AI
Artificial intelligence: systems that produce predictions, content or decisions from data and objectives.
AI vocabulary
One word at a time. Search for a term or an idea. No technical knowledge required.
04 / WORDS MADE CLEAR
20 terms
No matches. Try a shorter word or a related concept.
Modello / Model
The mathematical system that processes an input and produces a result. It is not the whole app.
Sources: Google Developers · ML Glossary ↗
LLM
A large language model: it works with language and can generate text. Its answers are not guaranteed to be true.
Sources: Google Developers · ML Glossary ↗ · NIST · Generative AI Profile, 2024 ↗
Prompt
The request or material you give AI to guide a response.
Sources: Google Developers · ML Glossary ↗
Token
A small piece of text processed by a model: a word, part of a word or a symbol.
Sources: Google Developers · ML Glossary ↗
Training
Training: the process that changes a model using examples.
Inference
Using an already trained model to obtain a result.
Sources: Google Developers · ML Glossary ↗
Machine learning
Techniques that let systems learn patterns from data instead of receiving every rule explicitly.
Generative AI
Generative AI produces content, such as text, images or audio, using what it has learned.
Sources: NIST · Generative AI Profile, 2024 ↗
Hallucination
False or unsupported content presented by a model as plausible.
Sources: NIST · Generative AI Profile, 2024 ↗
Context window
The amount of information a model can process together, often expressed in tokens.
Sources: Google Developers · ML Glossary ↗
Multimodal
A model that works with multiple types of data, such as text, images and audio.
Sources: Google Developers · ML Glossary ↗
Agent
An AI system that can choose steps and use tools to pursue a goal within its permissions.
GPU
A processor that performs many calculations in parallel, often used for AI.
Sources: NVIDIA · CUDA introduction ↗
Data center
A facility housing servers, networks, storage and the infrastructure that keeps them running.
Sources: IEA · Energy and AI ↗
Deepfake
Synthetic or manipulated images, video or audio that make a person or scene appear authentic.
Sources: NIST · Generative AI Profile, 2024 ↗
Bias
A systematic distortion that can make results skewed or unfair.
Sources: UNESCO · Ethics of AI ↗
RAG
The system retrieves information from documents or databases to help form an answer. It does not eliminate errors.
Fine tuning
Additional training that adapts an existing model to a task or domain.
Sources: Google Developers · ML Glossary ↗
Open source model
A model available to use, study, modify and share. Downloadable weights alone do not meet the OSI definition.