https://coursiv.io/blog/categories/ai-glossaryRecent content in AI Glossary on Coursiv BlogHugo -- 0.147.0en-UShttps://coursiv.io/blog/glossary/action-spaceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/action-spaceThe complete set of things an agent is permitted to do — its tools, APIs, and permissions.https://coursiv.io/blog/glossary/activation-functionMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/activation-functionThe non-linear step applied at each unit, without which a deep network would collapse into a single linear one.https://coursiv.io/blog/glossary/adversarial-exampleMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/adversarial-exampleAn input perturbed just enough to fool a model, while looking unchanged to a person.https://coursiv.io/blog/glossary/agent-memoryMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/agent-memoryHow an agent retains information across steps and sessions, beyond the context window.https://coursiv.io/blog/glossary/agentic-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/agentic-aiAI that acts — plans, uses tools, self-corrects — rather than only producing text.https://coursiv.io/blog/glossary/agentic-loopMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/agentic-loopThe cycle an agent repeats: reason about what to do, act, observe the result, decide again.https://coursiv.io/blog/glossary/ai-agentMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-agentSoftware that plans and takes actions toward a goal on your behalf, rather than only answering.https://coursiv.io/blog/glossary/ai-alignmentMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-alignmentThe problem of making AI systems pursue what we actually want, including things we never thought to specify.https://coursiv.io/blog/glossaryMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossaryA plain-English glossary of the AI terms that actually come up at work — from tokens and RAG to the EU AI Act. Compiled from Google's ML glossary, NIST and Regulation (EU) 2024/1689.https://coursiv.io/blog/glossary/ai-governanceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-governanceThe internal structure of decisions, owners, and controls for how an organisation builds and uses AI.https://coursiv.io/blog/glossary/ai-incidentMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-incidentAn event where an AI system causes or nearly causes harm — increasingly something you must record and report.https://coursiv.io/blog/glossary/ai-literacyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-literacyUnder the EU AI Act: the skills and understanding staff need to use AI systems knowingly — and a legal obligation since February 2025.https://coursiv.io/blog/glossary/ai-regulatory-sandboxMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-regulatory-sandboxUnder the EU AI Act: a supervised environment where innovators can develop and test AI systems under a regulator's eye.https://coursiv.io/blog/glossary/ai-safetyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-safetyThe field concerned with preventing AI systems from causing harm, from everyday failures to systemic risks.https://coursiv.io/blog/glossary/ai-slopMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-slopLow-quality AI-generated content produced for volume, not value.https://coursiv.io/blog/glossary/ai-systemMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ai-systemUnder the EU AI Act: a machine-based system that operates with some autonomy and infers outputs from inputs.https://coursiv.io/blog/glossary/algorithmic-biasMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/algorithmic-biasWhen a system produces systematically different outcomes for different groups, without justification.https://coursiv.io/blog/glossary/apiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/apiThe interface your code uses to send a prompt to a model and get a response back.https://coursiv.io/blog/glossary/api-keyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/api-keyThe secret credential that authenticates and bills your API calls — and must never reach a browser.https://coursiv.io/blog/glossary/artificial-general-intelligenceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/artificial-general-intelligenceA hypothetical system matching human breadth across essentially any cognitive task — no agreed definition, no agreed timeline.https://coursiv.io/blog/glossary/artificial-intelligenceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/artificial-intelligenceSoftware that performs tasks we normally associate with human thinking — recognising, predicting, generating, deciding.https://coursiv.io/blog/glossary/attention-mechanismMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/attention-mechanismThe operation that lets a model weigh which parts of the input matter most for each part of the output.https://coursiv.io/blog/glossary/autonomyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/autonomyHow far a system acts without human involvement — a dial, not a switch.https://coursiv.io/blog/glossary/backpropagationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/backpropagationThe algorithm that works out how much each weight contributed to the error, so gradient descent can fix it.https://coursiv.io/blog/glossary/batch-inferenceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/batch-inferenceRunning many predictions together as a job instead of one at a time in real time.https://coursiv.io/blog/glossary/batch-sizeMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/batch-sizeHow many examples the model processes