Last week Google's self-driving cars team just released another monthly report on accidents involving its growing fleet of autonomous vehicles. The results of this data? Google robocars have still yet to cause a single accident. As in August, Google reported no additional accidents. Every month it is becoming more clear that Google Cars are actually getting better at avoiding accidents. During the last month alone, accidents per million miles traveled fell from 7.9 to 7.4.
This is the first monthly crash report with zero new crashes. As I covered in my last post on this topic, Google Cars have caused zero accidents to date.
Here's some more data. During the past month, Google added 20 more cars to the road, nearly doubling their fleet size to 48, The increase meant that the number of miles traveled in aggregate grew quickly. The number of accidents reported involving Google Cars over the past five years is a whopping total of 16. The significant majority of those incidents involved a human driver rear-ending a Google Car at a stop sign or intersection. And nearly all of them have been in California, where Google Cars have driven the most miles to date.
Tellingly, a significant percentage of the human drivers that crashed into the Google Cars were texting or otherwise distracted with mobile phones. This illustrates why humans are so poorly equipped to control a multi-ton vehicle. They are easily distracted. We are wired for distraction and unable to ignore notifications from our phones, the urge to rubberneck, or the desire to talk on the phone while driving, which research has shown, is equally dangerous when the driver is using hands-free devices like headsets. That research underscores an uncomfortable reality. Our brains do not multitask well. Driving and talking on the phone split our mental processing power in two, reducing our capabilities commensurately. That's not even covering how controlled substances can reduce our inhibitions, impair our judgement and entice to drive while in under the influence, a leading cause among the 33,000 car accident casualties recorded every year in the United States.
The economic lift from ridding the roads of human-driven vehicles would be over $190 billion per year, according to the consultancy McKinsey. That would primarily come from reducing property damage caused by low-speed collisions. McKinsey does not look at reduced costs from reductions in higher-speed, more damaging collisions. Those would include hospital bills, first-responder costs, and insurance costs for disability. That estimate may be conservative. Experts believe that we could eliminate nearly all accidents if we went to 100% self-driving cars directed by a uniform traffic control and collision avoidance network. In that scenario, the cars would communicate directly with each other and via a specialized network.
This would open up many other interesting possibilities. Cars could drive much faster and could travel more closely together. Of course, ownership of a car would be less attractive when it would be easier and cheaper to hail a car service from a smartphone. This on-demand availability would make car ownership less attractive to tens of millions of people. With the new bespoke Google Cars, we are getting glimpses of that future. Those cars do not bother with many of the accoutrements for drivers. There is no ignition systems or key, only a green start button, for example. Removing the pretense that humans will need anything more than bare-bones ability to control a car will allow engineers to remove many wasteful features that weigh down a vehicle and reduce its fuel economy.
But as we grow to trust the robo-cars, then steering wheels could disappear entirely. Rather, software could deliver a gentle braking mechanism that occurs when an autonomous car is unable to brake or becomes technically compromised. As Marc Andreessen, the famous venture capitalist and futurist points out, objections to automated cars echo older objects to automated elevators. Yes, humans used to control those entirely and switching to automated systems made elevators safer and vastly more efficient. The data doesn't lie. The Google Cars are adding mile after mile with no dangerous side effects. Google is also improving their systems in ways that human drivers never could.
For example, after an embarrassing stand off with a cyclist in a which a Google Car would appeared confused by the rider's balancing act and unwilling to enter a four-way intersection, Google quickly adjusted the driving algorithm so that cars would gently edge into the intersection to signal intent and avoid stalemates. Humans, sadly, are far harder to program or improve because our systems are not software. Let's admit it. The data continue to mount. Let the robots drive, please, and help more of us arrive alive.