Main branches of Artificial Intelligence

Artificial Intelligence (AI) derives from a series of models or branches that can be used in different areas of people's lives as well as in the professional world to automate processes that, carried out in the traditional way, can be unproductive, inefficient, slow or costly.

Each subfield of Artificial Intelligence is constantly changing and being updated. Therefore, it is expected that many more will appear and that over time others will become "obsolete", cease to be used, or bring out improved versions without eliminating that type of AI (as is the case, for example, with the Deep Learning and Machine Learning).

Exploring the main branches of Artificial Intelligence

Machine Learning

It is focused on the development of algorithms that allow computer systems to learn from previously provided data. We can observe different types of machine learning.

1. Supervised Learning. In this type of learning, the machine learning model is trained on data that has known correct answers. The goal is for the model to learn to predict the correct labels or answers for new data. Such as, for example, the classification of emails as spam or non-spam or image recognition.
2. Unsupervised Learning (Unsupervised Learning). In this case, the machine learning model is trained with data that no have known answers. Its purpose is that the model learns to find patterns or behavioral structures in the data. For example, the segmentation of customers into similar groups or the detection of anomalies in the data of the different variables available.
3. Reinforcement Learning. Finally, this learning model is trained through a trial-and-error process, where it receives feedback in the form of rewards or punishments based on the actions it takes in an environment. The idea is that the model learns to take the correct actions in different situations to maximize the total reward. Applications of reinforcement learning include automation of complex tasks in robotics and strategy games.
4. Neural Networks. They are information processing models inspired by the structure and function of the human brain, which are used in machine learning and computer vision. They can also be classified into different types:

• Feedforward neural networksThey are useful for classification and regression problems.
• Recurrent neural networksThey are useful for prediction problems, such as time series analysis.
• Convolutional neural networks: They use convolution operations to extract important features from the input.
• Neural networks of long-term memory (LSTM): designed to model long-term dependencies. 
• Neural Networks Transformer or GPT (Generative Pretraining Tranformer)Recurrent and convolutional neural networks are a variant of recurrent and convolutional neural networks, useful for problems of natural language processing, machine translation and object recognition.
• GAN (Generative Adversial Networks): Antagonistic Generative Networks employ deep learning models to generate and/or manipulate images, photos, videos and audio editing. They essentially rely on an algorithm based on a system of two neural networks - the Generator and the Discriminator - that compete with each other.

▶ You may be interested in: The future of automation thanks to Machine Learning

Natural Language Processing

It is the branch of Artificial Intelligence which deals with the development of algorithms and techniques that allow computer systems to understand, interpret and generate natural human language. It combines computational linguistics, which is concerned with developing formalisms that describe the rules of human language so that they can be adapted by a computer, with deep learning models. 

Within natural language processing we find a number of models.

1. Virtual assistants. For example, Siri, Alexa and Google Assistant use NLPs to be able to understand the user's questions and requests and try to provide as coherent an answer as possible.
2. Machine Translation. Machine translation tools, such as Google Translate, use NLPs to analyze the source language and produce a translation in the requested language.
3. Sentiment analysis. It is used to analyze large amounts of text, such as social media posts or customer comments, to determine the general public opinion or sentiment on a topic or product.
4. Information extraction. NLP is used to extract valuable information from large text datasets, such as news, PDFs, legal documents or financial reports.
5. Text generation. Automatically generates consistent and relevant text, such as product descriptions, news and article summaries.
6. Categorical classification. In this case it is used to automatically sort text into predefined categories, such as news, sports or technology, for easy organization and searching.
7. Semantics. Enables more advanced and precise search engine searches, such as voice search or search for specific queries.

Computer Vision (Computer Vision)

It is an Artificial Intelligence that aims to emulate the way human vision works. AI allows computer systems to be endowed with lines of thought and artificial vision gives them the ability to observe and understand what they are looking at. viewing.

There are different types of machine vision used for different problems, let's see what they are:

1. Object detection. Detects and locates specific objects in an image or video.
2. Object tracking. It consists of tracking an object in time as it moves in a video.
3. Recognition. Used to identify specific patterns in an image or video, such as faces, shapes, textures, etc.
4. Ranking. Objects or images are classified into different categories.

Robotics

A branch of Artificial Intelligence that deals with the creation of robots and autonomous systems that can perform tasks independently. There are several types of artificial robotics that we will see below:

1. Autonomous roboticscan operate independently, without the need for human intervention.
2. Collaborative roboticsrobots are robots that work together with humans in the same space and can collaborate on specific tasks.
3. Service roboticsare designed to perform service tasks such as cleaning, delivery of food and beverages, etc.
4. Medical roboticsare used in medicine to assist in surgical and diagnostic procedures.
5. Industrial robotics: are intended for repetitive and hazardous tasks in manufacturing and production environments.
6. Exploration roboticsare robots used in space and terrestrial exploration missions.

Cognitive Robotics (Cognitive Robotics)

It is the branch of robotics that focuses on the development of robots that can process and understand complex information, and make decisions accordingly.

Expert Systems

These are computer programs that use specific knowledge and rules to perform specific tasks, such as decision making. They solve problems in a specific domain. The types of expert systems that exist are:

1. Case-based. They use a database of previous cases to solve new problems. These systems search for cases similar to the current problem and adapt the solution from those cases.
2. Based on neural networks. They learn from input data and infer an appropriate output. 
3. Based on fuzzy logic. They use fuzzy set theory to handle uncertainty and vagueness. These systems are useful for problems where the answers are not binary (yes/no), but are on a continuum and even ill-defined, e.g., can handle predicates of the type: "if the temperature rises, then the valve opens a little".
4. Based on genetic algorithms. They use an optimization technique inspired by biological evolution to find the best solution.
5. Based on intelligent agents. Intelligent agent-based expert systems use software agents that can interact with the environment and make decisions.
6. Rule-based inference engines. Starting from a set of axioms and a set of rules, they are able to infer new rules. In this way, they make it possible to act accordingly in the face of contexts and/or issues that were not previously addressed.

Gamco is a pioneer company in the creation of solutions for Artificial Intelligence software for business optimization and advanced risk management by developing predictive models and optimizing actions.