AI Terms: 180+ Artificial Intelligence Terms Explained

Artificial intelligence♥

Also called: AI

Computer systems that do tasks we normally link with human thinking, such as understanding language, recognising images, making decisions or solving problems.

Everyday

Chatbot♥

A program you talk to in everyday language by typing or speaking. Modern AI chatbots are built on large language models.

EverydayFree chatbots worth trying

Generative AI♥

Also called: GenAI

AI that creates new content, such as text, images, music, voice, video or code, instead of only sorting or analysing existing data.

Everyday

Machine learning♥

Also called: ML

A branch of AI where software learns patterns from examples (data) instead of following rules a programmer wrote by hand. The more good examples it sees, the better it usually gets.

Everyday

AI winter

A period when funding and interest in AI dropped sharply after early hype failed to deliver. The field went through AI winters in the 1970s and late 1980s.

Advanced

Algorithm

A set of step-by-step instructions a computer follows to solve a problem or complete a task.

Everyday

Artificial general intelligence

Also called: AGI

A hypothetical AI that could learn and perform almost any intellectual task a human can, across many fields. It does not exist yet, and experts disagree on when or whether it will.

Intermediate

Computer vision

The part of AI that lets computers understand images and video, for example spotting faces, reading text in photos or finding objects in a scene.

Intermediate

Conversational AI

AI designed to hold a back-and-forth conversation in natural language, by text or voice. Chatbots and voice assistants are examples.

Everyday

Deep learning

A kind of machine learning that uses neural networks with many layers. It powers modern speech recognition, image recognition and chatbots.

Intermediate

Expert system

An older style of AI built from hand-written if-then rules supplied by human experts, popular in the 1980s for tasks like diagnosis and troubleshooting.

Advanced

Model

Also called: AI model

The trained result of machine learning: a file of learned patterns that takes an input (a question, a photo) and produces an output (an answer, a label, an image).

Everyday

Narrow AI

Also called: weak AI

AI built for one specific job, such as recommending films or detecting spam. Almost all AI in use today is narrow AI.

Intermediate

Natural language processing

Also called: NLP

The part of AI that deals with human language: reading, understanding, translating, summarising and writing text.

Intermediate

Neural network

A computer model loosely inspired by the brain: layers of connected ‘neurons’ (simple maths units) that pass numbers to each other and adjust their connections as they learn.

Intermediate

Recommendation system

Also called: recommender

AI that suggests what you might like next, such as videos, songs, products or posts, based on your past behaviour and people like you.

Everyday

Robotics

The field of building machines that sense and move in the physical world. AI gives robots the ability to see, plan and adapt.

Everyday

Speech recognition

Also called: speech-to-text

AI that turns spoken words into written text, used in voice typing, captions and voice assistants.

Everyday

Superintelligence

Also called: ASI

A hypothetical AI far smarter than the best humans in nearly every area. It is a subject of research and debate, not something that exists today.

Advanced

Text-to-speech

Also called: TTS

Technology that reads text aloud in a synthetic voice. AI voices now sound much more natural than the old robotic ones.

Everyday

Turing test

A test proposed by Alan Turing in 1950: if a person chatting by text cannot reliably tell a machine from a human, the machine passes.

Intermediate

Virtual assistant

Also called: voice assistant

An AI helper on your phone, speaker or computer that answers questions and does simple tasks, like setting timers or sending messages, by voice or text.

Everyday

Context window♥

Also called: context length

How much text (counted in tokens) a model can take into account at once, including your messages, files and its own replies. Beyond that, older parts are dropped or forgotten.

Everyday

Large language model♥

Also called: LLM

An AI model trained on huge amounts of text so it can understand and write language. It works by predicting the next piece of text, one token at a time. Most modern chatbots run on LLMs.

Everyday

Multimodal AI♥

AI that can work with more than one type of input or output, such as text, images, audio and video, in the same conversation.

Everyday

Reasoning model♥

Also called: thinking model

A model that works through a problem step by step before it answers, often showing a ‘thinking’ stage. It is better at maths, logic and coding but slower.