before updating its weights once.https://coursiv.io/blog/glossary/benchmarkMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/benchmarkA standard test set used to compare models — useful for direction, unreliable as a guarantee.https://coursiv.io/blog/glossary/biasMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/biasSystematic skew in a model's outputs — a data and design property, not a moral accusation.https://coursiv.io/blog/glossary/biometric-categorisationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/biometric-categorisationUnder the EU AI Act: assigning people to categories based on their biometric data — restricted, and prohibited for sensitive attributes.https://coursiv.io/blog/glossary/ce-markingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ce-markingThe mark declaring that a high-risk AI system conforms to applicable EU requirements.https://coursiv.io/blog/glossary/chain-of-thought-promptingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/chain-of-thought-promptingAsking the model to work through its reasoning step by step before answering.https://coursiv.io/blog/glossary/chunkingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/chunkingSplitting documents into passages small enough to embed and retrieve usefully.https://coursiv.io/blog/glossary/classificationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/classificationPredicting which category something belongs to.https://coursiv.io/blog/glossary/computer-useMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/computer-useAn agent operating a computer the way a person does — reading the screen, moving the cursor, typing.https://coursiv.io/blog/glossary/conformity-assessmentMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/conformity-assessmentThe procedure demonstrating that a high-risk AI system meets the Act's requirements before it goes on the market.https://coursiv.io/blog/glossary/content-credentialsMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/content-credentialsCryptographically signed metadata recording how a piece of media was created and edited.https://coursiv.io/blog/glossary/context-engineeringMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/context-engineeringDeciding what information goes into the context window, in what order, and what gets left out.https://coursiv.io/blog/glossary/context-windowMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/context-windowThe maximum number of tokens a model can consider at once — prompt, documents, and its own answer combined.https://coursiv.io/blog/glossary/copilotMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/copilotAn AI assistant embedded in a tool that suggests while you keep control.https://coursiv.io/blog/glossary/data-poisoningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/data-poisoningCorrupting training data so the resulting model misbehaves, often only on a specific trigger.https://coursiv.io/blog/glossary/data-provenanceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/data-provenanceA documented record of where data came from, how it was collected, and what may be done with it.https://coursiv.io/blog/glossary/deep-learningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/deep-learningMachine learning built on neural networks with many layers — the approach behind almost every modern AI system.https://coursiv.io/blog/glossary/deepfakeMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/deepfakeSynthetic audio or video that convincingly depicts a real person doing or saying something they did not.https://coursiv.io/blog/glossary/deployerMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/deployerUnder the EU AI Act: whoever uses an AI system under their own authority in a professional capacity.https://coursiv.io/blog/glossary/differential-privacyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/differential-privacyA mathematical guarantee that a result barely changes whether or not any one person's data was included.https://coursiv.io/blog/glossary/diffusion-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/diffusion-modelThe architecture behind most AI image and video generation: start from noise, remove it step by step.https://coursiv.io/blog/glossary/distillationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/distillationTraining a small model to imitate a large one, keeping most of the quality at a fraction of the cost.https://coursiv.io/blog/glossary/downstream-providerMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/downstream-providerUnder the EU AI Act: a provider whose system integrates a general-purpose AI model made by someone else.https://coursiv.io/blog/glossary/embeddingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/embeddingA list of numbers representing meaning, where similar things end up close together.https://coursiv.io/blog/glossary/emotion-recognition-systemMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/emotion-recognition-systemUnder the EU AI Act: a system that identifies or infers emotions or intentions from biometric data — banned in workplaces and schools.https://coursiv.io/blog/glossary/epochMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/epochOne full pass of the training algorithm over the entire training dataset.https://coursiv.io/blog/glossary/eu-ai-actMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/eu-ai-actThe EU's regulation on artificial intelligence — the first broad, binding AI law, applying by risk tier.https://coursiv.io/blog/glossary/evaluationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/evaluationSystematic