This is a glossary of terms related to artificial intelligence (AI). It is constantly growing and a non-exhaustive list so stay tuned for more terminology as it advances.
affective computing
A form of computing that relates to the recognition, interpretation, and simulation of human emotions.
algorithm
A set of rules that are followed, usually by computers, in the aid of problem solving.
artificial general intelligence
Also known as AGI, artificial general intelligence describes a type of AI that can understand or learn any human task. This is not to be confused with the AIs that are currently in the news.
artificial neural network
A network of artificial neurons or nodes where their connections are studied and used in areas such as predictive modeling and speech recognition.
automated reasoning
A branch of computer science that studies reasoning within computing, allowing for this process to be automated.
clustering
A machine learning task where data is grouped by their similarities. This is used to improve AIs that look for patterns and allow them to make more accurate predictions.
few shot learning
A machine learning method where there are limited training samples to use
in-context learning
A machine learning method where language models learn tasks with limited input-output examples and without improving parameters
knowledge representation
A branch of AI related to conveying information about the world to a computer system so it can solve complex problems. This is an important element in large language models so they can produce accurate outputs and mimic dialogue when given prompts.
machine learning
Also known as ML, this is a field of AI where computers ‘learn’ to complete tasks efficiently using a variety of methods and data.
natural language classification
A natural language processing problem where text is categorised based on the words or ‘entities’ within them and their relationships to a given set of categories.
natural language processing
A branch of AI where machines learn and interpret context within written and spoken human language.
natural language understanding
A branch of natural language processing dedicated to machine reading comprehension.
recommender system
Also known as a recommendation engine, this system filters information and offers suggestions that a user would find helpful, based on their needs. For example, when streaming content on Netflix, a recommendation engine suggests other series or films to watch based on what they’ve watched before, what is popular, or the category of media they’ve otherwise rated.
responsible AI
A form of AI that aims to build and deploy systems and devices that will help the world. This means a strict focus on the ethical impacts of AI and reducing noise and bias that could harm people and the environment. Although the terms aren’t used interchangeably, ethical AI follows similar processes for similar causes.
sentiment analysis
A natural language processing (NLP) problem that attempts to determine whether a given dataset or data points expresses positive, negative or neutral opinions.
text-to-image model
A machine learning model which usually combines a language model and a generative image model to convert natural language into a generated image. Examples of text-to-image model’s include OpenAI’s DALL-E 2, Google Brain’s Imagen, and StabilityAI’s Stable Diffusion.
unsupervised learning
A form of machine learning that uses algorithms to analyse and group together unlabelled datasets.
zero-shot learning
A machine learning problem related to classification, where the machine has to predict if data goes into a class it hasn’t learnt about.