Everyday

Token♥

The small chunk of text an AI model reads and writes, often a word or part of a word. AI limits, speed and prices are usually counted in tokens.

Everyday

Attention

Also called: self-attention

The mechanism inside a transformer that decides which parts of the input matter most for each word it is processing.

Advanced

Benchmark

A standard test used to compare AI models, for example sets of maths problems, coding tasks or exam questions.

Intermediate

Chain of thought

Also called: CoT

Step-by-step reasoning written out before the final answer. Asking a model to think step by step often improves results on hard questions.

Intermediate

Closed model

Also called: proprietary model

A model you can only use through the maker’s app or API. Its weights and training details are kept private.

Intermediate

Diffusion model

The kind of model behind many AI image and video generators. It learns to start from random noise and remove it step by step until a clear picture appears.

Intermediate

Distillation

Also called: knowledge distillation

Training a smaller ‘student’ model to copy the behaviour of a larger ‘teacher’ model, giving a faster model that keeps much of the skill.

Advanced

Embedding

A list of numbers that represents the meaning of a word, sentence or image, so that similar meanings sit close together. Used for search and recommendations.

Advanced

Foundation model

A large, general-purpose model trained on broad data that other apps and models are built on top of, often after fine-tuning for a specific job.

Intermediate

Frontier model

A term for the most advanced and capable AI models available at any given time, usually from the biggest AI labs.

Advanced

GPT

Also called: Generative Pre-trained Transformer

Short for Generative Pre-trained Transformer: a language model that is first pre-trained on lots of text, then generates new text. The letters appear in some product names, but the idea is a general one.

Intermediate

Inference

Using a trained model to get an answer. Training is learning; inference is the model doing its job each time you send a prompt.

Intermediate

Knowledge cutoff

Also called: training cutoff

The date after which a model has no training data. It will not know about later events unless it can search the web or read files you give it.

Everyday

Latency

The delay between asking an AI something and getting the response. Lower latency feels faster.

Intermediate

Mixture of experts

Also called: MoE

A model design where only some specialist sub-networks (‘experts’) switch on for each token. It lets very large models run more cheaply.

Advanced

Model card

A document from a model’s maker describing what the model is for, how it was trained and tested, its limits and known risks.

Advanced

Natural language understanding

Also called: NLU

The part of language AI focused on working out what text means, such as intent, sentiment and who did what.

Advanced

Open source AI

AI whose code and, ideally, weights and training details are freely available to use, study, change and share. Definitions vary on how much must be open.

Intermediate

Open-weight model

Also called: open model

A model whose trained weights are published so anyone can download and run it on their own hardware, usually under a licence with conditions.

Intermediate

Parameters

Also called: weights

The internal numbers a model learns during training. Model size is often given in parameters, such as 7 billion; more parameters usually means more capability but more computing power.

Intermediate

Quantization

Shrinking a model by storing its numbers with less precision, so it uses less memory and runs faster, with a small loss in quality.

Advanced

Small language model

Also called: SLM

A smaller, cheaper language model that can run on a phone or laptop. It is less capable than the biggest models but faster and more private.

Intermediate

Temperature

A setting that controls how random an AI’s output is. Low temperature gives predictable, focused answers; high temperature gives more varied and creative ones.

Advanced

Tokenization

Splitting text into tokens before a model processes it. For example, ‘unbelievable’ may become ‘un’, ‘believ’ and ‘able’.

Intermediate

Transformer

The neural network design behind modern language models. Introduced in 2017, it uses ‘attention’ to weigh how every word in a passage relates to every other word.

Intermediate

Vector

A list of numbers. In AI, text and images are turned into vectors so a computer can measure how similar they are.

Advanced

Vector database

A database built to store embeddings and quickly find the ones most similar to a query. It is often used to give chatbots access to your documents.

Advanced

Prompt♥

The instruction or question you give an AI. Clear prompts with context and an example usually get better answers.

EverydayReady-to-use prompts

AI search

Also called: answer engine

Search that gives a written answer drawn from several web pages, usually with source links, instead of only a list of results.

