Dangers Of Self-Driving Cars That Rely On Remote Human Operators

Dr. Lance Eliot, AI Insider

Image for post
Image for post

[Ed. Note: For reader’s interested in Dr. Eliot’s ongoing business analyses about the advent of self-driving cars, see his online Forbes column: https://forbes.com/sites/lanceeliot/]

There are three major modes involved in remote access to a true self-driving driverless autonomous car.

Remote monitoring. The simplest mode of remote access is remote monitoring of a self-driving car.

In this case, the remote capability allows someone to know whether the car is turned on or not, or whether the car is moving or not, or whether the car has been involved possibly in a crash or not.

You might already be familiar with this kind of feature since conventional cars now have this. You can press a button and speak with a human in a faraway remote center, and tell them you are out-of-gas and they’ll call for roadside assistance. Or, if you are lonely, I suppose you can just carry on a conversation with another human, with them sitting somewhere in say Nova Scotia, while you are stuck in your car and enduring the daily bumper-to-bumper freeway traffic.

Remote non-piloting control. In this second and more advanced mode, the remotely based human can actually take over some controls of the car, but only in a very limited way.

They can for example remotely start the car.

They can unlock or lock the doors of the car.

Notice that these actions are not associated with the piloting of the car. A remote non-piloting mode is one that allows a remote human to do anything other than actually drive the car.

Being able to take action of a non-pilot nature is relatively easy to technologically make happen, and also reduces the amount of skill needed by the remote operator. The remote operator is able to do what the remote monitoring does, and has the added benefit of being able to control non-driving aspects of the car.

The range of non-piloting controls can vary quite a bit in terms of particular car models, some of which allow for more or less kinds of non-piloting controls capabilities.

Remote piloting. Finally, there is remote piloting, which is the third and most advanced mode of remote access.

In this case, the remote operator can actually drive the car. They are able to turn the steering wheel, apply the brakes, use the accelerator, and overall operate the car.

To be able to do so, the car is fitted with various sensor devices such as cameras and distance detectors. The remote operator sits in front of a console and drives the car. You’ve undoubtedly seen movies that show a remote drone operator piloting an semi-autonomous drone, often in battle scenes where they are viewing a suspect and then launch a missile to get the bad guy.

Now that I’ve laid out the three modes of remote access, we can take a look at the latest trends in self-driving cars and AI.

I’ve stated that a Level 5 self-driving car is one that can be operated by AI and do whatever a human can do.

If that’s the case, would we need a remote piloting capability in a Level 5 car?

One answer is no, we would not need a remote pilot since the AI is supposed to be able to fully pilot the car without any needed human intervention.

Some argue that well, yes, that’s the definition of a Level 5 car, but wouldn’t you feel safer to know that a human was able to remotely pilot your car, just in case the AI went haywire or fell asleep at the wheel?

I would argue that I might actually feel less safe if a remote operator took over the controls of my car.

Imagine someone that is getting paid minimum wage, and they are sitting in a remote location and driving my car, while I’m in it. Do they really have the needed driving skills? Are they licensed to drive in my particular state? Do they know the rules of the road in my area? Can they really see sufficiently via the camera all the facets needed to safely drive the car? Are they potentially going to be driving my car, while it is on the highway racing along at 65 mph, and maybe suddenly the remote operator reaches for their coffee mug and oops my car goes flying into a ditch?

You might argue that drones are being flown all the time remotely, by both highly trained operators and also by that teenage kid down the street that got one for his birthday.

I get that.

One little difference.

There’s no human inside that drone!

Of course, the drone can go awry and hit someone, but the point is that currently we aren’t allowing human-occupied drones to be piloted remotely publicly. We certainly have the needed technology to be able to remotely pilot cars, trucks, and even planes and ships that have people in them. But, we aren’t actively doing so. We still believe in the “human driver inside” aspect.

One concern about this movement toward remote piloting is that we are perhaps once again coming back to the human in the equation of driving a car.

The AI pursuit is that we are taking the human out of the equation.

