Self-driving cars rely on artificial intelligence (AI). In the 2010s, there were considerable advances in AI. This progress drove optimistic predictions for self-driving cars. 

But the limitations of AI became evident. One problem is the need for training data (billions of hours of footage of actual driving) and teaching the computer good driving behaviour.

In the future, it is likely that taxi and ride-hailing companies will offer self-driving cars instead of paying drivers. Self-driving cars can also make a difference in people with disabilities who can't move around freely.

However, polls show that most Americans don't want to be driven around by a self-driving car. Researchers argue that widespread self-driving cars first need significant changes made to the streets to make it easier to communicate information to those cars.

Many major car manufacturers have tested the waters with self-driving car research.

Two statistics are helpful in evaluating how advanced a self-driving car program is: 

• How many miles it has driven (training data and investment to get cars on the road), and 
• Disengagement (how many times human drivers had to take over.)

In 2018, Waymo drove 1.2 million miles with 0.09 disengagements every 1,000 miles. General Motors' Cruse drove half a million miles and 0.19 disengagements per 1,000 miles.

In 2018, a self-driving Uber car ran down and killed a pedestrian walking her bicycle across the street. The near-range cameras and ultrasonic sensors were not in use at the time of the crash. The system also had false alarms and failed to identify her because she did not cross at a crosswalk.

In 2018 Tesla's Autopilot system was implicated in a lethal accident where the car steered into a concrete divider. There were three more deadly Tesla crashes.

Some advocates claim that self-driving cars may reduce car trips as car ownership may become redundant. Others argue that people drive in a wasteful way.

But there's not much evidence that computers are more fuel-efficient drivers. Studies also show that people will go on more trips if the cars are more fuel-efficient or if a vehicle could drive itself.

It's not uncommon for technology to be dangerous when it is invented and then refined into a valuable part of life. The first aeroplanes were dangerous and commercially useless.

While companies boast about their progress, they hesitate to put their cars on the road when they're not ready. They are aware that killing someone, as has happened, will probably spell doom for their business. Despite the progress made, some sceptics think that fully self-driving cars might never happen.