The Case for Autonomous Cars
Written by Michael Kumar
From a young age, I knew that I was supposed to want a car. I knew that on my sixteenth-and-a-half birthday I was supposed to beg my parents for one last ride to the hallowed halls of the Department of Motor Vehicle for my road test before being set free into the boundless and unexplored strip malls of suburban New Jersey. Before that point, I had merely survived, but on that day, I would live.
Truth be told, however, I hate cars. I hate driving. Yes, having the ability to drive is certainly convenient, and I’m glad that during high school I could commute from school and baseball practice in my whip (a.k.a. 2008 Volkswagen Jetta — rims to come later). But the physical act of driving — that is God’s version of monotonous hell. I am by no means a “car guy”.
Coming to New York City, I flourished in passive act of sitting the back of a cab and was only minorly peeved with how long it takes to refill a MetroCard (why can’t I do it on my phone?). But somehow in this city with millions of people and a dearth of drivers licenses, I became with friends with two “car guys”.
The other day I was walking with one such friend when I saw a nice looking car. I studied it closely, and was able to gather from the jaguar on the the logo that this car was indeed a Jaguar. I asked him what type of Jag it was, casually showing off my deductive reasoning skills, and he quickly told told me the model (it was something with letters and numbers). He could tell I liked the car, so he quipped, “You know you have to drive this kind.” (My enthusiasm for driverless cars has turned into a bit of a running joke.) I’m not a car guy. I’m an autonomous car guy.
When I was a kid, I watched a Sebastian Thrun TED Talk. When I watched him tell the story of his friend that died in a car accident and how it inspired him to create a better car, I felt like the smartest guy in the room. I told everyone I knew (basically my parents and my brothers), and my dad—hoping that I would finally come around to this whole engineering thing—showed me the DARPA Grand Challenge, a collegiate autonomous car competition. We followed the live updates and when Carnegie Mellon won, www.cmu.edu become my homepage.
Mass adoption of autonomous cars will not be like changing my homepage—cars will not go from manually-operated to autonomous overnight. Rather the automation process will be frustratingly slow-moving, like taking the Garden State Parkway on a beach weekend. And while Google Maps may not be able to determine the ETA, it can tell us that the path to a world without human driving will pass through ADAS first.
ADAS, Advanced Driver Assistance System, is an umbrella term for automotive systems that automate certain aspects of the driving experience. They allow cars to parallel park themselves and power such things as automotive cruise control, a feature that allows you lock in a distance to the cars in front of you during highway driving (as opposed to standard cruise control which allows you to lock in a speed). These systems are powered by sensors that allow the cars to sense the road around, read signs, stop before a collision, and even change lanes. Drivers can usually override these systems, but as is the case with Tesla’s Autopilot and the Mercedes Benz’ Drive Pilot, the driver can choose to sit back and let the car do the work. In the automotive industry, these systems have been dubbed “semi-autonomous.”
Investment has flowed into ADAS companies, largely from automotive corporations feeling threatened by the emerging technology. GM acquired a pre-revenue startup, Cruise Automation, that had about 40 employees for excess of a billion dollars for the technology and talent in 2016. GM has also invested in Lyft with hopes of developing autonomous taxis that can be summoned with a mobile device (and I still can’t refill my MetroCard on my phone!). However, the winners in autonomous cars are yet to be determined and may not come from the traditional or existing automotive industry. Qualcomm, understanding the hardware implications of ADAS systems, acquired NXP Semiconductor, the largest supplier of chips for cars. Samsung, perhaps realizing that people will now need something new to do while they’re not driving, recently acquired Harman International, a car entertainment and services company. I suspect ridesharing giants Uber and Lyft may have changed their homepages as well, each having set up autonomous car research centers in Pittsburgh and begun testing driverless cars near the CMU campus.
When automated and autonomous cars become the accepted, I predict car ownership will become less ubiquitous. As Morgan Stanley Automotive Analyst Adam Jonas consistently points out, cars are typically used for less than 4% of the time. When car ownership becomes less essential to daily life and ridesharing platforms become cheaper and democratized, we can begin to embrace efficiency.I’m an autonomous “car guy” not because not only am I excited about not driving — I’m excited to eliminate parking lots and streamline traffic patterns.
Moreover, to the average family, the advent of autonomous cars means that parents can be assured that their children are safer traveling to school. They will save thousands of dollars on gas, insurance, and car payments and they will be freed up from driving, allowing them to focus their attention on the things that really matter.
When I tell people about my vision of a world with autonomous cars (that we summon using our smartphone-like augmented reality glasses and pay for with bitcoin), I get a common question in response: What are people going to do when we keep taking away these types of jobs? Driverless cars certainly a risky proposition, especially in the short run, but in my opinion it’s low hanging fruit for a world with technological progress and without the DMV.
Autonomous Cars: A Reality Check
Written by Sanchit Kumar
Humans are prone to making mistakes and thus we create technologies to solve our imperfections. Yet, technology has its own imperfections. Last May, a man was killed in a fatal crash involving a self-driving Tesla whose sensors failed to recognize a white truck in bright sunlight, questioning the longstanding belief that computers can operate a vehicle more safely than humans. Tesla CEO Elon Musk defends the technology in its current form, declaring that “the probability of having an accident is 50% lower if you have Autopilot on.” On the other hand, Ford argues that technology does not guarantee safety when you combine humans and robots, adding fuel to the debate over the future of driverless technology. In light of the lack of a regulatory framework for autonomous cars, it is dangerous to assume that technology will make us safer. The technology behind self-driving cars can be hacked and the cars can easily be weaponized. While carmakers around the world have outlandish images of what they envision vehicles to be in 20 years, the path to complete autonomy has a long way ahead.
First, the evidence simply does not suggest complete autonomy is foreseeable in the near future. In 2016, carmakers have carried out tests for their driverless technology in California, a state that experiences a mild climate and is hardly representative of global driving conditions. Google’s Waymo has done the best so far as it required one human intervention almost every 5000 miles. Its 60 testing cars drove about 11,000 miles – 3000 miles less than the annual US average – and required two interventions each. This is nowhere near perfect. Tesla’s self-driving cars performed a lot worse as its cars required 45 interventions each after driving only 140 miles – around 1 intervention every three miles. Each intervention is an accident which was potentially avoided. Given that the public will not tolerate any faults in technology, these facts are not encouraging.
Even if these technical problems are overcome, unforeseen externalities still pose as obstacles. One of the major challenges for driverless cars will be contemplating a world in which humans and driverless cars interact safely. Humans and algorithms have very different motivations to act and this loophole can be exploited when ethics and law diverge and do not provide the same answer. At times, our moral principles compel us to act illegally. For example, drivers might want to drive faster than the speed limit in times of an emergency. Should autonomous cars ever break the law? Are we sure autonomous cars will be properly motivated?
It is critical and inevitable for programmers to set principles and have a discussion about ethics, such as which choices are better than others. Is it better to save an adult or a child? What about saving ten adults or a child? Ethics in terms of numbers alone is incomplete as many other factors will exist that affect one’s choice. Programmers of driverless cars will have to account for an infinite number of scenarios. Human drivers might be forgiven for a bad and instinctive decision; however, programmers will not since they have the time to get it right.