19 Aug Artificial Intelligence Technologies and their categories
Our previous article explained what Artificial Intelligence is. If you have not read it yet, I recommend doing so, but if you do not have much time you can watch the following video as a summary. As a continuation, today we are going to explore the generic set of Artificial Intelligence technologies and their categories.
ARTIFICIAL INTELLIGENCE TECHNOLOGIES
There are many technologies and disciplines that involve Artificial Intelligence, which have their own branches of mathematical and engineering study. Let’s take a look at the most relevant technologies, starting with recognition systems through to machine learning systems.
Automatic speech recognition
Automatic speech recognition is a discipline belonging to acoustics which recognizes phonemes in a voice signal. The voice recognition systems process the signal collected by a microphone to identify the words pronounced by the user.
Natural language processing NLP
While speech recognition focuses on purely conversion of voice to text, Natural Language Processing NLP is a discipline that is more closely linked to the field of linguistics, and its objective is to understand what the user means when making a certain command, question or statement (either written or vocal) and what he expects to achieve. In addition, it analyzes the mood to find subjective patterns. In short, it is the field that helps communication (mainly sound and written) between machine and human.
Visual recognition is the discipline based on processing an image or video signal, with the aim of recognizing patterns, shapes, and in the best cases, accurately identifying the different elements in an image.
Text recognition could be considered a part of visual recognition, as its main objective is to recognize and identify text in image formats. It is common to use OCR (Optical Character Recognition) tools for this work.
Without going into technicalities, Big Data can be considered as a large volume of data. Big Data alone is not a technology but having a huge amount of data (preferably structured) available is vital for achieving objectives both in Business Intelligence analysis and in the application of certain Machine Learning algorithms.
Expert systems are those which contain all possible human knowledge about a particular topic. A classic example are the systems that play chess, which use a whole collection of movements and strategies, that have been input in their memory, to determine the best move (usually based on decision trees).
Robotics (either mechanical or robotic software, such as RPA) covers a wide range of devices. Whenever a system or robot shows signs of intelligence, for example, being able to make decisions, however basic they may be, we can be talking about Artificial Intelligence. Remember that AI does not have to be especially sophisticated, it exists at all levels, even the most basic ones, and it must be differentiated from the ability to learn from machines; that is, Machine Learning.
Machine Learning is the discipline, within Artificial Intelligence, that tries to get a system to learn and relate information the way a person would. To do this, it uses algorithms that are able to detect patterns in previous data, being able to create future predictions, as well as new trends such as Deep Learning and its neural network algorithms.
Deep Learning is a subdiscipline of Machine Learning. It is a learning system that is inspired by the functioning of the neural networks of the human brain to process information, with a very complex mathematical basis. Although it does rely on experience (whether previous data, generated by the environment or self-generated), it does not start from strict indications that determine what is correct and what is not, so the system can determine conclusions on its own.
Cognitive Intelligence is a combination of the previously mentioned technologies with the aim of creating artificial intelligence services capable of having human understanding. It is the union of visual recognition, sound, reading comprehension, NLP and Machine Learning to create systems capable of understanding information related to human interaction and responding accordingly.
ARTIFICIAL INTELLIGENCE CATEGORIES
It is not easy to categorize artificial intelligence and the truth is that it is best practice to categorize it based on the algorithms used by a particular system. However, some experts have tried to create artificial intelligence groups based on their approach.
According to computer scientists Stuart Russell and Peter Norvig, artificial intelligence can be divided into the following categories:
Systems that think like humans
These systems try to emulate human thought quite literally using artificial neural network models.
Systems that act like humans
These systems focus on acting as humans; They are more linked to classical robotics and are less flexible.
Systems that think rationally
These systems try to apply human logic when it comes to perceiving, reasoning and acting. They are not focused on emulating the neuronal behavior of the brain but are trained to act in a human way in a given environment. An example of this is expert agents.
Systems that act rationally (ideally)
They try to emulate human behavior in a rational way, obtaining their own conclusions to given environmental conditions. The differential point in these systems is trying to apply rationality to their decisions.
A more common categorization is one that divides 2 large groups:
Weak (or narrow) AI
Known by its acronym ANI (Artificial Narrow Intelligence), and although the name may seem somewhat derogatory, it covers all the Artificial Intelligence in existence today. It is Artificial Intelligence dedicated to solving a specific or set of problems in an optimal way, but without the possibility of extending to general problems without the relevant programming. Even the most advanced virtual assistants fall into this category.
Strong (or GENERAL) AI
Known by the acronym AGI (Artificial General Intelligence), it is Artificial Intelligence capable of matching or surpassing human intelligence in the capacity of reasoning and deduction. Today it is a utopia only existing in science fiction because although the machines already outperform humans in a multitude of capacities (including vision and auditory recognition in some areas), they do not have real feelings, native cognitive abilities, self-awareness or the ability to adapt to any scenario.
This post is the second installment of a series of articles that will go deeper into more technical and valuable aspects for the application of Artificial Intelligence in a company. In the previous post we explained what Artificial Intelligence is, and in this article, we have given an overview of several categories and technologies of Artificial Intelligence.
Coming up in the next article we will see what type of algorithms are the most used within AI, what applications we can find in the market today and what benefits it can bring to businesses, so I encourage you keep an eye on the blog page to not miss any news.
To learn what Artificial Intelligence can do for you using a BPM as an orchestrator, we recommend you consult the following page: