Autonomous cars: It’s time for a reality check!

Source: Dllu

As artificial intelligence becomes the hotspot in global academia, one of the major applications of it has attracted great attention in the recent days. In October 2014, the famous electric-automotive manufacturing company Tesla, initially released its semi-autonomous driving assist service Autopilot for its Model S and Model X. Also,  Waymo, the affiliated self-driving company of Google, launched its driverless ride-hailing service in Arizona four months ago. It seems like the future of taking a nap in a car with nobody behind the wheel is rapidly approaching. However, normal people are still skeptical of whether a computer is able to protect passengers’ lives in circumstances of unpredictable pedestrians, varying city roads and other variables.

Above is the real road testing video from Waymo

Improved safety of autonomous cars, compared to that of human-driven cars, is arguably a main push behind the self-driving cars rush. According to statistics from U.S. Department of Transportation, over 90% of car accidents are due to drivers’ action errors. By contrast, self-driving cars make determinations by processing millions of data from the surrounding environment in every seconds, and therefore they cannot do dangerous behaviors related to tiredness, drunkenness or anger.

But indeed, there exists a lot of limits to a computer-controlled vehicle. It’s nearly impossible to be programmed in advance to handle every imaginable events on real-world roads. Even though they have been already programmed to be acting as defensive drivers, they also have to be ready for other react to other nearby road users driving unsafely. A recent example is a lady turned left to cross three lanes and hit a self-driving Uber vehicle.

Source: AP Images

Other technical obstacles include how to recognize out-of-work traffic lights and hand signals/spoken commands from traffic control officers. Teaching a computer to learn the eye contacts and body languages from a bicyclist will require a huge amount of road testing data, and more accurate radar technology.

At last, cybersecurity and data privacy are the raising concerns in any networked devices. Companies can collect numerous data on pedestrians and vehicles, however, if the computer of an autonomous car is hacked, not only these privacy data might be leaked, but safety of passengers could be endangered.

There is still a long way to go for autonomous cars to have a deeper understanding of complicated human driving etiquettes. Bryan Salesky, the CEO of Argo. AI, says “Those who think fully self-driving vehicles will be ubiquitous on city streets months from now or even in a few years are not well connected to the state of the art or committed to the safe deployment of the technology.” The future is awaiting, but it will just take longer time than we expected.