The Uber Self-Driving Accident Proves Humans Are Bad at Driving

Every day in the United States and globally, people die or are terribly injured in automobile accidents.  Statistically speaking, automobiles are one of the most dangerous ways to travel with ~40,000 people dying annually in the United States and 1, 250,0000 globally. There are many fundamental challenges associated with driving for humans – we are maneuvering very powerful and heavy automobiles with a wide variety of factors that are encountered during our daily use. Yet, a few weeks ago, tragedy struck in the realm of emerging self-driving technology. A self-driving Uber hit and killed a Pedestrian in Arizona. Predictably, this had led lawmakers and the general public to question the wisdom of this technology.  While, we should always figure out the root cause whenever there is any preventable loss of life; this accident also proves why we need this technology to be perfected sooner rather than later.

During the testing phase, each self-driving Uber has 1-2 safety drivers who are supposed to monitor the roadways & intervene when necessary to avoid crashes. As we see from the video, the safety driver was distracted and did not have her hands above the steering wheel as required by protocol.  In 2015, the National Highway Transportation & Safety Administration (NHTSA) estimates that 3,477 lives were lost due to distracted driving. This means that nearly 10 lives are lost per day because drivers are not paying attention. How many times have we seen a vehicle leave their lane or be driving extremely slowly? When we pull alongside or pass the vehicle, more often than not we see someone looking down at their cellular phone. In fact, one could argue that a self-driving vehicle with perfected technology would have given this pedestrian the greatest chance to live to see another day.  Unfortunately, traveling within a self-driving vehicle makes the occupants more susceptible to distraction because the monotony leads to the human letting their guard down. Even Tesla’s autopilot has been susceptible to this issue. Since the computer is so good at controlling the vehicle, drivers erroneously believe that the technology is further along than it actually is.

Computers are very good at predictably doing the same thing over and over again. Using a multitude of powerful sensors, expansive processing power and intelligently-designed algorithms, self-driving vehicles should eventually exceed the capabilities of humans. In fact, one emerging key performance indictor for self-driving vehicles is miles traveled between human interventions. Uber’s competitor, Waymo (division of Google) reports that they are able to go on average 5,600 miles before requiring human intervention. In contrast, Uber was struggling to meet their internal target of 13 miles without human intervention.  While Uber’s technology shows a large performance delta between that of Waymo, it is still approximately the average one-way distance (16 miles) for most workers commute.  However, the ultimate goal will be to reduce fatalities and injuries associated with automobiles to zero.  Additional testing and algorithmic improvements should accomplish this within the next 2-3 years.

Airlines have been developing and improving automated technology for decades. Using many of the same fundamentals as self-driving vehicles, auto pilot has the human pilot input the flight plan and then the computer takes over to determine the best way to execute the mission based on the changing speed, altitude and weather. However, autopilot still has not eliminated the need for humans. In fact, pilots are required to intervene during difficult conditions (turbulence). Over the years, autopilot addressed one of the fundamental challenges of flight – boredom in the cabin which led to fatigue and accidents. Unfortunately, most humans still believe that driving their automobile to the airport for a flight is the safest part of their journey. The data says otherwise; air travel is one of the safest modes of transportation. In fact, airplanes have a fatality rate of 0.1 / 100,000 flight hours while highway travel is 3.0 / 100 million vehicle-miles. This means that as self-driving technology improves, one of the biggest challenges will be convincing humans just how dangerous it is for them to get behind the wheel.

An investigation is already underway to figure out what went wrong in this tragic accident. Ideally, this analysis will identify improvements to accelerate the implementation of this technology so that we can eliminate or dramatically reduce fatalities and injuries from automobiles by removing one of the biggest challenges to vehicles – the human.