measurement of whether an AI system does what you need — your tests, not the vendor's.https://coursiv.io/blog/glossary/explainabilityMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/explainabilityBeing able to give a human a meaningful account of why a system produced a particular output.https://coursiv.io/blog/glossary/fairnessMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/fairnessThe requirement that a system treat people equitably — with several mathematical definitions that provably conflict.https://coursiv.io/blog/glossary/featureMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/featureOne input variable a model uses to make its prediction.https://coursiv.io/blog/glossary/few-shot-promptingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/few-shot-promptingIncluding several examples in the prompt so the model infers the pattern you want.https://coursiv.io/blog/glossary/fine-tuningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/fine-tuningContinuing training on your own examples to specialise a general model's behaviour.https://coursiv.io/blog/glossary/foundation-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/foundation-modelA large model trained broadly once, then adapted to many downstream tasks instead of being built per task.https://coursiv.io/blog/glossary/frontier-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/frontier-modelA model at or beyond the current state of the art in general capability — the ones regulators watch most closely.https://coursiv.io/blog/glossary/function-callingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/function-callingThe mechanism by which a model asks your code to run a specific function with specific arguments.https://coursiv.io/blog/glossary/ganMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/ganTwo networks trained against each other — one generating fakes, one detecting them.https://coursiv.io/blog/glossary/general-purpose-ai-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/general-purpose-ai-modelUnder the EU AI Act: a model with significant generality that can perform a wide range of tasks and be built into many systems.https://coursiv.io/blog/glossary/generative-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/generative-aiModels that produce new content — text, images, audio, video, code — rather than only classifying or scoring existing content.https://coursiv.io/blog/glossary/gpuMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/gpuThe parallel processor that made deep learning practical, and the scarcest resource in the industry.https://coursiv.io/blog/glossary/gradient-descentMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/gradient-descentThe optimisation method that finds better parameters by repeatedly stepping downhill on the loss.https://coursiv.io/blog/glossary/guardrailsMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/guardrailsControls that constrain what a model can be asked or allowed to output.https://coursiv.io/blog/glossary/hallucinationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/hallucinationFluent, confident output that is simply false — the defining failure mode of language models.https://coursiv.io/blog/glossary/high-risk-ai-systemMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/high-risk-ai-systemUnder the EU AI Act: a system in a listed sensitive use case, permitted but subject to the strictest obligations.https://coursiv.io/blog/glossary/human-oversightMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/human-oversightUnder the EU AI Act: the requirement that people can understand, monitor, override, and stop a high-risk system.https://coursiv.io/blog/glossary/human-in-the-loopMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/human-in-the-loopA design where a person reviews or approves before an AI action takes effect.https://coursiv.io/blog/glossary/hybrid-searchMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/hybrid-searchCombining keyword and vector search, because each fails where the other works.https://coursiv.io/blog/glossary/hyperparameterMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/hyperparameterA setting you choose before training, as opposed to a parameter the model learns during it.https://coursiv.io/blog/glossary/inferenceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/inferenceRunning a trained model to get an answer — everything that happens after training is done.https://coursiv.io/blog/glossary/inference-costMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/inference-costWhat it costs to run a model in production, billed per token in and per token out.https://coursiv.io/blog/glossary/interpretabilityMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/interpretabilityUnderstanding how a model works internally, as opposed to just describing its outputs.https://coursiv.io/blog/glossary/iso-42001Mon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/iso-42001The international standard for an AI management system — certifiable, unlike the NIST framework.https://coursiv.io/blog/glossary/jailbreakMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/jailbreakA prompt crafted to get a model to bypass its own safety rules.https://coursiv.io/blog/glossary/knowledge-cutoffMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/knowledge-cutoffThe date after which a model's training data ends — everything later is invisible to it.https://coursiv.io/blog/glossary/labelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/labelThe correct answer attached to a training