A.I. and Deep Learning in Vehicular Automation. Will Self-Driving Cars Ever Take over the Road?

A vehicular camera identifying objects and their location

As chip developer NVIDIA enters in the continually growing arena of self-driving vehicles by partnering with both ride-hailing giant Uber and car manufacturer Volkswagen, it’s clear that the potential for the technology is quite high. Promising a more intelligent and safe means of transportation, automated driving could present a massive social and economic benefit in the near future. A great deal of inspired supporters as well as a network of huge business players are creating a favorable outlook on removing the human-error in driving. However, with something as sensitive and dangerous as today’s roads are, there could be many complications that arise and concerned drivers, both personal and business, are speaking out against this futuristic possibility.

Since 2018 is going to be the year of data analysis and machine-learning, the rapid development in the automated driving industry isn’t surprising. It’s all about information and using that information to improve software. Uber has been testing their already-implemented self-driving fleet in Pittsburgh since 2016. As stated by the company, data is essential for self-driving vehicles, relying on sensors and 360 degree cameras constantly mapping the environment around them and identifying approaching objects.

Artwork from a scene in the movie "Minority Report"

Driving imposes an ever-changing environment, laden with humans affected by emotion, careless construction zones, various bottlenecking locations, rush hour and sporadic high-volume traffic. Throw this in the mix with humans and computers driving these same roads at the same time, and you can get into some serious problems. How can a computer anticipate an enraged motorist? What happens when you are on the highway and a computer avoids an incident in their environment which inadvertently affects a human driver nearby. Their reaction time is mathematically nowhere near that of its computer driver counterpart, so now they are negatively impacted by the computer’s decision. This yields an extreme need for solid data analytics and coupling that information with A.I. to safely control vehicles in a changing world.

The biggest questions to ask are “Will there be a time when every driver is replaced by self-driving technology? What is the overarching result of this industry? Is this innovation attempting to remove human-error across the board? Is this even a financially attainable prospect?”

A scene from the movie, "irobot"

My answers are as follows: No. More intelligent cars that can assist in avoiding negative outcomes. No. And somewhat.

The practicality and financial feasibility for every individual to own a vehicle that is completely automated by A.I. is quite improbable. It’s much easier to argue that it would be a transition into a mixture of manually driven and semi-automated vehicles. The result will yield a more responsive vehicle that can identify potential hazards and collisions and assist in avoiding them. There is no way to remove the human-error associated with driving, because there will always be manual drivers. The only way we could achieve perfection is if the road in it's entirety would be automated and every vehicle would be completely controlled by a computer. That way, there would be no complications with trying to overcome the errors of an erratic human. However, it wouldn’t be financially attainable for everybody on the road. Some just simply wouldn't be able to afford a fully self-driven vehicle even if mandated by state or federal governments.

Even then, could you imagine that catastrophe if a network of cars, now all self-driven, were breached or overridden? Or the moment a car's sensor malfunctions and doesn't detect the red light, doesn't stop, and hits a pedestrian... game over. There is nobody to blame. Nobody to jail. The only thing left will be the fear of the massive weapon that is the car.

A fully automated, unmanned vehicle in a collision

Artificial Intelligence and machine learning is invaluable for so many future technologies. To use it as an objective to produce fully automated vehicles may prove to be too cumbersome, expensive, and dangerous. All we can do now is wait and see how the future will improve the automotive industry. My hypothesis is that it will improve sensors and vehicle responses to avoid incidences as an assisted measure... but not to replace the driver entirely.