EverydayAI search engines compared

Canvas

Also called: side-panel editor

A side panel in some chatbots where a document, code or app is shown and edited separately from the chat.

Intermediate

Citation

Also called: source link

A link or reference an AI shows for where a fact came from, so you can check it yourself.

Everyday

Custom instructions

Saved preferences you give a chatbot once (who you are, how you like answers) so it applies them in every chat.

Everyday

Deep research

An AI mode that runs many searches, reads dozens of sources and writes a long report with citations. It takes minutes rather than seconds.

EverydayAI research tools with real sources

Few-shot prompting

Giving an AI a few worked examples in your prompt so it copies the pattern for your real request.

Intermediate

Grounding

Tying an AI’s answer to trusted sources, such as search results or your documents, so it is less likely to make things up.

Intermediate

Memory

Also called: AI memory

A chatbot feature that remembers facts and preferences from past chats so you do not have to repeat them. You can usually view or delete what it stores.

Everyday

Output

Also called: completion, response

What the AI produces in reply to your prompt: text, code, an image, audio and so on.

Everyday

Prompt engineering

The skill of writing and refining prompts to get better, more reliable results from an AI model.

Everyday

Prompt library

A saved collection of prompts that work well, kept for reuse by a person or a team.

Intermediate

Prompt template

A reusable prompt with blanks to fill in, such as a topic, audience or tone, so you get consistent results.

Intermediate

Regenerate

Asking the AI to produce a fresh answer to the same prompt. Because AI output varies, a second try can be better.

Everyday

Role prompting

Also called: persona prompt

Asking the AI to act as a particular expert or character, such as ‘act as a patient maths tutor’, to shape its answers.

Intermediate

Summarization

Using AI to shorten a long text, video or meeting into the main points.

Everyday

System prompt

Also called: system message

Hidden or background instructions that set an AI’s role, rules and tone before the conversation starts.

Intermediate

Tone

The style of an AI’s writing, such as formal, friendly or funny. You can ask for a tone in your prompt.

Everyday

Zero-shot prompting

Asking an AI to do a task with no examples, relying only on what it already knows.

Intermediate

Deepfake♥

Fake video, audio or images made with AI to show a real person saying or doing something they never did.

Everyday

AI avatar

A digital character or portrait created with AI, sometimes animated to speak in a video.

Everyday

AI music generation

AI that composes songs or background music from a text description, sometimes with vocals.

Everyday

AI voice cloning

Also called: voice cloning

Creating a synthetic copy of a real person’s voice from a recording. Useful for accessibility and dubbing, but also abused in scam calls.

Everyday

Aspect ratio

The width-to-height shape of an image or video, such as 16:9 for widescreen or 9:16 for phone stories.

Everyday

Content credentials

Also called: provenance

Tamper-evident information attached to a file that records how it was made and edited, including whether AI was used.

Advanced

Image recognition

AI that identifies what is in a picture, such as objects, plants, landmarks or text.

Everyday

Image-to-image

AI that changes an existing picture based on your instructions, such as changing its style, background or lighting.

Intermediate

Inpainting

Editing part of an image with AI: you mark an area and describe what should replace it.

Intermediate

Lip sync

AI that moves a face’s mouth in a video to match new speech, used in dubbing and talking avatars.

Intermediate

Negative prompt

A list of things you do not want in an AI image, such as ‘blurry, extra fingers, text’.

Intermediate

Optical character recognition

Also called: OCR

Technology that reads printed or handwritten text from images and turns it into editable text.

Everyday

Outpainting

Also called: generative expand

Extending an image beyond its original edges with AI, filling in new scenery that matches.

Intermediate

Seed

A number that sets the starting randomness of an AI image. Using the same seed and prompt usually gives a similar picture again.

Advanced

Style transfer

Applying the look of one image, such as a watercolour or sketch style, to another image.

Intermediate

Synthetic media

Also called: AI-generated media

Any image, audio, video or text created or heavily changed by AI rather than captured from real life.

Intermediate

Text-to-image

AI that creates a picture from a written description, called a prompt.

