ICDL AI Essentials key terms

Computer systems designed to perform tasks that normally require human intelligence, such as recognising patterns, understanding language, making decisions, or generating content.

A branch of AI where computers learn patterns from examples instead of following only hand-written rules. The resulting model improves as it is exposed to more relevant data.

A defined set of steps or rules a computer follows to solve a problem or complete a task. In AI, algorithms process data and produce predictions, classifications, or generated outputs.

The examples, records, or content used to teach a machine-learning model. Its quality, relevance, and balance strongly shape what the model learns and how fairly it behaves.

The output of a training process: a mathematical structure that has learned patterns from data and can make predictions, classifications, or generate new content when given an input.

A computing structure loosely inspired by connected brain cells. Layers of simple processing units adjust their connections during training so the network can recognise patterns in complex data.

Machine learning that uses neural networks with many layers. It is especially effective for images, speech, and language, but usually needs large amounts of data and computing power.

AI systems that create new content, such as text, images, audio, or code, in response to instructions. They learn patterns from existing material rather than copying a fixed answer from a database.

A generative AI model trained on very large amounts of text to predict and produce language. It can answer questions, summarise, translate, and draft text, but may state false information confidently.

The field of AI concerned with helping computers understand, interpret, and produce human language. It supports translation, sentiment analysis, chatbots, and text summarisation.

AI techniques that let computers interpret images and video. Common uses include recognising objects, reading text from photos, checking product quality, and analysing medical scans.

The stage where a trained AI model is used to produce an answer, prediction, or generated output from new input. It happens after training and does not normally change the model's learned weights.

A voice- or text-controlled service that helps with everyday tasks such as setting reminders, answering questions, playing media, or controlling devices. It typically combines speech recognition, search, and language processing.

An AI tool that suggests content, products, or services based on patterns in a user's behaviour and the choices of similar users. It aims to predict what someone may find useful or appealing.

A vehicle that uses sensors, cameras, and AI software to perceive its surroundings and drive with limited or no human control. Safety depends on reliable sensing, cautious design, and clear human oversight.

A technology that analyses facial features in images or video to identify or verify a person. It raises privacy and bias concerns because faces are sensitive personal data and accuracy can vary across groups.

A connected household product, such as a speaker, thermostat, or camera, that can be controlled remotely or automated. Many use AI to recognise speech, learn routines, or detect unusual activity.

A conversational AI tool that responds to user messages in text or speech. It may answer questions, guide a process, or provide support, and its reliability depends on its design and knowledge sources.

Tailoring content, recommendations, or interactions to an individual's preferences and behaviour. AI enables this at scale, but it can also create filter bubbles or misuse personal data.

A feature that suggests the next word or phrase while a person types. It uses language patterns and context to speed up writing, though suggestions can be inaccurate or inappropriate.

The instruction or question a person gives to a generative AI tool. Clear prompts state the task, context, desired format, and any constraints so the system can produce a more useful response.

The practice of designing and refining instructions to get better results from a generative AI system. It includes specifying context, examples, format, tone, and constraints, then testing and improving the prompt.

When a generative AI system produces information that sounds plausible but is false, made up, or unsupported. Users should verify important claims against reliable sources before relying on them.

The amount of text or other input a generative AI model can consider at one time. Content outside this limit may be ignored, so long documents often need splitting or summarising.

Further training of an existing AI model on a smaller, specialised dataset so it performs better for a particular task or style. It requires suitable data and careful testing to avoid unwanted side effects.

The process of checking an AI system's response for accuracy, completeness, and suitability before using it. It may involve fact-checking, comparing sources, testing code, or asking a knowledgeable person to review.

Improving a result by sending follow-up prompts that clarify, correct, or extend earlier responses. Instead of expecting one perfect answer, the user reviews each output and guides the AI step by step.

An AI system that can work with more than one type of input or output, such as text, images, audio, or video. This allows tasks like describing a picture or answering questions about a document.

A systematic tendency in AI outputs that unfairly favours or disadvantages particular people or groups. It can come from unbalanced training data, flawed labelling, or choices made by developers.

The protection of personal information from misuse, unauthorised access, or unnecessary sharing. AI tools can collect and process sensitive data, so users should know what is stored and who can see it.

Legal rights that give creators control over how their original works are used, copied, or distributed. AI raises questions about training on protected material and about ownership of AI-generated output.

Synthetic image, audio, or video created with AI to show someone saying or doing something that never happened. Deepfakes can deceive audiences, damage reputations, and undermine trust in evidence.

False or misleading information spread regardless of intent. AI can produce and amplify it quickly, so checking sources and verifying surprising claims is an important safety habit.

Openness about how an AI system works, what data it uses, and what it cannot do. Clear disclosure helps users judge reliability, spot limitations, and hold organisations responsible.

The involvement of people in monitoring, reviewing, or overriding AI decisions. It is essential in high-stakes settings such as healthcare, finance, and transport, where errors can cause serious harm.

The principle that a person or organisation remains responsible for the effects of an AI system they use or deploy. Responsibility cannot simply be shifted to the tool itself.