We want to have cars that drive on a self-driving basis.

If we put the human back into the role of driving, this time remotely driving the car rather than driving while inside the car, aren’t we also saying that we really don’t need the AI to be able to be autonomous?

If we put in place remote piloting, maybe it reduces our urgency and determination to make self-driving cars that are good enough to really be a Level 5. Remote piloting can be perceived as a crutch and lead to less funding and attention toward the Level 5. We won’t reach our desire for truly autonomous drivable cars and trucks if we become reliant on the human-based remote pilot.

Even if you argue that the remote pilot is secondary and that the AI will be the primary driver of the car, this still raises other issues.

When will the AI handover control to the remote pilot?

When the control is handed-over, will the remote pilot know what is happening? Can the remote pilot react quickly enough?

Should the remote pilot be able to take over control from the AI? Under what circumstances?

What kind of humans should be allowed to have this kind of life-or-death capabilities and decision making?

Will there be regulation to ensure that the remote pilots of cars are properly skilled, trained, and have a track record to proof their reliability? Will this create a security hole that then would allow for hackers to take over your car and drive you to a kidnap spot or worse still aim your car into a brick wall?

For Air Traffic Controllers (ATC’s), we require all sorts of stringent conditions on their skills and duties.

Right now, the remote pilots for cars are basically whatever a car maker decides they need to be.

Suppose we took the same attitude toward air traffic controllers. I don’t think any of us would feel safe going up in a plane.

This whole topic of the remote pilot labor force is still in its infancy.

Mark my words, if we continue along this path, we’ll someday maybe have as many remote human pilots of cars as we did when we had phone operators. I say this because suppose we have millions of cars on the road that are able to be piloted by a remote car operator.

Think about how many human pilots you would need to staff for this job. You need to have enough of them to be able to instantly step into driving a car, of which any of those millions of cars on the road might need at any moment. A car that says to you, “we’re sorry but all operators are busy,” would not be very pleasant when your car is rounding a blind curve and the AI has handed the controls over to the remote center.

In short, this notion of a remote piloting capability sounds good at first glance, but it raises enormous questions about trust.

Would you trust your life to a remote human that you don’t know and aren’t sure that they can pilot your car and that maybe they can’t see what your car sees?

If the human remote operator suddenly has problems with the camera in your car, and it is all fogged up due to weather, what happens then. Some might also perceive this remote access as a Big Brother kind of aspect, whereby your privacy is being invaded by someone remote. For the remote monitoring, most people that have cars with this feature are not even aware of the possibilities of being remotely spied on. It can happen.

For remote piloting, some like the idea because then the police can take over a car being driven dangerously by crazy criminal that is leading a harrowing car chase. Yes, that’s true, but the police could also possibly take over your car, even if you are completely innocent. It’s a dual-edged sword.

All in all, we need to be on our toes about the remote piloting trend. It does offer a potential safeguard for the advent of self-driving cars, but it is not a silver bullet.

By the way, if you are an Uber or Lyft driver, you might want to start playing video games involving driving cars, because soon you might be trading-in your actual car to become a remote pilot for self-driving cars.

Just think, no need to dress up anymore, and you can comfortably do your driving while in your pajamas at home.

For free podcast of this story, visit: http://ai-selfdriving-cars.libsyn.com/website

The podcasts are also available on Spotify, iTunes, iHeartRadio, etc.

More info about AI self-driving cars, see: www.ai-selfdriving-cars.guru

To follow Lance Eliot on Twitter: https://twitter.com/@LanceEliot

For his Forbes.com blog, see: https://forbes.com/sites/lanceeliot/

For his Medium blog, see: https://medium.com/@lance.eliot

For Dr. Eliot’s books, see: https://www.amazon.com/author/lanceeliot

Copyright © 2019 Dr. Lance B. Eliot

Written by

Dr. Lance B. Eliot is a renowned global expert on AI, Stanford Fellow at Stanford University, was a professor at USC, headed an AI Lab, top exec at a major VC.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store