EverydayFree AI art generators

Text-to-video

AI that generates a short video clip from a written description.

Everyday

Upscaling

Also called: AI upscaler

Using AI to enlarge an image or video and add detail so it looks sharper at a bigger size.

Everyday

Watermark

Also called: AI watermark

A hidden or visible mark added to AI-generated content so it can later be identified as made by AI.

Intermediate

Fine-tuning♥

Extra training of an existing model on a smaller, focused dataset so it gets better at a specific task or style.

Intermediate

Training data♥

Also called: dataset

The examples a model learns from, such as text, images or audio. Its quality and variety shape what the model can do and its biases.

Everyday

Backpropagation

The method a neural network uses to work out how much each weight contributed to an error, so each can be adjusted.

Advanced

Classification

A task where AI sorts inputs into categories, such as spam or not spam.

Intermediate

Clustering

Grouping similar items together without labels, such as grouping customers by shopping habits.

Advanced

Data labeling

Also called: data annotation

The work of tagging data with the right answers, often done by people, to train and check AI models.

Intermediate

Data scraping

Also called: web scraping

Collecting large amounts of content from websites automatically. AI companies scrape the web for training data, which raises copyright and privacy questions.

Intermediate

Epoch

One complete pass through the whole training dataset during training.

Advanced

Federated learning

Training a shared model across many devices, such as phones, without sending the raw data to a central server.

Advanced

Gradient descent

The step-by-step method used to adjust a model’s weights to reduce its errors during training.

Advanced

Labeled data

Also called: annotation

Data that people have tagged with the correct answer, used to train and test supervised models.

Intermediate

Loss function

The formula that measures how wrong a model’s predictions are during training. Training tries to make the loss smaller.

Advanced

Model collapse

A drop in quality that can happen when AI models are trained mostly on AI-generated content instead of human-made data.

Advanced

Overfitting

When a model memorises its training examples so closely that it performs badly on new data it has not seen.

Advanced

Pre-training

The first, biggest stage of training a language model, where it learns general language and knowledge from a huge amount of text.

Intermediate

Regression

A task where AI predicts a number, such as a house price or tomorrow’s temperature.

Advanced

Reinforcement learning

Also called: RL

Learning by trial and error: the AI takes actions and gets rewards or penalties, and gradually learns what works.

Intermediate

Reinforcement learning from human feedback

Also called: RLHF

A training step where people rate a model’s answers, and the model is tuned toward the kinds of answers people preferred.

Advanced

Self-supervised learning

Training where the data supplies its own answers, such as hiding a word in a sentence and asking the model to predict it. It is how language models learn from raw text.

Advanced

Supervised learning

Training with labelled examples, where each example comes with the right answer, such as photos tagged ‘cat’ or ‘dog’.

Intermediate

Synthetic data

Artificially generated data, often made by AI, used to train or test models when real data is scarce, costly or private.

Intermediate

Training

The process of feeding a model lots of examples so it learns patterns. Big models train for weeks on thousands of chips.

Everyday

Underfitting

When a model is too simple or undertrained to capture the real patterns, so it performs badly even on training data.

Advanced

Unsupervised learning

Training on data without labels, where the model finds patterns or groups on its own.

Intermediate

Agentic AI♥

AI that acts with some independence toward a goal: it breaks the goal into steps, uses tools and checks its own progress.

EverydayOur plain-English guide to agentic AI

AI agent♥

Also called: agent

An AI system that can plan and carry out multi-step tasks for you, such as browsing websites, filling forms or editing files, rather than only answering questions.

EverydayAI agents you can try

Retrieval-augmented generation♥

Also called: RAG

A method where the AI first looks up relevant information in documents or the web, then writes its answer using what it found. It reduces made-up answers and keeps responses current.

Intermediate

AI coding assistant

An AI tool that suggests, writes, explains or fixes code inside a code editor or terminal.

EverydayThe best AI tools for coding

AI copilot

Also called: AI assistant

A general term for an AI helper built into an app that works alongside you, suggesting text, code or actions as you go.

