Both tech giants and traditional automakers have been investing in self-driving technology for the better part of the last decade. Contemporary autonomous driving prototypes are incredibly advanced in comparison to solutions Google started developing back in 2009, but they still have a long way to go before they become a viable alternative to human drivers. Regardless of that, the auto industry already started including some limited autonomous technologies in its latest products. The new Tesla Model S is a good example of that endeavor seeing how the Palo Alto-based company is offering to equip this electric car with a package of sensors that enables the so-called Tesla Autopilot feature. As the name suggests, Tesla Autopilot allows the car to operate on its own in certain conditions, but it's still far from offering a fully fledged self-driving experience.
Namely, this technology utilizes sensors such as radars to identify the surroundings of one's vehicle and control the car accordingly. However, Tesla is specifically marketing this feature as designed for assisting drivers and not for completely autonomous driving. Unfortunately, Tesla Autopilot lulled some people into a false sense of security, which is how one Joshua Brown was killed in a car accident this May after his Tesla Model S crashed into a trailer it failed to detect. An incident such as this one is something that sticks in people's minds even if both Tesla and authorities agree the driver was at fault for expecting their driving assistance technology to do more than just assist.
So, we're still somewhat away from tech giants and automakers commercializing completely autonomous vehicles as there's a lot more work to be done on that front. As Danny Shapiro, one of the executives of NVIDIA's car division recently stated, the industry has yet to figure out how to completely replicate the behavior of human drivers. At the moment, a self-driving vehicle needs at least 20 sensors to achieve a level of perception a human driver has. In addition to cameras, radars, and GPS mapping solutions, contemporary autonomous cars also rely on something called a LIDAR, which is an acronym for Light Detection and Ranging, a sensing technology which shoots lasers to measure variable distances. However, even when all of these technologies work seamlessly and the car "sees" everything around it, it still needs to interpret that data in real-time.
That's why exhaustive testing is a crucial part of the effort to commercialize self-driving vehicles and teach deep learning technologies powering them how to react to various objects that one might encounter on roads. However, sooner or later, these prototypes learn everything they can from operating in controlled conditions and must start driving on public roads to improve further. Now, given how these are still prototypes we're talking about, they're not going to perform flawlessly. For example, Uber recently got into a lot of trouble after one of its self-driving cars was recorded running a red light in San Francisco.
When Shapiro was asked about these issues, NVIDIA's executive was rather straightforward on the matter by asserting that some accidents will simply be unavoidable. In addition to the possibility that a self-driving car encounters an object it doesn't recognize and can't respond accordingly, Shapiro said that some accidents were and will be caused by unpredictable human behavior. As he explained, while autonomous vehicles are already statistically safer than human drivers, this technology won't eliminate traffic accidents as long as there are human drivers and pedestrians on public roads. In fact, "crazy behavior of humans" is one of the main reasons why Shapiro believes people need to cut self-driving cars some slack.
Needless to say, that's easier said than done, as evidenced by numerous surveys which suggest people aren't ready for the autonomous driving revolution. While most consumers who've already had the pleasure of operating vehicles with semi-autonomous driving assistance like the Tesla Model S are more inclined to trust self-driving cars with their lives, people, in general, are still scared by the idea of a fully automated vehicle. Sure, driving can be a stressful activity, but if people find the idea of being driven by a computer even more stressful, then that's potentially the biggest obstacle the industry will have to overcome before autonomous vehicles become a common occurrence on public roads. Sensors, radars, and artificial intelligence solutions will certainly improve by a wide margin in a few years, but it remains to be seen whether people's perceptions can be changed as quickly.