This is a somewhat unique post compared to my previous posts on the blog. My hope is to transition to more frequent posts and this is more or less a trial post. As always, the idea for this post was generated while I was riding my bicycle to work.
For the first time, I am referencing a current event on this blog. I recognize that paying attention to current events is best viewed as entertainment rather that seeking knowledge. However, I see this particular event as worthy of discussion here because I believe that the fatal Tesla accident has the potential to push autonomous cars in two different directions.
Full Discloser: I am an autonomous car BULL. From the vantage point of a bicycle, riding through somewhat heavy traffic in a city every working day of the year, I witness the insanity of people-driven cars. My favorite example is when I was a block away from home. I was waiting patiently at a stop light and there was a group of cars coming from the north. The light turned from green to yellow and a car that was half a block away from the intersection appeared to be accelerating so I waited and observed. This driver was young, likely a college student. He had a burrito in one hand with a cell phone in the other. He was texting and chewing and apparently steering with his left elbow. He crossed the intersection approximately 3 seconds after the light had turned red and then cut off a car 200 feet beyond the light, forcing this driver to slam on the brakes. No matter what your feelings are about self-driving cars, I am pretty sure taking that foolery off the road will be a net positive for civilization.
Getting back to the Tesla fatality. There is the somewhat likely possibility that everyone is going to freak out about this death and suggest that we just cannot trust this technology. We have already seen this with absurd headlines on Google News and a full segment on the Today show (and presumably the NBC nightly news) dedicated to pointing out the horror of this one traffic related death. These sentiments could become so loud that government regulators listen to these fears and push ubiquitous autonomous driving cars into the next decade and beyond.
The purpose of this post is to counter the mainstream view of this event and push this event in the positive direction.
First of all, Tesla’s response was the most rational response any company could have to the event. They first explain that the current death rate of drivers using autopilot in their cars is well below average human-driven cars. I am going to expand on this to make this clear. According to their numbers, there has been one death in 130 million miles during autonomous driving and that in the world there is one death per 60 million miles. In the United States, we have a lower death rate than the rest of the world per mile. As everyone is aware, Americans drive a lot! The most recent fatality statistics in the US indicate there were 38,000 car-related deaths per year (2015). In that same year US drivers logged 3.06 trillion miles. That is one death per 80.5 million miles [3.06 trl miles per year/38000 deaths per year = 80.5 mil miles per death]. So the death rate for autonomous driving is still significantly lower than that of all US drivers. If we extrapolate the fatality rate of Tesla autonomous driving cars to all cars in the US [(1/130 mil)*3.06 trl] there would be significantly fewer traffic related deaths in the US per year: 23,538 deaths. Please note this is almost 15,000 people who would be alive today (per year) if all cars performed like Tesla’s autonomous driving cars.
If that doesn’t sound like a lot of savings of death to you, let me put it into this perspective. On all of the lists of things that all responsible parents are suppose to do, is to douse your babies and children with sun screen. Americans spent around half a billion dollars on sunscreen in 2013. Not only this, many kids are deprived of the outdoors largely because of fear of the sun. This fact has many additional costs that I will elaborate on in a future post. Why do we freak out about the sun? Allegedly, it is is because of skin cancer. Do you know how many people die from skin cancer every year? Less than 15000 people. Here is the estimate 2016 death rate from the main skin cancer type that kills people. If everyone was driving an autonomous Tesla, we could throw out that smelly sun screen and spend every day frolicking in parks and forests and still be ahead in fatalities, especially since there is little evidence that staying out of the sun/wearing sunscreen has any effect of mortality related to skin cancer.
Which would you prefer?
Going back to Tesla’s response. They have been relatively clear that their autopilot feature is not to be used in the way the driver who was killed was using it. Tesla has placed many warnings and safety systems in place that tells the driver she still needs to keep her hands on the wheel at all times. The driver was probably not following the rules. We actually know this particular driver has had a lot of experience with the autopilot feature. He was even known to post videos of his experience online, including an incident (tweeted by Tesla CEO Elon Musk) where his Tesla avoided a truck which cut him off (link to the deceased’s YouTube channel).
Unfortunately, reliance on highly functional technology led this driver to have too much confidence with the system and he pushed the boundries beyond what Tesla was comfortable with. This is human nature. For example, people that work on the 67th floor of a New York City skyscraper are not prepared to walk up 67 floors worth of stairs everyday. They have been trained to have confidence in the technology of elevators. If the elevators malfunctioned a lot of people would not get to work. It is also well-known that pilots that rely heavily on autopilots systems happen to make more mistakes. No different with the Tesla.
