The First Inch
On-Demand · $149
Team Training
Keynote
What It Is
Industries
Resources
Learn
AI Glossary
FAQ
Company
About
Affiliates
Contact
Get The First Inch · $149
Log in
The AI world is full of jargon. Here's 24 of the terms you'll actually hear — each one explained in plain English, with a real example. No degree required.
Software that can do things we used to think only humans could do — like writing, answering questions, recognising images, or having a conversation.
When your phone suggests the next word as you type, that's a simple form of AI.
The message or question you type to an AI tool. It's your instruction — the better your prompt, the better the response.
"Write a thank-you email to a client who just renewed" is a prompt.
The type of AI behind tools like ChatGPT and Claude. It was trained on enormous amounts of text and predicts what words should come next to form a useful response.
Think of it as an incredibly well-read assistant that has read most of the internet.
The small chunks of text an AI reads and writes — roughly a word or part of a word. AI tools have limits on how many tokens they can handle at once.
The word "unbelievable" might be split into three tokens: un-believ-able.
When an AI confidently states something that's wrong or made up. It's the single most important thing to watch for — always verify important facts.
An AI might invent a realistic-sounding book title or statistic that doesn't actually exist.
How much information an AI can "hold in mind" at once during a conversation — including your prompt and everything said so far. Bigger windows handle longer documents.
Like short-term memory: once the conversation gets very long, early details may drop off.
All the text and information an AI learned from before you ever used it. This is why AI knows so much — and why it can carry the biases of what it read.
An AI's knowledge has a cut-off date — it doesn't automatically know today's news.
Taking a general AI model and giving it extra training on specific material so it gets better at a particular job or speaks in a particular style.
A company might fine-tune a model on its support tickets to handle customer questions in its own voice.
AI that creates new things — text, images, audio, video, code — rather than just analysing or sorting existing information.
Asking AI to write a poem or generate an image are both generative tasks.
An AI you interact with by chatting back and forth, like texting. Most modern AI assistants work this way.
ChatGPT, Claude, and Gemini are all chatbots.
A specific version of an AI, often with a name and number. Newer models are usually more capable. You'll sometimes get to choose which one to use.
GPT-4, Claude Opus, and Gemini Pro are all models.
The internal "settings" an AI learned during training — often billions of them. More parameters can mean a more capable (but more expensive) model. You rarely need to think about these.
Like the connections in a brain: more connections, more capacity to learn patterns.
An AI that can handle more than just text — it can also "see" images, "hear" audio, or work with multiple types of input together.
Uploading a photo and asking "what's in this picture?" uses a multimodal model.
The moment an AI actually generates a response to your prompt. When people say a model is "thinking," this is what's happening.
Every time you hit enter and wait for an answer, that's inference.
A behind-the-scenes instruction that shapes how an AI behaves across a whole conversation — its personality, rules, and role — before you type anything.
A customer-service bot might have a system prompt telling it to always be polite and never discuss competitors.
A setting that controls how creative or predictable an AI's responses are. Low temperature is focused and consistent; high temperature is varied and surprising.
For a factual answer you want low temperature; for brainstorming wild ideas, higher.
A technique where an AI looks up relevant information from a specific source before answering — so it can give accurate, up-to-date answers grounded in real documents.
An AI that searches your company handbook before answering an HR question is using RAG.
An AI that can take actions on your behalf — not just answer, but actually do multi-step tasks like booking, searching, or operating other software.
An AI agent might research flights, compare them, and fill in the booking form for you.
The skill of writing prompts that get great results. It's the most valuable AI skill you can build — and the whole point of FLUENT.
Knowing to add context, examples, and a clear format is prompt engineering.
A way for software to talk to an AI directly, so developers can build AI into their own apps and websites. You don't need one to use AI tools day-to-day.
When an app says "powered by GPT-4," it's using OpenAI's API behind the scenes.
AI models that anyone can download, inspect, and run themselves — as opposed to closed models you can only access through a company's product.
Meta's Llama models are open source; you can run them on your own computer.
When an AI's responses unfairly favour or disadvantage certain groups — usually because of patterns in its training data. Worth being aware of, especially for important decisions.
An AI trained mostly on one kind of writing might assume that style is "correct" for everyone.
The safety rules built into an AI to stop it producing harmful, dangerous, or inappropriate content.
Guardrails are why an AI will refuse to help with something genuinely dangerous.
Zero-shot means asking AI to do something with no examples. Few-shot means giving it a couple of examples first to guide it — which usually improves results.
Showing AI two sample product descriptions before asking for a third is few-shot prompting.
Try a different word, or browse the full list.
You don't need to know the jargon to use AI brilliantly. But now you do anyway.