Skip to Main Content

Artificial intelligence (AI)

This Research Guide will support your research and learning journey in Artificial Intelligence.

What is AI

The term AI (Artificial Intelligence) is a developing and contested term whose meaning changes depending on the context of its use. The term was coined by John McCarthy (in D Israel 1991), an American computer scientist and cognitive scientist, in 1955.

AI can refer to human-like intelligence displayed by an artificial entity, the entity itself, scientific branches that aim to develop techniques or knowledge fields that replicate or approximate human-level or superior intelligence, or specific technologies or techniques that utilise the knowledge gained from these scientific branches.

However, the term "AI" can be misleading in everyday language as it may not indicate the level or extent of human involvement in a technology, leading to confusion about what constitutes AI.

(Israel. D. 1991. "A short sketch of the life and career of John McCarthy", Artificial and Mathematical Theory of Computation, p2)


Further Resources: 


AI glossary

Select a term below to view definition and examples (if available).

Arguably what many people think of when hearing "artificial intelligence", AGI is a hypothetical intelligent agent of equal or superior intellectual ability to humans or animals. Also known as "strong AI", although some reserve that term for an agent that possesses 'a mind' or 'consciousness'.

Also known as: strong AI

Examples: HAL 9000 (2001), Data (Star Trek), Ships in Iain Banks' Culture series

A step-by-step procedure for solving certain kinds of problems or carrying out certain tasks. An algorithm may be expressed in natural language, a programming language or in machine code.

Examples: Euclidean algorithm

The science and engineering of extracting, analysing and understanding information from images or videos. The information and knowledge gained can be applied to enabling robot navigation, object detection, image restoration or medical matters.

Also known as: machine vision, robot vision

Examples: automatic inspection in manufacturing, missile guidance, facial recognition technology

Dark patterns are elements of digital user interface that are maliciously crafted to deliberately obscure, mislead, coerce and/or deceive users into performing actions or making choices they did not mean to do. 

Also known as: deceptive design pattern

Deepfakes use digital manipulation to replace one person's likeness with another in a convincing manner. They are often used to create fake images, audio, and video content. These deepfakes are created by either transforming existing source content by swapping one person for another, or by entirely creating new content where the person is depicted doing or saying something they did not actually do or say.

Also known as: synthetic media, audio deepfakes, wild deepfakes

A multiple layer use of machine learning algorithms to extract greater detail from an input or data. The method may or may not involve human involvement through the means of data that is generated or annotated by humans. Deep learning is used in areas such as computer vision and automatic speech recognition.

Also known as: deep reinforcement learning

Computer system which uses an inference engine and a knowledge base provided through a user interface to allow users to make queries of it for the purpose of receiving appropriate advice as if they were consulting with a human expert or experts. Modern expert systems are able to incorporate new knowledge more easily and thus be more quickly updated than previous approaches.

Also known as: knowledge-based system

Example: MATHLAB, Real time process controls

An AI system that humans are capable of comprehending and trusting, or the methods involved in creating such an AI system. Explainable AI allows an understanding of how an AI system has produced a specific output.

Also known as: xAI, Interpretable AI, Explainable Machine Learning

Example: clinical decision support systems,

A technology which builds on existing material to generate text, images or other media in response to inputted prompts. The generative AI is built through machine learning applied to a data set.

Also Known As - Generative artificial intelligece, GenAI

Example - ChatGPT, Bard, DALL-E, Midjourney

The basis for chatbots like ChatGPT, Bard or LLaMA. The models are trained on a large amount of text from sources such as Wikipedia, web forums, newspaper or journal articles and books. The models use human feedback to improve algorithms which are used to deduce or infer connections and patterns between these text elements

Classification of data based upon models which have been developed and predictions made for future outcomes based upon those models. machine learning uses algorithms to 'learn', or improve the accuracy of predictions or pattern recognition. Machine learning is applied in areas such as computer vision and to technologies such as large language models

Also known as: predictive analytics, statistical learning

Example: deep learning, reinforcement learning, supervised learning, unsupervised learning

An interdisciplinary field of computer science and linguistics concerning text and speech, seeking to use machine learning approaches to accurately gather information from documents or speech as well as gain the ability to categorize and/or organize such documents.

Also known as: NLP

Example: Optical character recognition, text classification, speech recognition, sentiment analysis

A network of processing elements (artificial neurons) where a collection of nodes are linked to each other by one- or two-way nodes, with some connection to the outside world which provides input or output for the network. Using algorithms, the network is able to calculate weights from examples and then use that to predict occurrences in other circumstances.

Also known as: artificial neural network, simulated neural network

An approach to structuring input to a generative AI user interface that seeks to reliably produce a particular desired result. This can include malicious use to create security exploits or to pollute a set of data.

Robots frequently incorporate elements of various computer science or AI disciplines in order to interact with and navigate the physical world.

Example: autonomous vehicles, industrial robots, humanoid robots

A neural architecture or model that is used in fields such as natural language processing and computer vision. The transformer applies mathematical techniques to detect influences and dependencies between data elements. Involved in training large language models on large language datasets, real world applications that result from the use of transformers include machine translation, document generation and computer code generation. Transformers can be used for modalities other than text (audio, images) by breaking down the material into parts and treating them as they would do words in a text-oriented example.

Also known as: Transformer architecture, transformer model

Example: Generative Pre-Trained Transformer (GPT), Bidrectional Encoder Representations from Transformers (BERT)

The data used in the first instance to develop a machine learning model, from which the model creates and refines its rules. A model will then be assessed and improved (fitted) by carrying out predictions on at least one validation data set and one test data set, possibly more. The aim is to create a model that performs well when applied to new, unknown data.

Also known as: training dataset, learning set, training set

A test conceived by Alan Turing in 1950 aimed at revealing whether a machine demonstrated 'artificial intelligence', which involved asking questions via a text input of an entity behind a closed door. If a person was unable to determine the identity of the entity through questioning, it was said to have passed the test and exhibited 'artificial intelligence'.