Our modern cell phones are another example. When you take out your phone to unlock it, it scans your face with the camera and recognizes it as you using AI. When you want to photograph something with your mobile, AI is used to figure out what it is you want to photograph so that the picture is the best, and when you search among all your old pictures in your mobile, AI is used to figure out what each picture represents. When you receive email, AI technology is used to figure out whether it is spam or not.
AI is short for artificial intelligence. It’s a collective name for a range of different technologies that try to make computer programs work more like a human. It is, for example, about being able to reason and plan, understand common language, be able to learn from data, recognize patterns and put together several different impressions to arrive at a conclusion.
An important prerequisite for AI is something called machine learning. In a regular old-fashioned computer program, a person has to type in everything that the program should be able to do – that is, program it. The program acts based on a clear system of rules written in its code, and when something is not in the code, nothing happens in the program. These rules are called algorithms. An algorithm is simply a series of instructions that describe how, with what and in what order a task should be solved.
Machine learning is a method where the program is instead allowed to create a rule system on its own by feeding it with a large amount of data. A classic example is image recognition. The reason why the picture library in your computer can find pictures that represent dogs is that the system has been trained to recognize dogs in pictures. When you talk to the mobile phone’s voice assistant, AI is used to figure out what you’re saying.
You may have fed it a few hundred images and told it that these images represent a dog. Then the system has to independently look at millions of pictures, and try to find other pictures of dogs by comparing them with the first ones. In this way, the computer system quickly becomes quite good at recognizing dogs. Machine learning is a very important component for AI. Instead of writing direct programs with instructions, we make programs that can learn by example or by recognizing patterns.