Road traffic safety cannot be traded lightly. Many people, including this writer, can attest to the life-alerting experience of traffic accidents. What if there is a way to reduce traffic accidents? Should we keep ignoring it? Let’s not rush to the answer. Let us first take a detour with history.
As soon as multi-storied buildings started to pop up, elevators were invented. These were different from what we see today. The safety features were not as great as today. Numerous accidents happened either due to design flaws or operational errors. Nevertheless, the elevator’s motion is simple – going up and down with controlled acceleration and deceleration. Still, there used to be operators employed whose sole job was to close the door and press the button to take elevators to the desired floor.
Until the 1940s, this was a common scene in the USA; even in the early nineties, it was also common in Bangladesh. Slowly but surely, people realized that the elevator operation was simple and did not need a full-time employee.
Adopting any technology is a gradual shift and depends on many socio-economic factors. The extinction of the liftman was also such a case. Between 1920-1960, there were a few elevator strikes in New York City when the operators refused to work in demand of better wages. The 1945 strike brought the city to a standstill. These strikes created incentives to design elevators that would be easy to use and not need any operator. At some point, it became the norm that lifts would work on button presses without needing a trained professional.
Cities like Vancouver, San Francisco, Seattle, and some European cities have foggy climates year-round. Although pilots face a greater challenge landing aircraft in those cities, accidents are rare in these airports. It is because pilots rely heavily on sensors and onboard computers to evaluate the surrounding environment and make decisions.
Fog can create a barrier to human visibility, but modern aircraft have sensors that can see through dense fog and make the landing a breeze. When planes fly at a high altitude, the onboard computer system takes over most of the navigational tasks. A pilot can intervene and manually override a computer’s decisions, but pilots are overly dependent on the computer for navigation. We still have a psychological barrier that does not allow us to board a plane where computers fly, but we got used to riding an elevator without an operator. We will become comfortable boarding a plane without a human pilot someday.
So now the question arises, what are the issues if we outsource the tasks of driving a car to computers? Imagine you want to use an app on your phone to summon a car. The ride-sharing company sends you a car with a human driver who takes you to your destination. Now, let us assume this. The company sends you a driverless car, it takes you to the destination, and you pay for it from the app. No fuss.
If you ask 100 people how far that future is away, half of them will say 10 years down the road. But the future is already here. Although the technology is not perfect yet, level 5 autonomous driving cars are here now. Like any technology, it is still nascent and has some kinks that need to be ironed out. In the last 2-3 years, many cities have allowed these cars to be driven experimentally on public roads, and the data so far makes us hopeful. The cars have already amassed a few billion miles of driving and have a near-perfect driving record.
Each driverless vehicle has about 30 cameras, lidar, radar and other sensors. Because of these, the vehicle can better assess the environment than humans. The statistics available today attest that autonomous vehicles are much better drivers than humans. They do not break the rules, and if there is a situation where the onboard computer cannot decide, it can be remotely resolved or navigated using connected technologies.
Some vehicles coming to the market do not even have steering wheels or seats for drivers. The biggest hurdle at this point is not the technology but the lack of a legal framework. It is not just the politicians, ranging from the public to insurance professionals — many of us are still struggling to accept the new reality.
It has become necessary for countries to create a legal framework for driverless vehicles and develop a strategic vision to replace human drivers with computers. Because driverless cars will cause fewer traffic accidents than human drivers. The number of lives lost each year due to traffic incidents can be dramatically reduced to a fraction of what it is today if all vehicles become driverless.
Elevators have become operator-less, most aircraft operations are computer-assisted, and cars, buses and trucks will become autonomous also. Driving as a profession may even become extinct. In developed nations, most long-distance trains still have drivers; however, airport shuttles and short-distance freight trains are usually driverless nowadays.
We must recognize that advances in machine learning and big data, new sensor technologies, and onboard computers with lots of processing power enable autonomous vehicles to make decisions faster than humans. Most vehicles now have redundant navigation capability to guard against a single failure mode.
Here is another economic argument for driverless cars. When we buy a vehicle for personal use, it stays parked 95% of the time — a significant investment that remains underutilized. So, there are now discussions about changing the ownership model. If there is a way to have on-demand access to reliable and fast driverless vehicles, there is no reason to buy a traditional car and keep it idle most of the time. There will be no time wasted finding parking after reaching the destination. The car can drop the passenger and move on to the next one.
A few decades ago, philosophers conceived the ‘Trolley Problem.’ We are here going to present it as a driving scenario. Suppose you are driving a car extremely fast and realize the brakes are not working. Looking ahead, you see that 5 people are crossing the road, and a collision will be fatal for them. On your right side is an old man standing, and if you steer the car in that direction, the man will die, but those 5 pedestrians will survive. Also, there is a wall on your left and crashing into that wall will surely save all of them except you. Now the question is which way you should decide.
The problem might appear overly dramatic but not improbable. We want to understand the implication of each decision and solve this problem now. Because, in an autonomous vehicle, you will not make the decision. The onboard computer or the programmer writing the software needs to know the ‘right’ decision. And ‘right’ is questionable here since we are dealing with moral philosophy.
As computers are fast and accurate at computing things, they are usually programmed to solve this type of problem using laws of probability. It tries to compute a decision tree with an estimated loss for each node and then opts for the node that minimizes the loss. It may find out that running over the old man minimizes loss and attempts to do that. Some benevolent programmers might decide to kill the driver but none of the pedestrians. It all sounds very utilitarian in principle, but you will probably not buy such a car.
The odds of a trolley problem scenario are extremely low, particularly when all vehicles will be driverless. There is no point in letting so many lives be lost in fatal vehicle accidents just because of this remote possibility of ending up in a trolley problem. Stories portrayed in science fiction and movies swayed some people’s opinions against computers and robots with decision-making abilities. But like COVID vaccines, the rewards outweigh the risks for autonomous vehicles.
Commercial trucking is expensive in the United States due to high driver wages and regulations on how long a driver can operate without a break. This incentivized some companies to invest in autonomous trucking. It will probably take a strike from truck drivers like elevator operators in New York to show people that autonomous trucking is safer and cheaper than human truck drivers. There are about two million professional drivers in the United States, and all these jobs may be lost. But, making and maintaining these fleets of autonomous vehicles, there will be a huge need for programmers, engineers and a variety of new professions which do not even exist today. There is no way to ignore the economic importance of this future possibility.