Large Language Models (LLMs) are artificial intelligence systems designed to understand, generate, and respond to human language. They are termed large because they contain billions of parameters that enable them to process complex patterns within language data.LLMs are a subset of machine learning models capable of understanding and generating human-like text. They acquire these abilities by learning statistical relationships from text documents during an intensive computational training process that involves self-supervised and semi-supervised learning. They are trained on vast datasets, often collected from the internet, which can include diverse textual sources such as Wikipedia pages, books, social media topics, and news articles. Large language models emerged around 2018 and exhibit high performance across a wide range of tasks. The internal architecture of these models is based on transformers. Transformers are artificial neural networks that rely on attention mechanisms to process long sequences of tokens and typically have tens of millions to billions of trained parameters. They can capture dependencies and relationships between words and sentences, as well as syntax, semantics, and context in natural language, making them suitable for generating text. In this context, generative AI takes an input text and predicts the next token or word repeatedly to produce human-like text.As of March 2024, the largest and most powerful ones are built using decoder-only transformer architectures. Meanwhile, some other implementations are based on architectures such as different types of recurrent neural networks and Mamba (a state space model).