example.https://coursiv.io/blog/glossary/large-language-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/large-language-modelA model trained on huge amounts of text to predict the next token, which turns out to be enough to write, summarise, translate, and reason.https://coursiv.io/blog/glossary/latencyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/latencyHow long a request takes — the difference between an assistant that feels alive and one that feels broken.https://coursiv.io/blog/glossary/learning-rateMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/learning-rateHow big a step training takes each update — the hyperparameter most likely to ruin a run.https://coursiv.io/blog/glossary/llmopsMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/llmopsMLOps for systems built on language models — prompts, evals, cost, and safety instead of training runs.https://coursiv.io/blog/glossary/loraMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/loraA cheap fine-tuning method that trains a small add-on instead of updating the whole model.https://coursiv.io/blog/glossary/loss-functionMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/loss-functionThe formula that scores how wrong a model's prediction is — the thing training tries to minimise.https://coursiv.io/blog/glossary/machine-learningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/machine-learningA way of building software where the system learns patterns from examples instead of following hand-written rules.https://coursiv.io/blog/glossary/meta-promptingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/meta-promptingUsing a model to write, critique, or improve prompts for another model.https://coursiv.io/blog/glossary/mixture-of-expertsMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/mixture-of-expertsAn architecture where only a fraction of the model's parameters activate for any given token.https://coursiv.io/blog/glossary/mlopsMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/mlopsThe practices for getting machine learning models into production and keeping them working.https://coursiv.io/blog/glossary/modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/modelThe trained artefact — the learned parameters plus the architecture — that turns an input into an output.https://coursiv.io/blog/glossary/model-cardMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/model-cardA short standard document stating what a model is for, how it was evaluated, and where it fails.https://coursiv.io/blog/glossary/model-context-protocolMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/model-context-protocolAn open standard for connecting AI applications to tools and data sources through one common interface.https://coursiv.io/blog/glossary/model-driftMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/model-driftSilent degradation as the world moves away from the data a model was trained on.https://coursiv.io/blog/glossary/multi-agent-systemMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/multi-agent-systemSeveral specialised agents working together, usually with a coordinator dividing the work.https://coursiv.io/blog/glossary/multimodal-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/multimodal-aiA model that handles more than one kind of input or output — text, images, audio, video.https://coursiv.io/blog/glossary/narrow-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/narrow-aiAI built for a specific task, which is every AI system that currently exists.https://coursiv.io/blog/glossary/negative-promptMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/negative-promptIn image generation, a list of things you want kept out of the result.https://coursiv.io/blog/glossary/neural-networkMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/neural-networkA model made of layers of connected units that pass numbers forward, loosely inspired by neurons.https://coursiv.io/blog/glossary/nist-ai-rmfMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/nist-ai-rmfA voluntary US framework for managing AI risk, organised around Govern, Map, Measure, Manage.https://coursiv.io/blog/glossary/on-device-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/on-device-aiRunning a model on a phone or laptop instead of sending data to a server.https://coursiv.io/blog/glossary/one-shot-promptingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/one-shot-promptingGiving exactly one worked example alongside the instruction.https://coursiv.io/blog/glossary/open-weights-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/open-weights-modelA model whose trained weights you can download and run yourself — not necessarily open source in the strict sense.https://coursiv.io/blog/glossary/orchestrationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/orchestrationThe layer that decides which model, tool, or agent handles each step, and in what order.https://coursiv.io/blog/glossary/overfittingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/overfittingWhen a model memorises its training data instead of learning the pattern, and fails on anything new.https://coursiv.io/blog/glossary/parameterMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/parameterOne of the internal numbers a model adjusts during training; parameter count is a rough proxy for model capacity.https://coursiv.io/blog/glossary/post-market-monitoringMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/post-market-monitoringUnder the EU AI Act: providers must keep watching how their system performs in the real world after launch.https://coursiv.io/blog/glossary/precision-and-recallMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/precision-and-recallThe