Short answer: These are 40 common AI terms in plain language, grouped by theme. They are simplified for general readers. Where a term has a formal or legal meaning, check the authoritative source before relying on it. Useful references are ISO/IEC 22989:2022 on AI concepts and terminology, the NIST AI Resource Center glossary, and the Personal Data Protection Act 2012 for "personal data".
What are the core ideas?
- Artificial intelligence (AI): software that performs tasks that normally need human judgement.
- Machine learning: systems that learn patterns from data instead of fixed rules.
- Algorithm: a set of steps a computer follows.
- Model: the learned pattern a system uses to make predictions or outputs.
- Training: the process of teaching a model using data.
- Inference: using a trained model to produce an output.
- Dataset: a collection of data used to train or test.
- Parameter: an internal number a model adjusts while learning.
- Neural network: a model loosely inspired by connected brain cells.
- Deep learning: machine learning using large neural networks.
What are the generative AI terms?
- Generative AI: AI that creates text, images, audio or code.
- Large language model (LLM): a model trained on large amounts of text to produce language.
- Prompt: the instruction you give a model.
- Token: a small chunk of text a model processes.
- Context window: how much text a model can consider at once.
- Hallucination: a confident output that is false or invented.
- Fine-tuning: further training a model for a specific task.
- Retrieval-augmented generation (RAG): giving a model relevant documents to ground its answer.
- Embedding: a numeric representation of meaning used for search.
- Multimodal: handling more than one type of input, such as text and images.
What are the practical terms?
- Agent: an AI system that takes steps or uses tools toward a goal.
- Chatbot: software you converse with.
- API: a way for programs to talk to each other.
- Open-source model: a model whose weights or code are published under an open licence.
- Automation: a task done by software without manual steps.
- Human in the loop: a person reviews or approves outputs.
- Benchmark: a standard test for comparing models.
- Latency: the delay before a response arrives.
- Edge AI: AI running on a local device rather than in the cloud.
- Computer vision: AI that interprets images and video.
What are the governance and risk terms?
- Bias: systematic unfairness in outputs, often from data or design.
- Fairness: treating people equitably in decisions.
- Transparency: being open about how and where AI is used.
- Explainability: being able to give reasons for an output.
- Accountability: clear responsibility for outcomes.
- Personal data: data about an individual who can be identified from that data, or from that data and other information the organisation has or is likely to have access to (Personal Data Protection Act 2012, section 2).
- Data governance: rules for managing data quality, access and use.
- Deepfake: synthetic media that imitates a real person.
- Red teaming: deliberately testing a system to find weaknesses.
- Model card: a document describing a model's purpose, data and limits.
Why do definitions differ between sources?
The field moves quickly and organisations define terms for their own purposes. Treat this list as an orientation, and prefer official definitions in contracts and policies.
Keep learning
Our free AI Fundamentals for Professionals course is listed on the Academy page (checked 21 November 2026). The free monthly newsletter is at asia-ai.org; membership is at asia-ai.org/join.
This article is general information, not legal advice.