The Trolley Problem

Your self-driving electric car is zipping along the road and out of nowhere, a child runs onto the street right in front of you. The car’s onboard AI algorithm has two choices – either swerve onto the side of the street and potentially harm the passengers or continue straight and run over the child. But perhaps the technology has accelerated so quickly that these kinds of surprises simply won’t occur, as sensors would detect the child from the periphery long before the road. This is known as the Trolley Problem in AI circles, and this ethical dilemma has sparked furious debate.

Perceived safety concern is the key barrier to self-driving adoption. Publicly available technology today (like Tesla’s autopilot) is limited to Level 2 Vehicle Autonomy, where most aspects of driving are automated, but the driver must be ready to take over at all times. However, only with Level 4/5 automation, where the car can deal with normal driving circumstances without a human, will self-driving technology ever be successfully adopted on a massive scale.

The barrier to achieving Level 4 autonomy is the Trolley Problem. It shows the clear moral conflicts with allowing computers to take control – a black-box algorithm gets to decide who lives or dies instead of a human. If it chooses to avoid the pedestrian and harm the vehicle’s passengers, people would not trust this technology to drive them. If it kills the child, the government wouldn’t accept it. This paradox must be resolved in order to scale adoption, ride-sharing, and higher rate of vehicle utilization.

I had an opportunity to interview Jeff Dean, an AI expert and one of the first employees at Google, who believes these conundrums would never actually take place. In fact, he claimed that the technology would be able to predict the child running onto the road long before it ever happened, preventing the situation entirely. 

Researchers around the world are looking for a technical solution to the Trolley Problem. Without a clear answer, societies and governments around the world will not adopt self-driving cars. This technology unlocks a massive opportunity for cheaper transportation and electric vehicles (EVs) powered by renewable energy, leading to substantial socioeconomic improvement.