two metrics that actually tell you whether a classifier works when one class is rare.https://coursiv.io/blog/glossary/prohibited-ai-practiceMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/prohibited-ai-practiceUnder the EU AI Act: uses of AI banned outright in the EU, in force since February 2025.https://coursiv.io/blog/glossary/promptMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/promptThe input you give a model — instructions, context, examples, and question, all as text.https://coursiv.io/blog/glossary/prompt-cachingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/prompt-cachingReusing the processed form of a repeated prompt prefix so you do not pay full price for it again.https://coursiv.io/blog/glossary/prompt-engineeringMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/prompt-engineeringDesigning and iterating on prompts to get reliable output — closer to spec-writing than to magic words.https://coursiv.io/blog/glossary/prompt-injectionMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/prompt-injectionHiding instructions in content the model reads, so a third party effectively takes over the prompt.https://coursiv.io/blog/glossary/prompt-templateMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/prompt-templateA reusable prompt with variable slots, so a working prompt becomes a repeatable asset.https://coursiv.io/blog/glossary/providerMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/providerUnder the EU AI Act: whoever develops an AI system or GPAI model and puts it on the market under their own name or trademark.https://coursiv.io/blog/glossary/quantizationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/quantizationStoring model weights at lower numeric precision to cut memory and cost, with a small quality trade-off.https://coursiv.io/blog/glossary/rate-limitMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/rate-limitThe cap a provider puts on how many requests or tokens you can use per interval.https://coursiv.io/blog/glossary/react-patternMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/react-patternAn agent design that interleaves explicit reasoning with actions, one step at a time.https://coursiv.io/blog/glossary/reasoning-modelMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/reasoning-modelA model trained to spend extra compute working through a problem before answering.https://coursiv.io/blog/glossary/red-teamingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/red-teamingDeliberately attacking your own AI system to find failures before someone else does.https://coursiv.io/blog/glossary/regressionMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/regressionPredicting a number rather than a category.https://coursiv.io/blog/glossary/regularizationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/regularizationAny technique that deliberately constrains a model to stop it memorising the training set.https://coursiv.io/blog/glossary/reinforcement-learningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/reinforcement-learningLearning by trial and error, guided by rewards rather than labelled answers.https://coursiv.io/blog/glossary/responsible-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/responsible-aiThe practice of building and deploying AI with its effects on people deliberately considered.https://coursiv.io/blog/glossary/retrieval-augmented-generationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/retrieval-augmented-generationFetching relevant documents first and putting them in the prompt, so the model answers from sources instead of memory.https://coursiv.io/blog/glossary/rlhfMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/rlhfTraining a model on human preference judgements so it becomes helpful and well-behaved rather than merely fluent.https://coursiv.io/blog/glossary/robustnessMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/robustnessWhether a system keeps performing when inputs are noisy, unusual, or deliberately adversarial.https://coursiv.io/blog/glossary/role-promptingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/role-promptingAssigning the model a role to shape its vocabulary, depth, and framing.https://coursiv.io/blog/glossary/sdkMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/sdkAn official library that wraps a provider's API in your programming language.https://coursiv.io/blog/glossary/self-supervised-learningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/self-supervised-learningTraining where the data generates its own labels — the trick that made large language models possible.https://coursiv.io/blog/glossary/shadow-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/shadow-aiEmployees using AI tools that the organisation has not approved and cannot see.https://coursiv.io/blog/glossary/speech-to-textMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/speech-to-textTurning spoken audio into written text.https://coursiv.io/blog/glossary/structured-outputMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/structured-outputForcing the model to return data in a fixed schema so downstream code can parse it.https://coursiv.io/blog/glossary/supervised-learningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/supervised-learningLearning from examples that come with the right answer attached.https://coursiv.io/blog/glossary/synthetic-dataMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/synthetic-dataTraining or test data generated by a model rather than collected from the real