November 28, 2018
It seems like within a span of just a few years, the rapid evolution of AI allows technology, once only seen in science fiction movies to be soon witnessed on our roads. From self-driving cars, to a car that assesses its damage for maintenance purposes, the automobile industry is investing heavily in the power of artificial intelligence to make cars safer and more efficient.
There is a common misconception that the technology we see today was introduced in this decade. However, the concept of a self-driving car can be dated back to the 1920s. In the year 1925, an automobile inventor created a radio-controlled car; he demonstrated the car driving through Manhattan using a transmitter from a second car and also showed the car starting its engine, sounding horns and shifting gears. Likewise, a computer scientist theorized in the year 1969 that cars of the future would be computer-controlled, and could act as “automatic chauffeurs.” A passenger can input their destination using a keyboard, and the car will autonomously drive to that destination. The passenger can also command the car to take stops or detours using the input device.
While a car with these exact specifications is yet to be designed, similar attributes can be expected to make an appearance in cars of the near future. Features such as parking assist have already existed since the early-to-mid 2000s whereby the car can steer itself into a parking spot with little to no assistance from the driver. The next avenue to be explored by car manufacturers was, not unexpectedly, driver assist. Cars equipped with driver assist can accomplish several tasks, such as emergency braking in the event of an impending collision, or cross-traffic detection whereby the car detects any cars are approaching yours and give a warning signal. Some cars come with the capability of driver-assist steering, in which case the car takes control of the steering wheel for reorientation and stabilization of the vehicle.
While these features are indeed impressive, they aren’t quite “science fiction movie” territory. This is where driverless cars come in. It is expected that by the year 2019, completely driverless cars will become commercially available; also, an AI company that develops self-driving cars is planning on deploying driverless taxis in Singapore. The science that has allowed us to accomplish this feat is known as deep learning, a method of machine learning that utilizes algorithms which are inspired by some of the functions of the human brain. The system that a driverless car runs on will need hours and hours of training data before it can be considered safe to be deployed. Driverless cars also need to be equipped with cameras, radars, GPS, sonar and sensors for the software to be able to control the vehicle.
In addition to this, AI is being used for predictive maintenance which is a technique designed to predict when equipment requires maintenance rather than performing preventive maintenance, which can be costly and purposeless if the equipment is not in need of maintenance. It saves both cost and time, and one of the latest functions being added to this kind of maintenance is the use of a cloud service — an OTA (Over-The-Air) software updates and transfers operational and diagnostic data from systems and components of the automobile. In layman’s terms, an OTA software allows the software and systems on the car to be updated or repaired remotely, the same technology used by smartphones. This kind of function can have AI detect the most minute changes and automatically provide a solution if it can be repaired from the cloud, or alert the driver of impending failure. Facial recognition and driver identification will also be a significant function of future cars. The same technology can also be used to monitor the driver; studying head position, posture, eye gaze, and openness to determine whether the driver is attentive or distracted and in the case of the latter, alert the driver to keep their eyes on the road.
While all of this information sounds extremely promising and exciting, some scientists fear an “AI Winter” will significantly impair progress in the development of self-driving cars. Some experts and skeptics have doubts about whether self-driving cars will be safe enough to take the road by 2019. Their doubts are a result of some accidents caused by prototype models of self-driving vehicles. The issue encountered by the AI system was, in each case, a scenario had taken place that had not already been previously learned. As a result of this, the system was unable to detect and conclude what action needed to be taken accurately. The assumption that engineers will cover all “generalizations” is why many experts believe AI still has a long way to progress before it can be implemented appropriately, as it is not plausible that every unforeseen circumstance is taught. Conversely, scientists and engineers are optimistic that once self-driving cars in service, upgrades and improvements can be made with each edge-case that is encountered. Airplane technology and safety, for example, evolved steadily due to the intensive inspection of each accident.
Despite some hesitation on behalf of a small group of people, it appears that fully autonomous, futuristic self-driving cars that run on highly advanced technology will soon become the norm. The benefits of such a system heavily outweigh the downsides, and this technology can only be expected to evolve even further.