AI literacy doesn't mean learning more words. It means understanding which ones matter, which ones mislead, and which ones you can safely ignore. Glossairy clarifies AI language so teams can think clearly, communicate honestly, and make better decisions.
Software trained on data to recognize patterns and generate outputs. The core building block of modern AI.
The input you give a model — a question, instruction, or example — to get a useful output back.
The process of teaching a model by exposing it to large amounts of data. Happens before you ever use it.
Additional training on a narrower dataset to make a model better at a specific task or domain.
A chunk of text — roughly a word or part of a word — that models read and generate one at a time.
The amount of text a model can consider at once. Bigger windows mean more information in, but not always better answers out.
When a trained model generates an output. Every time you use ChatGPT, that's inference.
The technical architecture behind most AI models. Accurate but rarely useful in general conversation — it obscures more than it clarifies.
The internal settings a model learns during training. Often cited as a size metric (e.g., "70 billion parameters") but meaningless to most audiences.
Software that can take actions on your behalf — not just answer questions, but do things: schedule, send, look up, file.
Using software to handle repeatable tasks without manual effort. Not new, but AI makes it possible for less structured work.
A sequence of steps that gets something done. AI plugs into workflows — it doesn't replace them.
Connecting one tool to another so data flows between them. The plumbing that makes AI useful in practice.
Retrieval-augmented generation. A way to give a model access to your own documents so it answers from your data, not just its training.
A way to represent text as numbers so software can compare meaning — used for search, recommendations, and clustering.
A branded metaphor that's become generic. Vague enough to mean anything from autocomplete to a full AI assistant.
When a model confidently generates something that isn't true. Important to understand because it's the default failure mode, not a rare bug.
Patterns in training data that lead to skewed, unfair, or unrepresentative outputs. A real and measurable problem.
The effort to make AI behave in ways humans intend. Important concept, but often used loosely to mean very different things.
Models whose code or weights are publicly available. "Open source" is often used loosely — open weights is more precise.
Artificial general intelligence. A hypothetical future AI that can do anything a human can. Speculative, undefined, and used more for fundraising than clarity.
Having feelings or awareness. No current AI system is sentient. Using this word about AI actively undermines credibility.
A hypothetical point where AI surpasses human intelligence and accelerates beyond control. Science fiction, not strategy.
Models can produce outputs that look like reasoning. Whether they actually reason is an open debate. Using this word uncritically overstates what's happening.
Use Glossairy as a shared reference for your team — for clearer thinking, better narratives, and more effective go-to-market conversations.
Share this glossaryFree. No account required. Link directly to any term.