What is Artificial Intelligence?

Artificial Intelligence is a technical term that we use to describe a set of programs that make machines think, in a very loose sense of the word. Whenever people hear about Artificial intelligence or Machine Learning, their minds jump to Skynet and the Terminator, and that is usually due to a lack of understanding and the proper explanation. If we want people to lose their apprehension to Artificial Intelligence, we have to change their distorted concept of what it is and how it works.

Before, I said that Artificial Intelligence is a way we have to say that there is a computer program that helps the computer think, and while it is technically true, the machine will not learn how to deal with lots of different situations; Instead, it will become incredibly good at one kind of task. So, contrary to popular knowledge, Artificial Intellience is not used to create human-like behaviours in a computer, but to teach the computer how to be a tool that you can use to solve one particular problem.

But how does it work?

There are lots of ways to train Artificial Intelligence machines, the key word in this sentence being "train". Machines, just like humans, need to repeat a task over and over again in order to learn how to properly do it. The repetition helps us prepare for any random eventuality that may happen, and helps us learn how to react to any unexpected outcome in the task. So when it comes to learning, machines are like people: they need practice.

Our objective here is to train a machine to diagnose patients with heart conditions from looking at their magnetic resonance imagings. So, to train, it will need to look at some magnetic resonance imagings with established diagnosis. Keep in mind, the more practice our machine gets, the better it will be at performing the task, just like a person in real life, so if you give the machine 40 magnetic resonance imagings, its predictions will not be as precise nor as accurate as those of a machine that has had the chance to learn from 80, 100 or 200 magnetic resonance imagings. Of course, the more images are given to the machine, the longer it will take to train it. The key is to find a compromise between time and effectiveness.