AI for EverybodyLearn the intelligence era with better judgment.

Your progress: 0 complete

Glossary My Map Account

Glossary My Map AccountQuick answers Choose my route History of AI Sources About this platform

You are here

Current learning moveOrient

Clarify a term, inspect evidence, or place an idea in context.

Do nowSearch the term blocking you

Account

Connected glossary

Understand the language of AI

Clear definitions, concrete examples, common misunderstandings, and routes into neighboring concepts.

A concept network, not an isolated listFind a term, then follow its relationships into neighboring ideas.

Artificial intelligence
Algorithm
Machine learning
Deep learning
Neural network

What the Glossary is for

Understand one term without losing its context

The Glossary is a connected reference system, not a vocabulary test. Each definition links to examples, misunderstandings, neighboring concepts, and the learning object where the term matters.

Best when: When unfamiliar language interrupts what you are trying to understand.
First move: Search one word, open its full concept page, then follow a related term only if useful.

Search concepts80 results

A13

Infrastructure, openness, and governance

Accelerator

An accelerator is specialized hardware designed to perform selected computations faster or more efficiently than a general-purpose processor.

Agents and AI-native software

Agent

An AI agent is a system that continues across steps toward a goal and can select or use actions, tools, or information within defined controls.

Agents and AI-native software

Agentic system

An agentic system is the full technical and organizational arrangement that enables goal-directed continuation and action.

AGI and possible futures

AGI

AGI, or artificial general intelligence, is a contested family of terms for AI with broad, flexible capability across many tasks or environments.

Truth, trust, and safety · Agents and AI-native software

AI incident

An AI incident is an event where an AI system causes, contributes to, or creates a credible risk of harm, failure, or rights impact.

Truth, trust, and safety · Education and human capability

AI literacy

AI literacy is the context-sensitive knowledge, skills, and judgment needed to understand, use, evaluate, and shape AI systems responsibly.

Map of AI · Data, training, and inference

Algorithm

An algorithm is a defined procedure for turning inputs into results through a sequence of operations or rules.

Generative AI · Truth, trust, and safety

Alignment

Alignment is the effort to make an AI system’s behavior accord with specified human intentions, values, rules, or objectives.

Agents and AI-native software · Infrastructure, openness, and governance

API

An API, or application programming interface, is a defined way for software components to request functions or exchange data.

Map of AI

Artificial intelligence

Artificial intelligence is a field and family of systems that infer how to produce outputs from inputs in pursuit of explicit or implicit objectives.

AGI and possible futures

ASI

ASI, or artificial superintelligence, is a speculative term for AI capability that substantially exceeds humans across most or all relevant cognitive domains.

Agents and AI-native software · Practical AI use

Automation

Automation is the use of technology to perform part or all of a process with reduced direct human execution.

Agents and AI-native software

Autonomy

Autonomy is the degree of freedom a system has to choose, continue, or act without a person directing each step.

Data, training, and inference

Benchmark

A benchmark is a standardized task, dataset, procedure, and metric used to compare model or system performance.

Truth, trust, and safety · Data, training, and inference

Bias

Bias is a systematic tendency in data, measurement, design, model behavior, or institutions that shifts results in particular directions.

Data, training, and inference

Classification

Classification is assigning an input to one or more defined categories, often with a score or probability for each.

Infrastructure, openness, and governance

Closed model

A closed model is a model whose key artifacts, internal details, or operating access are not publicly released beyond provider-defined interfaces and terms.

Infrastructure, openness, and governance

Compute

Compute is the processing capacity and work supplied by hardware to train, run, and operate digital models and systems.

Generative AI

Context window

A context window is the limited amount of tokenized information a model can consider together during one generation or inference.

Infrastructure, openness, and governance

Data center

A data center is a facility and operating environment that houses computing, storage, networking, power, cooling, and security systems.

Data, training, and inference

Dataset

A dataset is an organized collection of examples or records used for training, evaluation, analysis, or system operation.

