Self-Driving Cars Are Coming — So Are the Questions Nobody's Answered

The case for self-driving cars

Self-driving cars are here, and they’re already saving lives. Tesla gets most of the headlines, but it’s far from alone: Volvo, Mercedes-Benz, Audi, Nissan, Intel and Volkswagen are all racing to build autonomous vehicles of their own.

The safety case is hard to argue with. In 2016 alone, more than 37,000 people died in car accidents in the United States. Most of the excitement around autonomous cars is about wiping out human error: speeding, drunk driving, texting behind the wheel.

A self-driving car can save lives in two ways. It can avoid the crash altogether, thanks to the right sensors and programming, or it can keep the people inside safer when one happens anyway.

Author Tom Vanderbilt explored the ethics of all this through both facts and personal experience, even driving his parents around with “the future” to hear how they felt about it.

But before we ask will it crash?, there’s a quieter question worth asking: what does this car know about me, and who else gets to see it?

Nine ways your car could be watching you

Here’s where it gets uncomfortable. The smarter a car gets, the more it quietly collects, and it’s worth knowing where all that data can end up.

Connectivity. Cars are constantly talking to each other and to their manufacturers, sharing data to coordinate routes and get you from home to work efficiently. The big question is whether manufacturers are being transparent about how that data is used. So little about privacy is shared with consumers that it’s hard to tell.
Insurance companies. Smarter cars mean smarter insurance packages. That sounds great, until you remember that insurers would also see how fast you drive and where you go on a regular basis. Discounts for “safe” driving could mean a lot of data being collected about you, possibly without you even knowing it.
Law enforcement. Police will eventually be able to access your car’s black box. Most people would agree they shouldn’t, unless they’re pursuing a crime that has already happened.
Hackers. If there’s one major risk, it’s this one. Any time there’s a chance of a cybersecurity breach, you can rest assured hackers will be all over it in short order.
Emergency situations. If the worst happens, nothing stops your car from locking the doors on you, or even driving itself into a lake.
Biometrics. Your fingerprints and retina scans might unlock your car, and catch law enforcement’s interest. Worse, they could become “user friendly” to criminals who find a way into these systems.
Your car knows where you are. Where you are, where you’ve been, where you’re going. Government agencies could get access to all of it without a warrant.
Your car knows what you’re doing. Sensors can record who’s in the car, who opens the door and when, and facial recognition can know exactly who’s inside at all times.
Public wifi. In-car wifi could be used to track your location and driving patterns. Could anyone else access that information?

None of this means self-driving cars are a bad bet. It just means the privacy conversation deserves as much attention as the safety one.

Who’s betting big

The money was following the hype. Ford’s 2016 executive vice president, Mark Fields, said the company expected to sell one million autonomous cars by 2023. Audi and Mercedes-Benz estimated that 40 percent of the vehicles sold in the United States would be self-driving by 2025. And Tesla was already hinting that its cars might not need drivers at all.

Worried about jobs? The forecasts told a more hopeful story. Deloitte’s mobility forecast expected businesses to hire operators to run whole fleets of these cars, and people will still be needed to maintain them. Strategy Analytics went further, projecting 700,000 automotive-service jobs created by advanced driver-assistance systems. If you were looking for a job in the automotive field, it was a good time to start looking.

How the AI actually drives

Self-driving cars aren’t a new idea. They were first introduced over 100 years ago, and their popularity has risen and fallen many times since. What’s new is AI, and it’s taking them to a whole new level.

Under the hood, it comes down to three things: sensors (cameras, radar), software (the AI) and actuators (the steering wheel). The AI takes in everything the sensors see, combines it with GPS and mapping data, and decides what to do in that moment.

As the car moves, it keeps updating its internal map of the road and reacts to the world in real time. If it starts raining, it slows down to avoid hydroplaning.

The hard part is everyone else. Roads are shared with pedestrians and other drivers, and back then, nobody knew yet how a self-driving car would read what they were about to do. Full autonomy depends on cracking exactly that.

Not everyone agrees on the ethics

When a crash is unavoidable, who should a self-driving car protect first: the people inside it, or the people in its path? The “right” answer depends a lot on where you’re asking.