Everyday

AI note taker

Also called: meeting assistant

An app that joins or records meetings, writes a transcript and summarises decisions and action items.

Everyday

API

Also called: application programming interface

A way for one program to talk to another. Developers use AI APIs to add AI features to their own apps.

Intermediate

Autonomous

Able to act without step-by-step human direction. Fully autonomous AI is still limited, and most products ask you to confirm important actions.

Intermediate

Computer use

Also called: browser agent

An AI ability to operate a computer or web browser like a person: looking at the screen, clicking, typing and scrolling.

Intermediate

Custom chatbot

Also called: custom assistant, custom bot

A version of a chatbot you set up with your own instructions and files for a specific job, such as a recipe helper or a study tutor.

Everyday

Human in the loop

Also called: HITL

A setup where a person reviews or approves an AI’s work at key steps before it acts, such as before sending an email or making a payment.

Intermediate

Knowledge base

A collection of documents, such as manuals, FAQs or notes, that an AI assistant can search to answer questions accurately.

Intermediate

Model Context Protocol

Also called: MCP

An open standard for connecting AI assistants to outside apps and data, such as files, calendars and databases, through a common plug-in style interface.

Advanced

Multi-agent system

Several AI agents working together, each with its own role, such as one researching, one writing and one checking.

Advanced

Orchestration

Coordinating several AI models, tools or agents so they work together on a larger task in the right order.

Advanced

Plugin

Also called: connector, extension

An add-on that connects an AI assistant to another service, such as email, a calendar or a shopping site, so it can use that service’s data.

Everyday

Semantic search

Search that matches meaning rather than exact words, so ‘cheap flights’ can find a page about ‘low-cost airfares’.

Intermediate

Tool use

Also called: function calling

When an AI model calls outside tools, such as a calculator, web search, calendar or database, to get information or take actions.

Intermediate

Vibe coding

Building software mainly by describing what you want to an AI in plain language and accepting the code it writes, with little manual coding.

Everyday

Workflow automation

Using software, now often with AI, to run a series of tasks automatically, such as filing emails and updating a spreadsheet.

Everyday

AI bias♥

Also called: algorithmic bias

When an AI system produces unfair or skewed results, often because its training data was unbalanced or reflected human prejudice.

Everyday

Hallucination♥

Also called: confabulation

When an AI states something false or made up, such as a fake fact, quote or source, as if it were true. Always check important facts.

Everyday

AI alignment

Also called: alignment

The work of making sure AI systems pursue the goals and values their makers and users intend, and do not cause harm.

Intermediate

AI detector

Also called: AI content detector

A tool that estimates whether text or images were made by AI. Results are often unreliable and should not be the only evidence.

Everyday

AI ethics

The study of moral questions raised by AI, such as fairness, privacy, jobs, consent and who is responsible when AI causes harm.

Everyday

AI regulation

Also called: AI law

Laws and rules that govern how AI is built and used, such as transparency, safety testing and banned uses.

Intermediate

AI safety

The field focused on preventing AI from causing accidents or harm, from everyday mistakes to large-scale risks.

Everyday

AI scam

Fraud that uses AI, such as cloned voices of relatives, deepfake videos of celebrities or AI-written phishing emails.

Everyday

Black box

A system whose inner workings are hard to see or understand, even for its makers. Large neural networks are often described this way.

Intermediate

Copyright and AI

The legal questions around AI training on copyrighted works and who owns AI-generated content. Rules differ by country and are still being settled in courts.

Intermediate

Data poisoning

Sneaking bad or misleading examples into training data to make a model behave wrongly.

Advanced

Data privacy

Protecting personal information. With AI, it covers what you type into chatbots, whether chats are used for training, and how long they are kept.

Everyday

Disinformation

False information spread on purpose to mislead. AI-generated images, voices and articles can be used for it.

Everyday

Existential risk

Also called: x-risk

The idea that very advanced AI could one day pose a threat to humanity’s future. It is debated among researchers.

Advanced

Explainable AI

Also called: XAI

AI designed so people can understand why it made a decision, which matters in areas like loans, hiring and healthcare.

