We come into contact with it daily through emails, social networks, search engines, online shopping, smart assistants, and much more. Artificial intelligence has already changed our lives considerably and will continue to do so in the years to come. Here is a mini glossary to introduce you to some of the key terms of this thriving technology.
The term artificial intelligence and its abbreviation “AI” are most often used to refer to a machine or software whose operation is based on the cognitive faculties of human intelligence. An AI must be able to learn, adapt and change its behaviour. This technology has made great strides in recent years by improving computers’ computing capacity, developing appropriate algorithms and making Big Data available. Deep learning is the most promising branch of AI. It should be noted that in its broader sense, AI also refers to the theories developed in this field.
An algorithm is a set of logically ordered operations that a system must follow to achieve a specific result or solve a problem. The algorithm transforms input data into output data, the latter representing the solution. We could compare the algorithm to a recipe since it includes:
- data (the ingredients or kitchen accessories to be used);
- the procedure to follow to achieve the desired result (e. g. cutting a vegetable in a particular way, returning it, etc.);
- the result (the finished meal).
In computer science and technology, a computer program consists of a set of algorithms.
If the algorithm is a recipe, the Big Data are the ingredients. They correspond to all the information we produce every day with new digital technologies. It can be said that they result from the meeting between the Internet, social networks and intelligent devices.
As a sub-domain of artificial intelligence, machine learning is a process by which an algorithm can improve itself as it achieves results. Depending on the type of learning used by algorithms — one can also speak of an “algorithmic model” — machine learning can be supervised, unsupervised, semi-supervised or reinforced.
Sub-domain of machine learning, deep learning is also its most advanced form. It can be applied in a supervised, unsupervised, semi-supervised or reinforced manner. Deep learning makes it possible to process Big Data using an artificial neural network inspired by the one of the human brain. Data processing is done through several layers of these artificial neurons that encode them according to an increasing level of abstraction:
- the first ones encode the main characteristics;
- intermediate layers, the specifications;
- the last layers, the details.
Deep learning can produce predictive models not only from a large amount of data but also without it being structured and labelled. While artificial neural networks are also used in machine learning, these networks include more neurons, layers and interconnectivity when it comes to deep learning.