AI Glossary — every term in Artificial Intelligence
Direct definitions, practical SME examples and context you won't find on Wikipedia. Search, read, apply.
The process of optimizing digital content so that it becomes the direct answer chosen by AI assistants and conversational search engines such as ChatGPT or Perplexity.
An autonomous system capable of planning tasks, using external tools, and making decisions to achieve a specific goal without constant human supervision.
The ability of an AI model to process only the relevant parts of information, ignoring noise and focusing on critical relationships between the input data.
Technique for interacting with language models that forces them to decompose complex problems into intermediate logical steps before presenting the final answer.
The process of fragmenting extensive documents into smaller, coherent parts so that IA systems can accurately process, retrieve, and cross-reference information.
Maximum amount of information an AI model can process at once before beginning to forget initial details.
A mathematical metric that measures the degree of similarity between two items, such as documents or products, based on the orientation of their numerical meaning rather than their length.
A numerical representation of a content's meaning that allows computers to measure semantic proximity between different concepts, texts, or images.
Systematic processes and measurement tools to verify if an AI model's output is accurate, safe, and useful for a specific business case.
Method of providing a few reference examples in the prompt to the AI so that it learns to perform a specific task without the need for additional training.
Process of training a pre-existing AI model with specific data so that it learns tasks, tones of voice, or terminology specific to a niche or company.
The ability of an AI model to identify that it needs an external tool to perform a task and generate the necessary command to execute it using structured data.
Content optimization strategy to ensure a brand is cited and recommended by generative AI models and conversational search engines.
Set of rules and restrictions applied to AI models to ensure that responses are safe, accurate, within the company's tone of voice, and free from hallucinations.
A phenomenon where AI models generate factually incorrect information or information without a basis in reality, presenting it with an excessive and misleading tone of confidence.
Data organization algorithm that allows for quickly finding similar information in large-scale databases, functioning as a network of intelligent 'shortcuts'.
Combines the precision of traditional keyword matching with the semantic understanding of artificial intelligence to optimize information retrieval in databases.
Artificial intelligence models trained on massive volumes of text to understand, generate, and process human language with high fluency.
Efficient training technique that allows adapting large-scale AI models to specific tasks without spending fortunes on computing or time.
Open standard that allows connecting AI models to external data sources and tools universally, replacing custom and isolated integrations.
A set of practices that automates and standardizes the AI model lifecycle, from the experimentation phase to production maintenance with reliability.
The ability of an AI system to process and relate different types of information, such as text, image, audio, and video, simultaneously and in an integrated manner.
The ability to understand the internal state of an AI system by analyzing the data it generates, allowing for the identification of the root cause of problems instead of just detecting failures.
Technology that converts different types of documents, such as images or scanned PDFs, into computer-editable and searchable data.
The process of structuring input instructions for generative AI models to obtain more accurate, consistent, and useful results for business tasks.
A manipulation technique where a user inserts malicious instructions into an AI model to bypass security filters, ignore original rules, or extract confidential data.
Technique that allows an AI model to consult up-to-date and private external documents before generating a response, ensuring greater accuracy and reducing hallucinations.
An AI component that reorganizes search results to place the most relevant ones at the top, correcting inaccuracies from the initial search engine.
Educational process that adjusts the behavior of a language model through feedback from human specialists to ensure safe, useful responses aligned with real intentions.
The strategic transition from traditional search engine optimization (Google) to visibility in generative AI systems and direct answers (ChatGPT, Perplexity).
Ensures that the AI model responds in a rigid and predictable technical format, such as JSON, instead of narrative free text.
Artificially generated information by algorithms or AI models that replicates the statistical characteristics of real data without exposing sensitive or private information.
Parameter that controls the balance between predictability and creativity in a language model's responses, defining the level of randomness in word choice.
The minimum unit of processing that AI models use to read and generate text, usually corresponding to pieces of words or individual characters.
Neural network architecture based on the attention mechanism, which processes data in parallel and efficiently identifies complex relationships in information sequences.