The AI Glossary

Agentic AI is software that can act on its own to get a job done. Instead of just answering one question, it plans the steps, takes actions, and uses tools until the goal is reached.

AI governance is the set of rules and processes an organization puts in place to make sure its AI systems are used safely, fairly, and in line with laws and values. It answers who decides what AI can do and who is accountable when it goes wrong.

AI inference is what happens when you ask an AI a question and it answers. The model uses what it learned during training to respond to your new input in real time.

AI safety is the work that goes into making sure AI systems do what they are supposed to do, without causing unintended harm. It covers everything from preventing biased outputs to ensuring powerful AI systems remain under human control.

Artificial intelligence is the science of making computers do tasks that normally need human thinking, like understanding language or spotting patterns. It is the broad umbrella that covers many smaller fields.

Computer vision is the part of AI that helps computers 'see' and make sense of pictures and video. It lets software identify objects, people, or text in an image.

A context window is like an AI model's working memory. It determines how much text the model can read at once — the larger the window, the more it can consider when generating a reply.

A diffusion model is the technique used by AI image generators like Midjourney and DALL-E. It starts with random noise and progressively refines it into a detailed image guided by your text description.

Embeddings turn words, sentences, or images into lists of numbers that capture their meaning. Things with similar meaning get similar numbers, so a computer can tell what is related.

Fine-tuning is taking a model that already knows a lot and giving it extra training on your own examples. This teaches it to do a specific job better, such as writing in your brand's voice.

Generative AI is technology that makes brand-new content, like writing, pictures, or code, instead of just sorting or labeling existing data. You describe what you want, and it produces something original.

An AI hallucination is when a chatbot confidently says something that is simply not true. It can invent facts, names, or sources that look real but are made up.

A large language model is an AI trained on huge amounts of text so it can read your question and write a useful answer. It powers chatbots and writing assistants.

Machine learning is a way for computers to learn from examples instead of being told exact rules. The more relevant data they see, the better they get at making predictions.

Model distillation trains a small, fast AI model to behave like a large, slow one. The result is a cheaper model that's nearly as capable, ideal for running at scale or on devices with limited compute.

Multimodal AI can understand and create more than one type of content — for example, looking at an image and answering a question about it, or turning a text description into a picture.

Natural language processing is the part of AI that helps computers understand and use human language. It is what lets software read your text, translate it, or figure out how you feel.

A neural network is a computer model inspired by how brain cells connect. It learns by adjusting many tiny connections until it can recognize patterns, like telling cats from dogs in photos.

Prompt engineering is the art of asking an AI the right way. By writing clear instructions and giving context, you get much better answers from the same tool.

Retrieval-augmented generation lets an AI look up relevant facts before answering. Instead of relying only on memory, it pulls in the right documents and uses them to give a more accurate reply.

A token is a small piece of text, like a word or part of a word, that an AI reads one at a time. Models count tokens to measure how much text they can handle.

A transformer is the underlying design used by most modern AI language models. It lets the model pay attention to every word in a sentence at once, so it understands meaning and context much better than older approaches.

A vector database stores information as numbers that capture meaning, then finds other items that are similar in meaning — not just exact word matches. It powers the 'find related content' intelligence in many AI applications.

Zero-shot learning is when an AI completes a task it has never been trained on specifically — following a new instruction based on its general knowledge, without seeing any examples first.