AI Glossary

The conceptThe raw collection of text, code, images, or audio used to teach an AI model.

Real-world analogyThe library of textbooks given to a student before they begin studying.

Why it mattersHigh-quality data is the foundational fuel for all artificial intelligence.

The conceptSpecialized computer hardware capable of running thousands of math calculations simultaneously.

Real-world analogyInstead of one genius math teacher solving equations one by one, a GPU is 3,000 standard calculators operating at once.

Why it mattersEssential hardware required to train and run modern AI rapidly.

The conceptA step-by-step mathematical rule or formula that tells the computer how to process information.

Real-world analogyA cooking recipe detailing exact ingredients, actions, and steps to get a predictable dish.

The conceptA branch of AI where algorithms learn patterns directly from data instead of relying on manually written code rules.

Real-world analogyTeaching someone to recognize dogs by showing them 5,000 dog photos rather than describing a dog with rigid rules.

The conceptAn AI software architecture loosely inspired by the human brain, using interconnected layers of digital nodes.

Real-world analogyA series of filtration strainers—each layer sifts out finer details until a clear answer emerges.

The conceptAn advanced subset of Machine Learning using multi-layered neural networks to master complex tasks like speech or image recognition.

Real-world analogyLearning a foreign language by living in the country for years until patterns become second nature.

The conceptThe internal mathematical knobs and weights adjusted during training that store the model’s knowledge.

Real-world analogyThe millions of tiny volume knobs on a massive music studio audio console.

The conceptThe intensive phase where AI processes massive datasets, adjusting its parameters to minimize mistakes.

Real-world analogyA medical student spending four years reading textbooks and taking practice exams under supervision.

The conceptThe final, trained software program resulting from an algorithm processing data into learned parameters.

Real-world analogyThe finished doctor who has completed medical school and is now qualified to see patients.

The conceptThe most advanced, highly capable, state-of-the-art AI models currently available at technology’s bleeding edge.

Real-world analogyFormula 1 race cars—the absolute pinnacle of speed and performance technology.

The conceptOpen models publish their internal weights publicly for anyone to download; closed models keep code secret behind private cloud APIs.

Real-world analogyOpen-source is a public sourdough recipe anyone can bake; closed-source is Coca-Cola’s locked vault formula.

The conceptAn AI model capable of understanding and generating multiple data formats simultaneously (text, images, video, audio).

Real-world analogyA person who can read written notes, listen to a podcast, view a diagram, and understand all three seamlessly together.

The conceptThe live execution stage where a trained AI model processes a user query and generates real-time answers.

Real-world analogySitting down to take the final exam—putting stored knowledge to work in real time.

The conceptThe basic chunk of text processed by AI models (roughly 100 tokens ≈ 75 words in English).

Real-world analogyLego bricks used to build words, sentences, and paragraphs.

The conceptThe text, image, or instruction provided by a user commanding the AI to perform a specific task.

Real-world analogyThe specific drink order given to a barista specifying size, milk type, and syrup shots.

The conceptBackground guidelines set by developers before user chat begins, defining the AI’s role, tone, and behavioral rules.

Real-world analogyThe employee orientation handbook instructing a customer representative how to sound professional.

The conceptThe practice of designing, structuring, and refining inputs to guide AI models into producing optimal outputs.

Real-world analogyLearning how to phrase precise questions in a courtroom to get clear answers from a witness.

The conceptThe maximum total memory capacity (in tokens) an AI can read and retain in a single active conversation.

Real-world analogyThe physical surface size of your desk. When it fills up, older papers fall off the edge.

The conceptA dial setting that controls output creativity and randomness (low = factual/predictable; high = imaginative/varied).

Real-world analogyA cooking spice dial: 0.0 is mild broth (strict facts), while 1.0 is extra spicy (creative stories).

The conceptWhen an AI generates false or fabricated information while presenting it with complete confidence.

Real-world analogyA tour guide who forgets historical details and seamlessly invents fictional stories on the fly.

The conceptReinforcement Learning from Human Feedback—refining AI outputs so they match human ethical values, safety, and helpfulness.

Real-world analogyParenting a child—praising polite behavior and correcting actions until social norms are adopted.

The conceptRetraining an existing general AI model on a specialized dataset for a niche industry (e.g., medical law or finance).

Real-world analogyA general practitioner doctor completing a 3-year fellowship in cardiology.

The conceptSafety filters and automated boundary rules that block dangerous, inappropriate, or harmful content.

Real-world analogyHighway guardrails physically preventing vehicles from sliding off steep mountain curves.

The conceptTranslating words, sentences, or images into numerical coordinates so computers can mathematically measure meaning.

Real-world analogyGPS map coordinates for ideas—concepts like “King” and “Queen” share nearby digital locations.

The conceptA database built to store and search numerical embeddings based on conceptual similarity rather than exact keyword matches.

Real-world analogyOrganizing a supermarket by flavor pairing and recipe themes rather than alphabetical ingredient names.

The conceptA software bridge allowing external applications (websites, mobile apps) to send inputs and receive outputs from an AI model.

Real-world analogyA restaurant waiter taking your order to the kitchen chef and returning with your meal.

The conceptA technique where AI fetches accurate information from live external documents or databases before writing an answer, reducing hallucinations.

Real-world analogyAn open-book exam where the student checks official reference books before answering.

The conceptAn open universal connection standard created so AI models can safely plug into local tools, file systems, and enterprise databases.

Real-world analogyUSB-C ports for AI—a single universal plug connecting any model to external accessories.

The conceptAn autonomous AI system equipped with tools, goals, and memory that breaks complex objectives into steps and executes actions.

Real-world analogyAn executive assistant who books flights, reserves hotels, and manages your schedule without asking step-by-step guidance.

The conceptA self-operating, end-to-end system that perceives its environment, makes decisions, and performs continuous real-world actions without human intervention.

Real-world analogyA fully automated robotic warehouse where machines sort, package, and ship goods with zero human operators needed.