The brief exchange highlights a massive thorn in the side of automakers and policymakers alike: Self-driving cars will have to share the road with human drivers, likely for decades to come. Those vehicles must therefore respond and adapt to the idiosyncrasies and mistakes of humans behind the wheel.
Harper’s question also points to the barriers, large and small, that still exist as automakers aggressively push toward a self-driving future. Tesla chief executive Elon Musk predicted at a conference Monday that almost all new cars will be able to drive themselves in 10 years. Last week, Ford announced plans to spend $1 billion over the next five years creating artificial intelligence for its own autonomous fleet.
Self-driving technology has already come a long way. Cars being tested on the road today use cameras and radar, for example, to detect the movements of nearby cyclists and pedestrians. There are already driver-assist technologies built into some cars that monitor the speed of vehicles around you, detect potential accidents before they occur, and automatically slow or stop the car to avoid a wreck.
In a world where all cars drive themselves, the technology could operate at peak efficiency. Proponents of autonomous cars say that speed limits could be raised and fatal collisions largely avoided as no traffic laws are broken and poor drivers become a nuisance of the past.
But that reality is a long way off, even if self-driving technology becomes standard in the next decade. The migration to autonomous vehicles is expected to be gradual, and will likely start with ride-sharing services, as the costs of personally owning a self-driving car remain prohibitively high.
That means man and machine will have to play nice on America’s roadways.
“There’s so many interesting scenarios,” such as honking a horn, Harper said Tuesday. When “another car with a driver comes across a self-driving car without a driver in there and they realize that, it will freak some people out. How that’s going to be dealt with will be part of the fun part of this process.”
Any daily commuter can tell you that driving involves a lot of social interaction, particularly the hand gestures and eye contact of fellow drivers. An autonomous vehicle today won’t understand if a driver waves for you to proceed through an intersection or flips the middle finger because you accidentally cut them off, said Nidhi Kalra, co-director of the Center for Decision Making Under Uncertainty at RAND, a nonprofit research institution. There are also driving customs to contend with that might be unique to a particular city or state.
“We have driving dialects if you will that are hard for autonomous vehicles to integrate with, and some of them have safety implications,” Kalra said.
Much of that communication between drivers will eventually be digital as cars increasingly collect data and share it with the vehicles around them. If a car speeds up, slows down or prepares for an abrupt stop, for example, that would be communicated instantaneously and electronically to the other vehicles on the road. Like self-driving technology, however, it will be years before the vehicle-to-vehicle communication is ubiquitous.
That challenge is being taken into account as autonomous vehicles are tested on public roads. Uber, Ford, GM and Waymo, among others, already have self-driving vehicles sharing streets and highways with motorists in driving environments as diverse as Pittsburgh, San Francisco, Detroit and Scottsdale, Ariz.
The honking issue is particularly tricky because it’s subject to interpretation. Even today, you might hear a honk whether you’re about to back into another car or a friend spots you driving from an adjacent lane. With the human brain, you can quickly determine the difference.
“It’s going to be really hard for an autonomous vehicle, even if it hears the honk, to figure out what that honk means,” Kalra said.
Read more from The Washington Post’s Innovation’s section.