Map of AI · Data, training, and inference

Deep learning

Deep learning is machine learning built with neural networks containing multiple layers of learned transformations.

Truth, trust, and safety · Generative AI

Deepfake

A deepfake is highly realistic synthetic or manipulated media that depicts a person, voice, event, or action in a materially false way.

Data, training, and inference

Distribution shift

Distribution shift is a meaningful difference between the data or conditions used to develop a model and those encountered later.

Generative AI · Data, training, and inference

Embedding

An embedding is a numerical vector that represents an item so useful relationships can be measured or learned.

Data, training, and inference · Truth, trust, and safety

Evaluation

Evaluation is the structured measurement and investigation of how a model or system performs against defined questions, criteria, and contexts.

Truth, trust, and safety

Evidence status

Evidence status is a visible label describing how strongly and consistently current sources support a particular claim.

Truth, trust, and safety

Explainability

Explainability is the extent to which relevant people can receive a useful account of how or why a system produced a result.

Truth, trust, and safety

Fairness

Fairness is the context-dependent quality of treating people or groups according to justified criteria, rights, needs, and procedures.

Data, training, and inference

Feature

A feature is an input property or representation used by a model when producing an output.

Generative AI · Data, training, and inference

Fine-tuning

Fine-tuning is additional training that adapts a pretrained model to selected tasks, domains, examples, or behavior goals.

Generative AI

Foundation model

A foundation model is trained broadly enough to be adapted or used across multiple downstream tasks and applications.

Agents and AI-native software

Function calling

Function calling lets a model propose a structured function name and arguments for software to validate and optionally execute.

Data, training, and inference

Generalization

Generalization is a model’s ability to perform usefully on relevant examples or conditions not seen during training.

Generative AI

Generative AI produces new text, images, audio, video, code, or other content in response to context and controls.

Generative AI · Truth, trust, and safety

Grounding

Grounding connects a model’s output to selected external information, observations, rules, or evidence relevant to the task.

Agents and AI-native software · Truth, trust, and safety

Guardrail

A guardrail is a preventive, detective, or corrective control intended to keep system behavior within defined boundaries.

Generative AI · Truth, trust, and safety

Hallucination

A hallucination is generated content that is unsupported, inaccurate, or invented while often remaining fluent and plausible.

Truth, trust, and safety

High-risk use

A high-risk use is an application where failure, misuse, or unequal performance can create serious consequences for people, rights, safety, livelihood, or essential services.

Agents and AI-native software · Truth, trust, and safety

Human in the loop

Human in the loop describes a process where a person performs a defined review, decision, correction, or approval within a system’s operating cycle.

Agents and AI-native software · Truth, trust, and safety

Human oversight

Human oversight is the wider capacity of responsible people and institutions to understand, govern, monitor, challenge, and intervene in an AI system.

Data, training, and inference

Inference

Inference is running a trained model on an input to produce a prediction, generation, score, or other output.

Data, training, and inference

Label

A label is a target value or category attached to an example for training or evaluating a supervised model.

Generative AI

Large language model

A large language model, or LLM, is a parameter-rich model trained on language sequences to predict and generate tokens in context.

Map of AI · Data, training, and inference

Machine learning

Machine learning is an approach in which a model’s behavior is shaped by finding patterns in data or experience rather than only by fixed task rules.

Agents and AI-native software

Memory

Memory is information a system retains or can retrieve for use beyond the immediate step or context.

Truth, trust, and safety

Misinformation

Misinformation is false, inaccurate, or materially misleading information shared regardless of whether the sharer intended to deceive.

Data, training, and inference

Model

A model is a learned or designed representation that maps inputs to outputs for a defined task or family of tasks.

Data, training, and inference · Truth, trust, and safety

Model card

A model card is structured documentation describing a model’s intended uses, evaluation, limitations, development context, and relevant risks.

Generative AI · Map of AI

Multimodal model

A multimodal model processes or produces more than one kind of data, such as text, images, audio, video, or sensor signals.

Data, training, and inference

Neural network

A neural network is a parameterized computing structure that passes information through connected units and transformations to learn an input–output relationship.

