We are living in a time of unprecedented change as new technologies continue to transform our societies and infrastructure. The inevitable introduction of driverless vehicles is quite possibly the most dramatic change the world faces since the introduction of, curiously enough, the automobile itself.

The idea of the autonomous vehicle has many obvious appeals. Every year in the US, over 30,000 deaths are caused by vehicle accidents attributable to human error. The expectation is that driverless cars would greatly reduce the number of these deaths since error-prone humans wouldn’t be driving them.

Additionally, it is expected that autonomous vehicles would lessen the impact our car culture currently has on the environment. There would be less pollution from fuel waste due to inefficient human driving, idling in traffic, and the excess driving caused by the time-consuming searches for parking. The number of cars on the road would decrease if the introduction of driverless cars spurs a drop in car ownership due to the automation of taxi service and other modes of public transportation. This could also lead to a reduction in suburban sprawl, as life without a car would be much easier in dense urban cities with new automated transportation options, which would in turn reduce human encroachment on wildlife habitats.

But, for all the benefits this new technology promises, there are many complications to deal with along the way.

The artificial intelligence that operates driverless cars currently uses a combination of preloaded maps, GPS, radar, LIDAR (laser-based radar), and other sensors to navigate and avoid collisions. In many ways it is superior to human vision, but in some very essential ways it is not. Driverless cars are good at avoiding obstacles such as construction, other vehicles and the odd pedestrian, but how well will it cope with roadways blanketed in snow, or leaves, or flooding from a heavy rainstorm? It has already been shown that a flurry of snowflakes can completely confuse their sensors, while human vision is better able to adapt to the challenge.

Will a driverless car be able to identify the siren of an emergency vehicle and know to pull over? What about a police officer directing traffic? These are common situations that are easy for humans to handle that would be a far more complicated challenge for artificial intelligence.

When an automated vehicle encounters a situation that it can’t handle, fallback to a human operator in most cases would not be a viable option – particularly if the passenger has been napping, having a few cocktails in the back seat, or otherwise enjoying the benefits of not driving. Some prototypes of driverless cars (Google, for example) don’t even have an option for human steering. What then? Will the car just stop in the middle of the road if it gets confused?

Another issue is how driverless cars will affect the insurance industry. Currently, it is fairly straightforward to assign blame in a car accident – it will be either one driver or the other or a combination of the two. If a driverless vehicle causes an accident, who is liable: the human passenger, the car manufacturer, the mapping platform, or the AI developer? What if the accident is the result of a faulty signal from another driverless vehicle? The way insurance companies apportion liability will be completely upended.

Our transportation infrastructure is also set to be upended. Will street lighting be optimized for radar or human vision? How well will a driverless car cope with street signs covered in graffiti or obscured by mud or foliage? Will they all need to be replaced by some sort of RFID alternative? How much will that cost? Who will pay? None of these questions, so far, have clear answers.

Another possibility is that the allure of having a chauffeur for each member of the family (and perhaps even the dog) increases demand for car ownership, thus undoing the advantages of having fewer vehicles on the road. Similarly, there could be a surge in suburban sprawl as people are offered the convenience of a relaxing and productive joyride to work as opposed to a grueling commute – great for happy commuters but far less so for sensitive ecosystems already under pressure.

The introduction of driverless cars may create nearly as many complications as they resolve, but technological progress has always been a complicated business. The potential rewards to be reaped from the application of artificial intelligence are too great to be ignored. A thoughtful approach to the integration of this new technology will be essential as the many complexities will have to be addressed and solved. But once we find those solutions our world will be transformed in ways we can only imagine.