot’s Ava, and Willow Garage’s Texai) that allow users to navigate through a remote working environment, interacting by means of a computer screen. So far these systems have limited functionality (some dub them “Skype on wheels”), and they’ve mostly been used for high-value problems involving costly experts. InTouch Health’s RP-7, for example, was designed to let doctors remotely diagnose stroke patients, since smaller hospitals often can’t afford a neurologist on staff.
The next wave promises much more capability per dollar. VGo’s robot can’t match the RP-7’s functionality, but at $6,000, it’s already a 12th the price. What’s more, DARPA recently issued a robotic challenge involving a complex set of tasks to be performed by a semiautonomous, remote-controlled humanoid robot—driving, walking through rubble, replacing a valve.
Progress toward the “avatarization” of the economy has been limited by two technical factors that don’t involve robotics at all. They are the speed of Internet connections and the latency involved in long-distance communication. Connecting a Thai worker to a robotic avatar in Japan with enough signal fidelity to carry out nonroutine work may be more difficult than engineering a cheap robotic chassis and related control systems.
How much bandwidth is enough? A “perfect” (just like being there) connection to a robotic telepresence system must accommodate a signal of 160 megabits per second. Theoretically, too, the distance between robot and worker shouldn’t exceed 1,800 miles: any farther and the operator could get confused by the time lag as signals travel round-trip. Realistically, however, avatar workers can probably be effective janitors or doctors even if they are farther away and sensory fidelity is weaker. The VGo runs on Verizon’s 4G network, for instance, and the U.S. military’s drone-control facility in Italy is 2,700 miles from Afghanistan.
High-end users in major U.S. and European cities will reach the 160-megabits-per-second threshold between 2014 and 2015 if current trends hold. Avatar workers are not far behind. Mexico, China, Poland, and Thailand have added 26.4 million high-bandwidth Internet users in the last 12 months. These countries have relatively low labor costs and are close to more developed countries. More than half of U.S. states are within 1,800 miles of the Mexican border; if workers in the Dominican Republic are considered as well, only Alaska and the northern tip of Maine are out of range.
Telepresence means that in theory, 10, 100, or 1,000 times as many workers could compete (virtually) for the same work. No matter how bad things get in Madrid or Houston, an avatar worker somewhere else could sell his or her labor for less. The same outsourcing logic applies to many high-wage jobs that rely on physical presence and motor skills, including the work done by cardiologists and machinists.
Big targets: Low-wage workers may one day operate robots in other countries. Here, an 1,800 mile range of operation is shown for various outsourcing centers. At longer ranges, time delays would make controlling robots more difficult.
Previous waves of outsourcing should remind us: the legal, political, and social obstacles to an avatar economy may prove greater than the technical ones. How will the meaning of work change when a gardener bot is controlled by a different remote worker every day? Or when one driver supervises 50 mostly autonomous taxis? What—and how much—work will be left in areas with the highest labor and housing costs?
Outsourcing physical work would bring huge economic gains, but it would also cause problems. In contrast to Cameron’s movie, Alex Rivera’s independent film Sleep Dealers offers a bleak vision of the avatar economy: the Mexican protagonist turns to the black market for risky surgical implantation of virtual-reality “nodes” that allow him to interface with stateside worker bots.
I believe outsourcing of nonroutine labor via robotic telepresence could begin to occur on a mass scale within a decade. Let’s take the time to manage the avatar economy thoughtfully while it is still young.
Matt Beane is a doctoral student in information technology at MIT’s Sloan School of Management, where he studies the effects of robotic telepresence and artificial intelligence on work.
Hear more about robots at EmTech MIT.
September 11-14, 2018
MIT Media Lab