ARTIFICIAL INTELLIGENCE VS MACHINE LEARNING : A CLARIFICATION

Artificial Intelligence vs Machine Learning

There is not a real difference between Artificial Intelligence vs Machine Learning.

Artificial Intelligence is, in general, the science whose goal is to optimize a machine to do better what it does, executing an algorithm whose aim is to minimize errors. This is possible through continuous measurements, by sensors, of the performance of the machine, or, in other words, of what the machine is doing and how well is doing it. This is why this science is called Artificial Intelligence: the machine is like to be conscious of what it is doing. But actually, it is not, there is nothing really intelligent.

We will see later some examples that will better clarify this introduction.

Machine Learning

Machine Learning is a branch of Artificial Intelligence. 

Machine Learning implements algorithms whose objective is to automatically find out the best answer to a request, namely, to answer with the minimum error. 

How? Based on data that we make available to the computer. Machine Learning requires a sufficient set of data, and this is a peculiar characteristic compared to other branches of Artificial Intelligence. We can say that Machine Learning is a data-based science, namely, a technology based on data that we make available to it. During the answering phase, the algorithm analyzes available data to find out the best answer to the request. The way these data are loaded on the computer does influence the performances of the algorithm. The more data are well ordered, the easier will be for the algorithm to find the best answer.

I know that this is not clear yet, but don't worry, we will do an example.

Machine Learning example: Google

The most famous example is Google. What happens when we digit some keywords on the search bar, or, in other words, we ask Google to show us some information? The search engine of Google, which is a particular mathematical algorithm, looks for contents (pages, articles, blogs, shops) in which there are words similar to the keywords that we are looking for. The algorithm selects the contents that fit better with our keywords and this is why we talk about Machine Learning. In other words, the algorithm iteratively minimizes the difference between the keywords that we digited and the words contained in the huge amount of pages inside its database. Finally, it shows the contents which better fits our keywords. Machine Learning refers to the fact that the machine is like to learn how to search in its database.

This search is influenced by many factors, such as the higher visibility of a page compared to another one (pages with more visualization will be shown in the first results). 15 years ago Google wasn't able to fit an internet search with the precision of today. Why? Simply because Google didn't have so much data as today.

Artificial Intelligence

Artificial Intelligence is a more general concept that comprises Machine Learning. Indeed, there is not only Machine Learning. Another diffused application of Artificial Intelligence is Robotics. How does an intelligent robot works? It is composed of three main components: sensors, computer, actuators. 

Sensors are the elements used by the machine to measure what it is doing itself (the receptors). The computer is the element that analyzes the measurements and decides which is the action to be executed (the brain). The actuators are the elements that execute the actions (the muscles). All the parts will be connected together by cables and electronics (the nerves).

Let make a simplified example.

Artificial Intelligence example: Robotics

Let consider a mechanical arm whose function is to take some objects on a table and to put it into a box.

Let install on the hand of the arm a sensor that can identify the position of the object (we can use a proximity sensor, optical or ultrasound). Once the object is identified, the sensor communicates to the computer the distance of the object, through an electrical signal. Let suppose that it is located 1 foot below. The computer analyzes the measurement, realizes that 1 foot is left and sends an electrical signal to the actuator (an electrical motor for example) to make the arm move 1 foot below.

Now, let suppose that for some imprecision of the machine, the object is not perfectly reached, namely, the distance is still not zero. The sensor measures the distance: 2 inches are left. Again, the information is sent in the form of an electrical signal to the computer, that analyzes and make the electric motor move the arm. This process is repeated until the distance from the object is null, namely, the arm can keep the object.

All this is possible because the computer executes a specific mathematical algorithm able to calculate which signal has to be sent to the actuator to move the arm such that the error is lowered, iteration by iteration. These algorithms belong to the class of Error Minimization Algorithms or Optimization Algorithms. If you are interested in a deeper analysis, typically adaptive filters and Least Mean Square algorithms are used. An extended bibliography regarding these algorithms is present on the web.

Of course, analogue-to-digital converters are needed, so that electrical signals (analogue) from sensors can be converted, transformed, to computer language (digital). In the same way, digital-to-analogue is needed, so that the information in computer language used by the algorithm can be converted into an electrical signal that makes the electric motor work.

Conclusion

In this article a general idea of what Artificial Intelligence and Machine Learning are is given, making some examples to better understand how they work. You should have understood that saying Artificial Intelligence vs Machine Learning doesn't make any sense, since they both share the same concept and working idea. The main peculiarity of Machine Learning with respect to others Artificial Intelligence applications is the necessity of data.

ScienceFull.