Intermediate

Guardrails

Also called: safety filters

Rules and filters that limit what an AI will say or do, for example refusing to help with dangerous or illegal requests.

Everyday

Jailbreak

A trick prompt designed to get an AI to ignore its safety rules. AI makers patch these as they find them.

Intermediate

Misinformation

False information spread without intent to deceive. AI can make it faster to produce and harder to spot.

Everyday

Opt out of training

Also called: training opt-out

A setting in many AI apps that stops your chats or content from being used to train future models.

Everyday

Prompt injection

An attack where hidden instructions in a web page, email or file trick an AI into doing something its user did not ask for, such as leaking data.

Intermediate

Red teaming

Deliberately attacking or testing an AI system to find weaknesses and harmful behaviour before bad actors do.

Advanced

Responsible AI

Building and using AI in ways that are fair, safe, transparent, private and accountable.

Everyday

Sycophancy

When an AI tells you what you want to hear, agreeing with you or praising your work, instead of giving an honest answer.

Advanced

Transparency

Being open about when AI is used, how it works and what data it relies on.

Intermediate

AI slop♥

Also called: slop

Slang for low-quality, mass-produced AI content, such as spammy articles, fake images and filler videos, flooding the web and social media.

Everyday

On-device AI♥

Also called: edge AI, local AI

AI that runs directly on your phone or computer instead of in the cloud. It can work offline and keep data more private.

Everyday

AI literacy

Knowing what AI can and cannot do, how to use it well and safely, and how to judge its output.

Everyday

AI Overviews

Also called: AI summaries

AI-written summaries shown at the top of some search results pages, with links to the sources they used.

Everyday

AI PC

A marketing term for a computer with a built-in NPU designed to run AI features locally.

Everyday

AI washing

Exaggerating how much a product really uses AI, to make it sound more advanced.

Intermediate

AI-powered

Also called: AI-driven

Marketing phrase meaning a product uses AI for at least some of its features.

Everyday

Automation

Using technology to do tasks with little human effort. AI extends automation to tasks that need judgement, like sorting email or answering questions.

Everyday

Autonomous vehicle

Also called: self-driving car

A car that uses AI, cameras and sensors to drive itself with little or no human control.

Everyday

Cloud AI

AI that runs on a company’s remote servers. You send your request over the internet and get the result back.

Everyday

Compute

The computing power, such as chips, servers and electricity, needed to train and run AI models.

Intermediate

Data center

A building full of servers. AI data centers use huge numbers of GPUs and a lot of electricity and cooling.

Everyday

Digital twin

A virtual copy of a real object, place or process, kept up to date with real data so it can be tested and predicted with AI.

Advanced

Generative engine optimization

Also called: GEO

Making web content easy for AI search tools and chatbots to understand, trust and cite, alongside traditional SEO.

Intermediate

GPU

Also called: graphics processing unit

A chip first made for graphics that can do many calculations at once, which makes it ideal for training and running AI.

Intermediate

Humanoid robot

A robot built in a roughly human shape, with arms, legs and a head, so it can work in spaces designed for people.

Intermediate

Hyperscaler

One of the giant cloud companies that run the huge data centers where most large AI models are trained and served.

Advanced

Moravec’s paradox

The observation that tasks hard for humans, like chess, are often easy for computers, while tasks easy for humans, like folding laundry, are hard for robots.

Advanced

NPU

Also called: neural processing unit

A chip in newer phones and laptops built to run AI features on the device quickly and with little battery use.

Intermediate

Scaling laws

The observed pattern that AI models tend to get better in predictable ways as you add more data, parameters and compute.

Advanced

Singularity

Also called: technological singularity

A hypothetical future point when AI improves itself so fast that technology changes beyond human understanding or control.

Advanced

TPU

Also called: tensor processing unit

A chip designed specifically to speed up machine learning maths.

Advanced

Usage limit

Also called: rate limit

A cap on how many messages, images or tokens you can use in a period. Free plans have lower limits than paid ones.

Everyday