Agents and AI-native software · Truth, trust, and safety

Observability

Observability is the capacity to understand a system’s relevant internal state and behavior from designed records, signals, and traces.

Infrastructure, openness, and governance

Open model

An open model is a model for which selected artifacts and usage rights are made publicly available, with the degree of openness stated explicitly.

Infrastructure, openness, and governance

Open source

Open source describes software whose source code is available under a license granting defined rights to use, inspect, modify, and redistribute it.

Agents and AI-native software

Orchestration

Orchestration coordinates multiple tasks, models, tools, services, agents, or people so their outputs form one process.

Data, training, and inference

Overfitting

Overfitting occurs when a model fits training examples or their noise too closely and performs less well on relevant new cases.

Data, training, and inference

Parameters

Parameters are internal numerical values that a model’s training procedure adjusts to shape its behavior.

Agents and AI-native software · Truth, trust, and safety

Permission

A permission is an authorized ability to access information, use a tool, or perform an action under defined conditions.

Truth, trust, and safety · Practical AI use

Personal data

Personal data is information that relates to an identified or reasonably identifiable person, directly or through combination and inference.

Data, training, and inference

Prediction

A prediction is a model’s estimated output for a given input, including a class, value, probability, sequence, or next element.

Truth, trust, and safety

Primary source

A primary source is original evidence or a first-hand record produced by the people, instrument, institution, or process directly involved.

Truth, trust, and safety · Practical AI use

Privacy

Privacy concerns appropriate control, expectations, and risk around how information about people is collected, inferred, used, shared, retained, and deleted.

Generative AI · Practical AI use

Prompt

A prompt is input supplied to a generative model or system to shape the task, context, constraints, or desired output.

Truth, trust, and safety

Provenance

Provenance is the recorded origin, custody, transformation, and history of data, media, evidence, or a system artifact.

Agents and AI-native software · Truth, trust, and safety

Recovery

Recovery is the ability to stop, contain, reverse, retry, compensate for, or safely continue after failure or interruption.

Data, training, and inference

Regression

Regression is a modeling task that predicts a numerical value rather than a discrete category.

Generative AI

Retrieval-augmented generation

Retrieval-augmented generation, or RAG, retrieves relevant material and places it in context for a generative model to use.

Map of AI · Agents and AI-native software

Robotics

Robotics is the field of designing, building, sensing, controlling, and operating machines that act in the physical world.

Truth, trust, and safety · Agents and AI-native software

Security

Security protects systems and information against unauthorized access, change, disclosure, disruption, or destruction.

Truth, trust, and safety · Practical AI use

Sensitive data

Sensitive data is information whose exposure, misuse, inference, or alteration could create substantial harm or requires heightened protection.

Agents and AI-native software

State

State is the information that describes a system or task at a particular point and is carried into later steps.

Truth, trust, and safety · Generative AI

Synthetic media

Synthetic media is text, imagery, audio, video, or other media created or materially altered by computational generation.

Generative AI · Agents and AI-native software

System instruction

A system instruction is configuration context intended to guide a model’s behavior across a session or task, often with priority over ordinary user input.

Generative AI

Temperature

Temperature is a generation setting that changes how strongly token sampling favors higher-probability options.

Generative AI

Token

A token is a unit into which a model’s input or output is divided for processing.

Agents and AI-native software

Tool use

Tool use is a system’s ability to invoke an external function, service, interface, or device and incorporate the result.

Data, training, and inference

Training

Training is the process of adjusting a model to improve performance on an objective using data, feedback, or interaction.

Truth, trust, and safety

Transparency

Transparency is making relevant information about a system, its use, evidence, limits, and governance available to the people who need it.

Truth, trust, and safety · Practical AI use

Verification

Verification is the process of checking whether a claim, output, action, identity, or result is adequately supported for its intended use.

Agents and AI-native software

Workflow

A workflow is an organized sequence of tasks, decisions, handoffs, information, and checks used to produce an outcome.

Loading page content...

…