The Blog

Who Gives A Truck? Are Self-Driving Vehicles Now Science Fact?

Why, then, if the first experiments in AI took place in the 70's and 80's are we not all driverless now? The simple answer is not that we have necessarily hit upon a new idea, but instead that the computing power to build the necessary "brain" just has not been availabl
filo via Getty Images

I still harbour a desire to own an E-Type Jag. And while I may one day just manage that, I may never be allowed to drive it.

"Self-driving" or Autonomous cars have been in the news a lot over the last few years, with trials seemingly springing up all over the world. Google will no doubt take the credit as being the pioneers of the recent resurgence, but the stalking horse may well be Uber who, if they are successful, may not only change the way we think about driving, but about the very notion of car ownership itself.

Let's take a step back as to why science fiction has suddenly become science fact.

Few futuristic movies of the past 50 years have not had some form of shiny, angular, somewhat wobbly car that transports passengers from one place to the other in response to a barked command. Not unusually the car fails to understand what is being said for mildly comic effect, most memorably for me the Johnny Cab driver in Total Recall, who when asked by a confused Arnold Schwarzenegger, "How did I get in this cab?", replies "The door opened, you got in". He then rolls his eyes, in a mock-human gesture and says "Hell of a day, isn't it?"

And in building the first self-driving cars, it is the "mock-human" element that has been lacking, until now.

Early attempts at autonomous cars in the 50's (although inspired by pre-WWII ideas), focused more on the environment than the car, by building a system of roads that could guide the car, rather than a car that could navigate existing roads. Magnets embedded in the surface would guide the cars, and radar would control distance between them. An idea doomed to failure without massive investment in infrastructure.

Only with the advent of "Artificial Intelligence" or AI did the basics of true autonomy become clear. Autonomous driving requires perception of the surroundings, reasoning to determine the conditions of the environment, planning the best course of action, and continuously learning to improve our understanding of the vast and diverse world. Like humans, you have to be taught the basics and then learn from experience.

Why, then, if the first experiments in AI took place in the 70's and 80's are we not all driverless now? The simple answer is not that we have necessarily hit upon a new idea, but instead that the computing power to build the necessary "brain" just has not been available. Endless comparisons can be made between the computing power used by NASA to put a person on the moon and the power of modern smartphones, but it does highlight the simple issue that until recently, we had all of the theory, and none of the practical tools, required to build a car that could drive itself.

The process required to teach a car to drive is called "Deep Learning" and the company behind it is NVIDIA. You have heard of Intel, Uber and Microsoft, but unless you are a hard-core gamer, NVIDIA may not be familiar. The chances are, though, that at least one device you own has an NVIDIA graphics card in it: The laptop I am writing this on used an NVIDIA chip to power the screen. And it just so happens that the parallel architecture that is needed to process millions of pixels simultaneously also mimics the human brain by being able to perform millions or even billions of calculations at the same time, exactly what you need if you are going to avoid a pedestrian jumping out unexpectedly.

NVIDIA has quietly got itself behind many of the world's autonomous driving projects, and while they may not be as high-profile as Google, they are working with something like 80 auto makers to bring self-driving cars to the streets, not in the future, but in the next five years. Already legal in some US states, the UK government has already committed to a "Modern Transport Bill" to allow autonomous vehicles on Britain's roads.

But like the Field of Dreams, will people really come round to the idea of driverless cars just because they exist? The adoption of electric vehicles should be a lesson to auto makers. The London Evening Standard reported an enormous rise in electric vehicle registrations in London in April of this year, a whole 22% year on year: to 800 in Q1 2016. To put this in perspective, 60,000 electric cars were registered in the UK in that period, against a total of over 900,000 cars, and that against a background of "Plug-in" grants, free or cheap charging points and no congestion charge in London.

Autonomous vehicles have an altogether bigger hurdle, as people do not trust a machine to do the driving for them.

I believe that the answer will, at first, lie in the logistics industry. A self-driving truck never tires, need only stop for fuel, and if it can be proven to work safely, has no passenger to worry about the computer taking control. And Otto, a company owned by Uber, has just completed the first ever commercial delivery using a self-driving truck, a 120 mile beer run in Colorado. Complete several billion miles of deliveries safely, and you give the public reassurance in the safety of the technology: Allow people currently "off-grid" for home deliveries the chance to get same-day autonomous deliveries from Amazon, and see their perceptions change.

From there, other pressures apply: Uber's main cost is its drivers, so if you want to be driven cheaply around a city, a self-driving taxi may be your only option. As insurers see that self-driving cars are much safer than humans, insurance premiums will go through the roof for "normal" cars, pricing me out of driving my mythical E-Type anywhere but in my dreams.

But the big change that I foresee is the way in which we own cars in the future. For most people in reach of even a small town, there will be a pool of available vehicles. Want to get to work? The one-seater pod will arrive. Need to get the kids to school? Stick them in a four-seater. Groceries? Those will already be delivered autonomously so why go to the store?

Auto makers need to get wise to this shift. They won't be selling cars because of their colour, or their top speed, or their gas mileage: They will be selling them to the company or companies that control huge fleets of self-organising, self-driving cars. They run the risk of being a contract manufacturer of vehicles designed by the Ubers of the future. But they have the chance to seize that opportunity for themselves, but not just building self-driving cars, but by thinking about how we will consume them: Only then can they survive as more than niche providers. Whether we as consumers care, to paraphrase Henry Ford, that cars are only painted in Uber's trademark black remains to be seen.