Are we less than 20 years from not only the end of widespread private car ownership but also the internal combustion engine that has been powering motors for the past couple of hundred years? A majority of technology analysts now think so.
The major catalyst to the first part of the change, the practical extinction of the privately-owned car, is expected to be AI-powered driverless car technology. Deep Learning developments in AI mean that we are within a few years of road-tested automated vehicle technology.
Deep Learning is an AI methodology which uses neural networks, a structure to Artificial Intelligence the mimics the biological structure of how our brains work. Information, such as the image of a road sign is, as our brains process information subconsciously and at lightning speed, broken up into elements such as shape, size, colour, symbols and words. Each element is assessed as the visual data is passed through ‘layers’ and matched with stored information.
Octagon, white rim, red centre, white letters, STOP spelled. When each of these elements is isolated by a layer of the neural network, the end result is a very educated guess what we are seeing is a STOP sign. Deep Learning neural networks work in the same way. Until recently, however, they were brittle and imprecise, requiring perfect conditions. Now, Deep Learning neural networks of the kind being developed by driverless cars can cope with ‘real life’ variations of perfect conditions such as fog or a branch partially obscuring something detected by computer vision. Judging where edges stop and start and knowing what a branch looks like so being able to combine it with a STOP sign means that AI is no longer thrown off the scent.
Mistakes are also ironed out by the ‘deep’ in Deep Learning, which is AI algorithms being fed massive banks of data which they learn from. That data means the algorithm has almost certainly encountered a STOP sign obscured by fog or a branch before. Or at least fog and a branch separately so they can be picked out as the information passes through the neural network’s layers. The speed information is processed at, necessarily practically instantaneous for driving, is now possible thanks to the latest technology in the world of GPU processors.
Companies such as Uber and Alphabet-owned Waymo, as well as car manufacturers such as BMW, Daimler, GM and Tesla, are pouring resources into the development of driverless cars and are almost there. When the technology is perfected, a development expected within the next couple of years, ride-hailing apps such as Uber, and rivals from companies such as Tesla, will start to send driverless vehicles. This comes with the caveat of potential delays around legislation and regulation catching up with the technology. Driverless vehicles will slash costs and turn private transport into a subscription or pay-as-you service that will mean private car ownership simply doesn’t make economic sense anymore.
As well as being driverless, the next generations of cars will also be fully electric rather than petrol or diesel powered or hybrids. Electric engines have developed significantly over the past decade. While they are still more expensive than internal combustion engines but costs are falling and will continue to do us. Running costs are also much less expensive than petrol and reliability and potential lifespan superior. Electric engines contain around 20 moving parts to 2000 in an internal combustion engine.
The combination of cutting the costs associated with a human driver as well as for fuel, with the assumption electric car prices will come into line with those of cars with internal combustion petrol engines today, could mean that within several years the cost of an Uber ride could drop to 10% what it does today.
It is forecast that within ten years of driverless cars freely hitting the roads, 95% of miles driven won’t involve a human driver. The combination of a more attractive economic model and greater convenience of transport-as-a-service with the distinct possibility that human drivers could even be legally forbidden due to the danger represented by human error could well mean that anyone who has recently bought a new car outright and plans to own it for 10 years could already own their last ever car.