Arguably one of the most exciting innovations out of the auto industry is the self-driving car. 

Besides the obvious excitement due to their novelty, manufacturers are also hoping they could be a safety breakthrough for greatly diminishing auto accidents. With approximately 94% of serious vehicle accidents caused by human error, self-driving cars could be a solution. 

Enthusiasts believe that self-driving cars might be safe enough to greatly reduce the 35,000 deaths that happen every year on American roads. 

So, how safe are self-driving cars, and should you be worried about driving next to them on the road? Let’s find out. 

Are self-driving cars really safer than non-automated cars?

Cars have levels of automation, ranging from 0 (no automation) to 5 (full automation), as defined by the Society of Automotive Engineers (SAE).

Levels 0 through 2 need human intervention, with AI providing minimal assistance if at all. Levels 3 through 5 consist of increasing levels of automation from partial (in which human intervention is needed in edge cases) to total (where human intervention is not needed at all).

Manufacturers have not yet developed the capabilities to program enough common sense into vehicles to enable them to handle edge cases.

What are edge cases?

Edge cases are a big reason self-driving cars aren’t roaming the streets. These are situations in which drivers must take control to handle the outcome. For example, a flood makes the road impassable, a deer or squirrel darts in front of your car, etc. 

We learn how to deal with these situations through practice and reasoning skills.

For example, if you see a ball roll into the street, you know enough to slow down and keep a lookout for children nearby. You become more alert when the car in front of you starts swerving dangerously to the sides because you suspect the driver may be distracted by their phone or intoxicated

If only it were this easy to code this level of contextual thinking and interpretation into machines, computers, and cars.

Self-driving car manufacturers have to anticipate every possible scenario and code for it. Machine learning, which is the foundation of AI, can only adapt to the extent that manufacturers can simulate scenarios for it to learn from.

And there are lots of unexpected scenarios. Everyone always seems to have at least one crazy driving story.

Given that there are estimated to be about 1.4 billion drivers in the world, you can see the scale of the mission ahead. Until these edge cases are effectively coded for, you can expect to continue seeing cars that are only partially self-driving that are still heavily dependent on human input and supervision.

Will we ever see fully self-driving cars on the road?

Fully autonomous (Level 5) cars are currently being tested in various parts of the world, but none are available to the public. We are still years away from that due to edge cases and challenges standing in the way which range from legislative and technological to environmental. 

Should you look forward to self-driving cars?

There are many benefits that self-driving cars promise should they materialize. For example, they could provide the elderly and disabled with greater mobility. 

Other benefits include: 

• Reduce transportation costs by 40% 
• Reduce traffic congestion by 30% 
• Improve walkability and livability
• Redefine the concept of parking spaces by freeing them up for other users
• Cut urban CO2 emissions by 80% worldwide

Share and comment!

What scares you (or excites you the most) about self-driving cars cruising up and down your street? Comment your thoughts below and don’t forget to share this post with someone who would appreciate the knowledge!