By Adam Dorr
The prefix tele comes from the Greek root tēle, meaning distant or far away. The ability to sense or act at any distance was an elusive dream for all of human history up until the 18th Century when it first began to yield to technological innovation. First the telegraph – writing at a distance. Then the telephone – sound at a distance. Then television. Today we lump them together as telecommunications.
Teleoperation emerged in the late 19th Century with systems for guiding weapons (namely torpedoes) and vehicles, and by the mid-20th Century the concept of remote control was widely familiar to the general public.
And finally, telepresence represents a modern incarnation of the ancient idea of possession that dates to classical antiquity in cultures around the world, wherein a distant mind (typically gods, demons, or other supernatural entities) can fully inhabit a body (often of an unwilling subject) or an inanimate object. In its technological sense, telepresence was perhaps most clearly first imagined in famed science fiction author Robert A. Heinlein’s short story “Waldo”. Since then, the notion of inhabiting robotic vessels has become a staple of science fiction, and real-world applications using virtual reality began to emerge in the 1990s.
Today, as we stand on the cusp of artificial labor (AL) replacing human labor in what will be one of the broadest and deepest technology disruptions in history, the convergence of virtual reality with humanoid robotics is poised to finally open the door to teleoperation and telepresence akin to what we have imagined in fiction for nearly a century. Indeed, teleoperation of humanoid robots could well be the killer app that virtual reality headset technology has been waiting for.
This is not to say that users will have an immersive “full dive” out-of-body experience any time soon – that will have to wait for mature brain-computer interfaces, or BCIs. But in a more crudely functional sense, the technology now exists to allow human operators to remotely control humanoid robot bodies with high-fidelity. It is important to explore the near-term implications of the new capability.
As our team at RethinkX has discussed previously, the painful truth is that AL will eventually disrupt all human labor. Virtually every “job” is in the crosshairs, and by 2045 there will be almost nothing a human being can do that a machine cannot do as well or better for a tiny fraction of the cost. And while it is true that the disruption of labor by AL, along with other disruptions of energy, transportation, food, and beyond will create new “jobs” (meaning collections of tasks and responsibilities) – AL will do those jobs too.
However, there will be a brief period of perhaps a decade or so prior to the 2040s when humanoid robots are rapidly joining the global economy across hundreds of industries as fast as we can build them, but are not yet able to independently perform some of the most challenging tasks that we humans do in our daily occupations. And it is in that interim that telepresence and teleoperation may have a significant role to play.
Robots on Wheels
We can see a clear precedent for this pattern in the teleoperation of autonomous vehicles (AVs). Waymo and other companies developing AVs use human workers to “remote in” to take over their cars when they get stuck in situations they are currently unable to handle. The need for intervention is obviously not continuous, which in turn means that one human worker can telesupport a number of AVs. Exactly how many is a closely guarded secret, but presumably the size of the fleet a single human can manage is growing over time as the vehicles become better and better drivers.
AVs are robots on wheels, limited to very narrowly circumscribed operating conditions. Humanoid robots, by contrast, will be deployed across vastly more diverse environments. In the early days, human support teleoperators (telesupporters?) may need to intervene even for relatively mundane situations where a robot simply gets stuck in an awkward position or is uncertain how to proceed because of safety concerns and so on. As time goes on, the need for this sort of superficial assistance will asymptote toward zero, but support teleoperators will still be required for solving complex problems that today require genuine human expertise. So although robots in the next few years may need “unskilled” tech support to remote in and help them if they get lost in the laundry room or get a hand stuck in something, even once those situations become vanishingly rare there will still be a decade or so in which human experts can be usefully called upon to take over a robot in particularly challenging situations.
Return of the House Call
Teleoperation and telepresence can go beyond just support for when a robot gets stuck. They also offer the opportunity to quickly and cheaply deploy a human expert into a situation. By the early 2030s, for example, humanoid robots will be starting to become a popular sight in homes and workplaces. Then when you need a consultation from an electrician or plumber or mechanic or doctor, it might be very helpful for that expert to remote into your robot instead of just talking to you over a phone or video call.
It is easy to imagine how this new service delivery pathway could support new business models, since the technology would enable any skilled person with a good VR headset to sell their expert labor to anyone else in the world with a robot. Telepresence and teleoperation become a proxy for the power of teleportation from the human expert’s perspective.
It is also easy to imagine how this capability could pose risks. Never mind the “Terminator” scenario where the robots themselves turn on their human “masters”, criminals who can forcibly take control of a humanoid robot would have an exotic new means of going places and doing things they shouldn’t. This only highlights the need for extremely robust and dependable security features, including hardware-level kill switches, built into robots from the outset.’
Moreover, telepresence raises questions about workforce participation. Should you need a work visa to provide services in a country other than your own? Again, we see some precedent for this in the teleoperation of AVs. To date, American companies have not used foreign workers to teleoperate their AVs (or at least have not admitted to doing so). But the issue has existed for several decades in the more abstract form of remote workers being able to take control of a computer while providing tech support. But would a foreign teleoperator “driver” supporting a small fleet of EVs need a work visa? And if that is not sufficiently on the nose, then what about when a foreign worker has effectively teleported across national boarders and is “walking around” one’s home or business in a robotic body? Note also that novel security issues arise here too, such as new opportunities for espionage.
Policing and Emergency Services
Speaking of security, there are obvious advantages to police, first responders, and emergency service personnel being effectively able to teleport instantly into a location to provide support – especially under hazardous circumstances. Few would object to skilled firefighters commandeering nearby robots in an emergency to save lives and property, perhaps. But should a police officer be able to execute a search warrant by taking over a robot inside someone’s home? What if the robot witnesses an adult abusing a child? Should an EMT be able to commandeer a robot without the owner’s authorization if that robot reports that its owner has become incapacitated? Should a government be able to seize control of all the robots in an area and turn them into riot police to suppress civil unrest? Should robots teleoperated by police be able to use lethal force?
On the Battlefield
Finally, on the topic of lethal force, we would be remiss not to also mention remotely controlled robotic soldiers as an obvious application of robotic telepresence. Needless to say, this too has been explored dozens of times in science fiction since Heinlein wrote “Waldo” in 1942. (Standout examples include “Avatar” and “Surrogates”, both released in 2009). On the one hand, it is discouraging to have to contemplate the use of such marvelous technology for such hideous purposes. But on the other hand, teleoperated robots that keep humans in the decision-making loop are morally preferable to fully autonomous killing machines. Unfortunately, the comparatively glacial speed at which humans think compared to AI means that any “teleoperation” of robotic soldiers is likely to be limited just to key decisions, such as whether or not to open fire.
An Urgent Conversation
Questions about robot teleoperation are thornier than one might imagine at first glance. Like any new technology, it is difficult to anticipate all of the risks, downsides, side effects, and unintended consequences in detail that teleoperation of humanoid robots pose. Unfortunately, time is short. Societies will need answers much sooner than most of us expect – certainly by the early 2030s, when humanoid robots are beginning to be deployed at scale. That means it is urgent to begin pilot project testing right now. In particular, the approaches from AI safety testing, which include red teaming, jail breaking, and other forms of adversarial testing, will need to be closely coordinated with existing governance institutions in order to ensure policies, regulations, and standards are reasonably effective from the start.