People on borrowed time
Can computers learn from experience just as you and I do every time we get behind the wheel? The answer apparently is yes, so thinking cars are set to become science fact, not science fiction. And the reality is that cars may become better drivers than humans.
Tech companies such as Intel and Google and major car corporations, including Toyota, are sending cars in to real world driving situations to "learn" how to avoid accidents.
And the developments are timely, as some European countries - Spain for example - are drawing up lists of guidelines on restrictions that over-65s may face on their driving licences in the not-to-distance future. These include limits on mileage per journey, speed, driving on motorways and driving at night depending on medical condition. A fully trained car could come to the rescue.
Scientists are working on deep-learning frameworks which will evolve today's semi-autonomous cars into self-driving vehicles that can handle all traffic scenarios, in all environments, by learning from experience rather than being pre-programmed for every eventuality.
Deep-learning frameworks is a form of artificial intelligence, which scientists plan to build into on-board computers which can learn to make big calls, like choosing between hitting a bus queue or a stray dog.
The deep-learning frameworks technology gets computers to mimic neurons (cells in the human body specialised in carrying messages through an electrochemical process - the human brain has approximately 100bn neurons) by creating artificial neural networks to quickly absorb and pass on information on road and traffic conditions.
Toyota has invested $1bn into its Toyota Research Institute which has autonomous vehicles as a priority and Intel has acquired Altera, a maker of programmable logic devices with automotive applications for $1.6bn. Experts at both companies are developing autonomous cars that not only adhere to set protocols like keeping between white lines, but react to unpredictable scenarios too.
Toyota research expert Dr Gill Pratt says that part of the Institute work is to focus on augmenting machine learning and measuring the robustness of systems for all types of scenarios, while Altera's partner in the UK, i-Abra, is developing deep-learning frameworks for autonomous cars.
"Ultimately you are trying to replace human sense and action with artificial intelligence by acting on data collected from the car's sensors, cameras, radar and maybe audio," says managing director Ian Taylor.
"If bricks fall off a truck the idea is to stop the car as quickly as possible," he says.
"The car can learn from experience and then share the experience across a fleet that is constantly connected. The moral maze is presented every day to humans - do you hit the deer or the tree? If a human can make the decision, a computer can too."
Taylor admits that machines are fallible but far less so than humans. "The fully autonomous car would not only have more experience than a 17-year-old driver, but also his father. Autonomous cars will also save fuel, cut journey times and save lives," he argues.