The Brilliant Noise AI Glossary

Updated monthly

agent / agentic
An AI system that can act semi-independently to complete tasks. It can make decisions, use tools, and even delegate work. “Agentic” describes its ability to behave proactively rather than passively.
Example: A sales agent that books meetings, adapts to responses, and sends follow-ups without human input.

ai literacy
The ability to understand what AI is, how it works, and how to use it confidently and critically. Essential for navigating AI transformation in organisations.
Example: AI literacy helped our leadership team redesign workflows using generative tools.

algorithm
A set of rules a computer follows to solve problems or make decisions. Algorithms are the foundation of AI systems.
Example: Spotify’s algorithm recommends new music based on your listening history.

artificial general intelligence (AGI)
A hypothetical form of AI that can perform any intellectual task a human can. Unlike today’s AI, AGI would be truly adaptable and self-learning.
Example: In science fiction, an AGI might write a novel, diagnose an illness, and hold a deep conversation — all in the same day.

artificial intelligence (AI)
Machines or software that simulate human-like thinking – such as learning, reasoning or problem-solving. The foundation of all smart automation.
Example: A customer service assistant that answers questions in natural language is using AI to mimic human support.

augmented intelligence
AI designed to enhance human capabilities, not replace them. Sometimes called intelligence augmentation (IA).
Example: A diagnostic AI suggests possible issues in scans – the doctor makes the final call.

bias
When an AI system produces unfair or skewed results, often due to biased training data or flawed assumptions.

bot
An automated software agent. Bots can be helpful (e.g. chatbots), harmful (e.g. spam bots), or even mimic humans online.
Example: A support bot that answers FAQs 24/7 is an example of automation in action.

chatbot
An AI-powered system that simulates human conversation. Used for support, sales, or information delivery.
Example: A website chatbot can reset your password or guide you to the right help article.

computer vision
The ability of AI to interpret and understand images and video. Powers things like facial recognition and automated visual checks.
Example: A quality control system that spots defects in product photos.

deep learning
A form of machine learning using layered neural networks to analyse data. It powers image recognition, speech-to-text and more.
Example: Your voice assistant uses deep learning to understand and respond to spoken questions.

diffusion model
A type of generative model that creates images by refining noise into a clear picture, often used in AI art tools.
Example: We used a diffusion model to generate ad visuals in our brand style.

explainable ai (XAI)
AI systems designed to be transparent – they help you understand how and why decisions are made.
Example: A loan tool that shows why it rejected an application based on income and credit score.

fine-tuning
Customising an existing AI model with your data to make it more accurate and brand-specific.
Example: We fine-tuned our chatbot with internal documents to answer employee questions.

generative adversarial network (GAN)
A model made of two AIs – one creates content, the other critiques it – leading to realistic images or data.
Example: GANs are used to generate lifelike faces of people who don’t exist.

generative ai
AI that creates new content – from copy to images – based on patterns it has learned.
Example: We used generative AI to produce 50 product descriptions in seconds.

generative pre-trained transformer (GPT)
A family of AI language models that generate human-like text based on prompts. GPT models power tools like ChatGPT.
Example: GPT-4 helped us write a press release draft from a short product summary.

hallucination
When an AI makes up an answer that sounds right but isn’t. Common in language models.
Example: An AI confidently cited a study that doesn’t exist – classic hallucination.

large language model (LLM)
An AI trained on huge volumes of text to understand and generate language. The backbone of tools like ChatGPT.
Example: We used an LLM to summarise 500 customer reviews into one insight.

machine learning
A type of AI that learns patterns from data and improves over time without being explicitly programmed.
Example: Our recommendation engine uses machine learning to personalise offers based on customer behaviour.

model
The trained engine behind an AI tool. It takes input (e.g. text) and generates output (e.g. a response or prediction).
Example: Our chatbot’s model was upgraded for a better tone of voice.

multimodal
AI that understands and combines different inputs like text, images, audio and video.
Example: A multimodal AI turns a product video into a transcript, summary and blog post.

natural language processing (NLP)
The field of AI focused on understanding and generating human language. Used in chatbots, voice assistants, and more.
Example: NLP helps email tools sort messages or suggest replies.

neural network
A type of AI inspired by the human brain, made up of layers of nodes that learn patterns in data. Core to deep learning.
Example: Neural networks help your phone recognise faces in photos.

orchestration
The process of coordinating multiple AI agents or tools to work together toward a shared goal. Orchestration defines how tasks are divided, in what order they happen, and how outputs are passed between agents — much like a conductor managing different instruments in a symphony.

Example: We used orchestration to connect a research agent, writing agent and fact-checking agent — so our content pipeline ran hands-free from brief to first draft.

prompt
The input you give an AI model to generate a response. The clearer the prompt, the better the output.
Example: "Write three headline options for a sustainability report" is a prompt that gets targeted results.

prompt engineering
The skill of crafting prompts to get better results from generative AI. A key skill for making AI work for you.
Example: We reworded a vague request into a detailed prompt — and got far better copy suggestions.

reinforcement learning
A training method where AI learns by trial and error, getting rewards for good decisions.
Example: Our recommendation engine got better over time by learning which offers users clicked most.

retraining
Updating a model with new data to keep it relevant and accurate.
Example: We retrained our AI to handle new terminology after launching in a different region.

responsible ai
Developing and using AI in ways that are fair, safe and ethical. Includes transparency, bias mitigation and accountability.
Example: Our loan algorithm was audited to ensure it didn’t unfairly penalise certain demographics.

shared language
A common understanding of AI terms and principles within a team or organisation. Helps alignment and clarity.
Example: We built a shared language to make AI conversations between tech and comms teams smoother.

supervised learning
A type of machine learning trained on labelled data with known answers. Great for prediction and classification.
Example: Our email filter was trained on thousands of examples labelled as spam or not.

synthetic data
Artificially generated data used to train AI models — especially useful when real data is limited or sensitive.
Example: We created synthetic customer profiles to train our AI without using real personal data.

token
A piece of text (often a word or part of a word) that language models use to process and generate content.
Example: GPT-4 can handle up to 100,000 tokens in one conversation — that’s around 300 pages of text.

turing test
A test of whether an AI can mimic human conversation convincingly. If a human can’t tell the difference, the AI passes.
Example: Some chatbots can pass parts of the Turing Test — but most still trip up with nuance and humour.

Further reading

Want to go deeper?

These are some of our favourite resources for understanding AI’s foundations, capabilities and implications. We've picked them for clarity, authority and real business relevance.

IBM – What is artificial intelligence?
A clear, business-ready introduction to AI from a trusted enterprise leader. Covers what AI is, the different types, real-world use cases, and how it can drive business value.

NVIDIA – A beginner’s guide to large language models (LLMs): A structured, accessible overview from a respected tech innovator. Explores how LLMs evolved, what they can do today, and why they matter in business applications.

Stanford HAI (Stanford University Human Centred Artificial Intelligence) – Emergent abilities in large language models: An expert-led, research-driven look at how powerful new capabilities arise in LLMs as they scale – essential context for understanding modern AI tools like ChatGPT.

DeepMind – Reinforcement learning: an introduction
A visual, well-organised explainer from a leading AI lab. Ideal for grasping how agents learn through feedback – foundational knowledge for grasping autonomous AI behaviour.

OpenAI – Guide to fine-tuning and custom GPTs
A practical, business-focused roadmap that explains how organisations can adapt AI models using fine-tuning or customization – highly relevant for anyone looking to tailor AI to real use cases.