The driver apparently was so reliant on the very effective autopilot software that he felt comfortable watching movies while his car did the driving for him. Thus, neither him or the car was able to identify the truck that pulled out in front of him. This is very sad, as the folks at Tesla have pointed out.
So why is this death in a self-driving Tesla a turning point where we either regress back into primal defensive mode and reject the technology or embrace the news as both tragic and promising.
What is different about this US car-related death versus the 37,9999+ other deaths that will occur in 2016 is that there is an opportunity from this death to prevent future deaths from occurring. Let us look at the current state of the world in both the best and worst case scenarios related to car accidents.
Fictional Dave is driving down the road and he is going slightly above the speed limit and approaches a curve. Dave is from out of town and is late for a business meeting so he continues to accelerate as he rounds the bend. Little does he know that there is a busy intersection with traffic backed up to just before the bend Dave is approaching. Dave had briefly looked at his cell phone to see how long before his next turn and before he looks up, he has plowed into the rear end of an SUV with 2 children in the back seat. Dave and these two children parish.
The local department of transportation investigates the accident and after dozens of meetings and engineering reviews by people with little experience in this situation, the city decides that a solution to this problem needs to be pursued. More months go by and this dangerous situation at this intersection remains but fortunately no accidents occur. 18 months later and after great expense, these experts devise and install a flashing sign warning of future drivers that an intersection is coming up and to slow down. I repeat, this is a best case scenario.
The worst case scenario returns us to the burrito-eating, texting driver mentioned above. Let us imagine that the car he cut off failed to slow down and runs into back side of Burrito Guy’s car. Burrito Guy has no control of the steering wheel and is distracted so his car swerves out of control and hits a pedestrian on the street, leading to a fatality. A full investigation by the local police determines that the Burrito Guy is at fault for distracted driving. They take away his license for 12 months and charge him with manslaughter.
In the better case scenario, there are minor changes to the roadway system at that particular position in the world but at great cost and there is no clear evidence that the changes made will actually prevent any future accident at that intersection, let alone at an intersection two states over. In the worst case scenario, the only improvement to the road system is that one bad driver out of millions is taken off the road for a short period of time. The next day on the same road, another driver is going to be equally distracted and equally deadly, he or she might just be lucky enough that the person who gets cut off knows to brake.
In the case of the Tesla fatality, all of the information from the crash including precise sensor and object recognition data is now available on the Tesla servers. In addition to the police and local investigation, Tesla will do extensive root cause analyses (perhaps there will even be cooperation between Tesla and others in this). Tesla is highly invested in the success of autonomous driving so they have the desire and resources to employ the most qualified experts and computer scientist to figure out what exactly went wrong and devise advancements in the existing technology to make sure no such event occurs again. Because Tesla will be fixing the driver itself, the improvements they make will apply to all cars in the world with the Tesla system. This mean an accident in Norway will improve the safety of an autonomous driving Tesla in Peoria. The implications of this is that driving intelligence will constantly improve (and will never be forgotten). This means that autonomous driving will get better and better the more miles that people drive.
To better illustrate this point, let us think about modifying the behavior of 214 million human drivers. There are precedents to doing this. Even simple, obvious messages require an incredible amount of effort and have limited results. How much money do you think has been spent on anti-drunk driving campaigns in the last 40 years? Yet there are still 10,000 alcohol related deaths per year in the US. Seat belt awareness has only made minor dents in seat belt utilization. So imagine saying at 7:15 am, the sun is in such position and you may or may not see a bright, reflective truck turning in front of you at westbound 100 Main Street. That sounds silly but the technology Tesla and Google have truly has this power. Every event, whether fatal or not, has the potential to save lives the next time this situation arises.
Image 1 Caption: Below are the vehicle related fatalities in the United states over the past century. I always found this graphic to be fascinating because it incorporates economic, technological, and sociologic changes. Note the near exponential grow of death from the popularization of motor vehicles until the start of the Great Depression. There is a flattening of the death rate through the depression followed but a steep drop related to World War II rationing and the deaths of young service men. After the war, the numbers climb as the industrial economy of the US explodes. In the 70’s, there was a realization that drunk driving was resulting in a lot of dead teenagers so there was a reduction in alcohol related deaths due to DUI enforcement. This was coupled with improved safety technology and more seat belt use resulting in a gradual decrease. In 2008, there was a precipitous drop in deaths due to the recession. By the end of 2016, we will be back to close to pre-recession vehicle mortality.
I really think that autonomous car driving technology has the potential to make this graph look like a dysmorphic parabola with vehicle related deaths approaching ZERO by 2030. There are other reasons I want autonomous cars to succeed but this is the major reason.
I hope some of you who read enjoy the content. If you do, please post comments below.