world.https://coursiv.io/blog/glossary/system-promptMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/system-promptStanding instructions that set a model's role, rules, and tone for an entire conversation.https://coursiv.io/blog/glossary/systemic-riskMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/systemic-riskUnder the EU AI Act: risk arising from the high-impact capabilities of the most capable general-purpose models.https://coursiv.io/blog/glossary/technical-documentationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/technical-documentationUnder the EU AI Act: the file describing how a system was built, tested, and controlled, kept ready for authorities.https://coursiv.io/blog/glossary/temperatureMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/temperatureThe setting that controls randomness in generation — low for consistency, high for variety.https://coursiv.io/blog/glossary/test-dataMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/test-dataData the model has never seen, used once at the end to estimate real-world performance.https://coursiv.io/blog/glossary/test-time-computeMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/test-time-computeSpending more computation at answer time — thinking longer — to get better results without retraining.https://coursiv.io/blog/glossary/text-to-imageMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/text-to-imageGenerating an image from a written description.https://coursiv.io/blog/glossary/text-to-speechMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/text-to-speechGenerating spoken audio from written text.https://coursiv.io/blog/glossary/text-to-videoMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/text-to-videoGenerating video clips from a written description.https://coursiv.io/blog/glossary/throughputMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/throughputHow much work a system handles per unit of time — tokens per second, requests per minute.https://coursiv.io/blog/glossary/tokenMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/tokenThe unit a model actually reads and writes — usually a word fragment, not a word.https://coursiv.io/blog/glossary/tokenizationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/tokenizationCutting text into the tokens a model can process — and a quiet source of odd model behaviour.https://coursiv.io/blog/glossary/tool-useMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/tool-useA model calling external software — search, code, APIs, databases — instead of answering from memory.https://coursiv.io/blog/glossary/top-pMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/top-pAn alternative randomness control that samples only from the smallest set of tokens covering probability p.https://coursiv.io/blog/glossary/tpuMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/tpuGoogle's custom accelerator chip, purpose-built for machine learning workloads.https://coursiv.io/blog/glossary/training-dataMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/training-dataThe examples a model learns from — the single biggest determinant of what it can and cannot do.https://coursiv.io/blog/glossary/transformerMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/transformerThe neural network architecture behind essentially every modern language model, built around attention.https://coursiv.io/blog/glossary/transparencyMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/transparencyMaking it clear what an AI system is, what it was trained on, and when someone is interacting with it.https://coursiv.io/blog/glossary/trustworthy-aiMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/trustworthy-aiAn umbrella for the properties an AI system needs to be relied on: valid, safe, secure, accountable, explainable, fair, privacy-enhanced.https://coursiv.io/blog/glossary/underfittingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/underfittingWhen a model is too simple to capture the pattern, and is wrong on training and new data alike.https://coursiv.io/blog/glossary/unsupervised-learningMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/unsupervised-learningFinding structure in data that has no labels — grouping, compressing, spotting outliers.https://coursiv.io/blog/glossary/validation-dataMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/validation-dataA held-out slice of data used to tune settings and check progress while training, without touching the test set.https://coursiv.io/blog/glossary/vector-databaseMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/vector-databaseA database built to store embeddings and find the nearest ones fast.https://coursiv.io/blog/glossary/vector-searchMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/vector-searchFinding content by meaning rather than by matching keywords.https://coursiv.io/blog/glossary/watermarkingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/watermarkingEmbedding a detectable signal in AI-generated content to mark its origin.https://coursiv.io/blog/glossary/weightsMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/weightsThe learned strengths of connections in a neural network — the substance of what a model knows.https://coursiv.io/blog/glossary/workflow-automationMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/workflow-automationChaining AI steps into a defined process that runs the same way every time.https://coursiv.io/blog/glossary/zero-shot-promptingMon, 01 Jan 0001 00:00:00 +0000https://coursiv.io/blog/glossary/zero-shot-promptingAsking the model to do a task with no examples — just the instruction.