What is a large language model (LLM)?
A large language model (LLM) is an AI system that learns from text data (words, characters, sentences) to respond to an input in a way that resembles a human being.
In general terms, language models are programmed to predict text. The system learns to calculate which character, word, or sentence is more likely to follow the one that the user has given as an input in a specific context. This type of AI is used in applications like SMS text messaging, Microsoft Word or Google Docs, which suggest words or phrases as a person is writing.
LLMs are language models with hundreds of millions or hundreds of billions of parameters, which means that they have learnt hundreds of millions of possible connections within text data. These connections are established by combining words and using mechanisms to learn and analyse the context of a text.
LLMs are the basis for most foundation models we see today. They can carry out a variety of tasks and be ‘built on top of’ by multiple text-based applications that can, for example, reply to questions, complete sentences, describe images or identify objects.