A 2018 study in Nature surveyed more than 2 million people worldwide on exactly that question, part of what researchers called the Moral Machine experiment. The results split along cultural lines: some regions leaned toward saving the greatest number of people regardless of who they were, others leaned toward protecting passengers over pedestrians, or the young over the elderly. There was no single answer everyone agreed on.

That’s a real problem for manufacturers trying to sell the same car in different countries. Someone has to program what the car prioritizes in a no-good-outcome situation, and the “right” choice might not travel well across borders.

Regulation varies just as much. Japan, for one, has been described as a leader on the manufacturing side, with automakers pushing advanced autonomous features under comparatively permissive rules for deploying them. Elsewhere, concerns about hacking and software spoofing have led to much stricter limits on what self-driving cars are allowed to do.

Either way, this isn’t just an engineering problem. It’s an ethics problem with no universal answer, and it’s not one the industry can code its way around.

Not just cars: flying cars and self-flying helicopters

If self-driving cars sound futuristic, wait until we leave the road entirely.

Flying cars used to be pure science fiction. Now they’re available to buy, provided you have over $200,000 lying around. They’re complicated too, taking around seven years of development before they’re operational. But once they’re up, some can fly for hundreds of kilometers at 200 mph (320 kph) or more, and when they touch down, they can easily turn back into driving cars.

Google had its own spin on the idea: a detachable passenger seat with rotors that can attach to any helicopter, built to get people from point A to point B as quickly as possible. The catch is that helicopters aren’t designed for those speeds. The fastest one ever flown, the X2, topped out at 279 mph in 2010.

Uber was chasing the same dream, with Airbus and Bell working on the vehicle design and Intel, Autoliv and Nvidia providing the tech. Flying cars weren’t as common as self-driving cars yet, but it looked like only a matter of time.

What Elon Musk was saying

Elon Musk, as usual, had bold predictions. At the Code Conference, he said Tesla’s cars would reach a level of autonomy greater than a human driving, possibly within two years, though he added that whether people would actually want to buy one was a separate question.

His bigger bet was on full autonomy: that within about five years, most miles driven would happen with nobody in the car at all.

Musk argued that it’s a misconception that self-driving makes roads more dangerous. Autonomous cars, he said, will save far more people than human drivers kill: they react faster, and they’re never drunk or distracted by their phones.

He also admitted Autopilot wasn’t perfect. Tesla once had to push out an update because the system was damaging cars when it braked during highway descents. Since then, he said, Autopilot had improved enough to operate on all three lanes of a highway, though Tesla still required the driver’s explicit approval before using it on public roads.

Who’s hitting the brakes

Not everyone was sold. Data from IHS Automotive suggested self-driving cars and trucks could be on U.S. roads in the next four years, with almost 16.9 million autonomous trucks and commercial vehicles by 2025. And the auto industry was grabbing all of that data to push its products.

Ford ran its own experiment: employees drove their own cars around while drinking beer and ice cream, logging their experiences on a dashboard camera. Ford said human drivers would remain in the driver’s seat, keeping full control over the vehicle.

David Glazier, an attorney with the Center for Automotive Research, was not impressed. He said he was stunned by it, calling it a disregard for safety and willful ignorance. How do you make sure these cars can interact with every other vehicle on the road? How do you test them in real-world situations? His verdict: the current technology wasn’t safe by any stretch of the imagination.

Still, he didn’t think the public had made up its mind. He placed most people somewhere in the middle: cautiously optimistic, not fully sold.

Then there’s regulation. Who regulates these cars: local municipalities, or states? Who’s liable when you get into an accident: you, or your car maker? California made a start with its Self-Driving Vehicle Law, in effect since 2015, which requires manufacturers to submit their testing and monitoring reports to the DMV on request. The bigger questions were still wide open.

Final thoughts

There’s a fun detail in the movie Drive: its famous car chase was filmed from a specially designed vehicle with cameras on the outside, giving audiences a view most of us never get from behind the wheel. Maybe that’s a little preview of where all this is headed: a ride where you get to enjoy the view instead of watching the road.

Whether that ride is years away or already parked in someone’s driveway, the safety case for self-driving cars is a strong one. The open question is whether the companies building them will take your privacy as seriously as their crash-test numbers.

Thanks for riding along.