When you hear about Machine Learning, what do you think first? Is it about learning a machine? Well, if you guess that, then it’s not correct. Machine Learning is about how a machine learns. As one of the topics in AI, Machine Learning is used and applied in many aspect of life, such as automation, face recognition, and many more. So, how this machine learns? There are 3 types of learning, which are supervised learning, unsupervised learning, and reinforcement learning. supervised learning, as the word suggest, it is one of the learning where there is a “teacher”. It is usually guided, in form of what the machine will get and what is the desired output. Unsupervised learning, is a way of learning where computer only given data, without giving any specific input and output. This kind of learning usually finds the pattern in the given data. Lastly, reinforcement learning. This usually done dynamically, which means that the data is varies and usually have a specific goal to achieve, such as driving a car.
Machine learning has been widely used for many things nowadays, mostly that I know is autonomous system. If we see systems like Tesla’s Autopilot, Google’s self-driving car, and lately, Nissan’s Piloted Drive, it is the product of Machine Learning. These technologies enable computer to self-drive itself, bringing benefits to user and road safety. They can almost control everything and do the same thing as human drivers do. If people thinks that autonomous systems like this are too advanced and might be not able to be used here, in Indonesia, well there are few technologies that able to used here, though it is a semi-autonomous system, more like an additional safety feature. Few examples that I can give are Lane Keeping Assist (where this system warns and able to make corrections based on the lane mark, which literally keeping the car in a lane) and Autonomous Emergency Braking (where this system reads or sense vehicles or barrier in front of them and tries to warn the driver if there are impending danger or crash. If there are no action, then the system will take over and brake the car into full stop).
An interesting idea about Machine Learning that might be created is autonomous flight system for commercial aircraft. The basic concept is almost the same like autonomous driving, but in commercial aircraft, it is like an enhancement to the current autopilot system. I do realize that current plane autopilot system able to guide itself into destination and even lands itself. But this one with enhancement in safety and reducing pilot workload. How? Plane able to take sufficient manoeuvre while in emergency situations such as 2 aircraft being too close, failure on control, and even few situations that might able resulting in a crash. Taking data from the current aircraft safety features, such as TCAS (Traffic Collision Avoidance System) and GPWS (Ground Proximity Warning System). Data from two avionics system able to be used for computer decision making in emergency situations, which, supposedly, help pilots to manoeuvre and avoid crash.