Glossary · AI & DataA/B TestingA/B testing is a controlled experiment that shows two versions of a page or feature to different users to learn which performs better on a chosen metric.Glossary · AI & DataAGI (Artificial General Intelligence)AGI is a hypothetical form of artificial intelligence able to learn and perform any intellectual task a human can, rather than excelling only at narrow tasks.Glossary · AI & DataAI AgentAn AI agent is a system built around a language model that plans steps, calls tools or APIs, and acts toward a goal with limited human direction.Glossary · AI & DataAI AlignmentAI alignment is the research and engineering effort to make AI systems pursue the goals and values their designers and users actually intend.Glossary · AI & DataAI BenchmarkAn AI benchmark is a standardized test set and scoring method used to compare models on a task, such as reasoning, coding, or question answering.Glossary · AI & DataAI HallucinationAn AI hallucination is output from a model that sounds confident and plausible but is factually wrong, fabricated, or unsupported by its sources.Glossary · AI & DataAI ModelAn AI model is the trained mathematical system, defined by its architecture and learned parameters, that turns inputs into predictions or generated content.Glossary · AI & DataAlgorithmic Bias (Bias in AI)Algorithmic bias is systematic, unfair skew in an AI system's outputs, usually arising from unrepresentative data, flawed labels, or design choices.Glossary · AI & DataArtificial Intelligence (AI)Artificial intelligence is the field of building software that performs tasks normally needing human judgment, such as perceiving, reasoning, and language.Glossary · AI & DataBackpropagationBackpropagation is the algorithm that computes how much each weight in a neural network contributed to the error, so gradient descent can update them.Glossary · AI & DataBatch SizeBatch size is the number of training examples a model processes before updating its weights once, affecting speed, memory use and training stability.Glossary · AI & DataBig DataBig data describes datasets so large, fast-moving, or varied that ordinary single-machine tools struggle, requiring distributed storage and processing.Glossary · AI & DataBusiness Intelligence (BI)Business intelligence is the practice and tooling for collecting, analysing and presenting business data in reports and dashboards to support decisions.Glossary · AI & DataChain of ThoughtChain of thought is a prompting approach where a language model writes out intermediate reasoning steps before answering, often helping multi-step problems.Glossary · AI & DataChatbotA chatbot is a program that converses with users in natural language through text or voice, from simple scripted menus to AI-powered assistants.Glossary · AI & DataClassification (Machine Learning)Classification is a supervised learning task where a model assigns each input to one of a set of predefined categories, such as spam or not spam.Glossary · AI & DataClusteringClustering is an unsupervised learning method that groups similar data points together without predefined labels, revealing natural segments in the data.Glossary · AI & DataComputer VisionComputer vision is the field of AI that enables machines to interpret images and video, for tasks like object detection, face matching, and text reading.Glossary · AI & DataContext WindowThe context window is the maximum amount of text, measured in tokens, that a language model can consider at once, covering both your input and its reply.Glossary · AI & DataCorrelation vs CausationCorrelation means two variables move together, while causation means one produces a change in the other; correlation alone does not prove causation.Glossary · AI & DataCosine SimilarityCosine similarity measures how alike two vectors are by the cosine of the angle between them, ranging from -1 to 1 regardless of vector length.Glossary · AI & DataCSV (Comma-Separated Values)CSV is a plain-text file format that stores tabular data with one record per line and fields separated by commas, widely used to exchange spreadsheet data.Glossary · AI & DataDashboard (Data)A dashboard is a single screen that visualizes key metrics and charts so people can monitor performance and spot problems at a glance.Glossary · AI & DataData AnalyticsData analytics is the process of examining data to find patterns and answer questions, typically summarizing what happened, why, and what to do next.Glossary · AI & DataData CleaningData cleaning is the process of finding and fixing errors, duplicates, missing values, and inconsistencies in a dataset before analysis or modeling.Glossary · AI & DataData GovernanceData governance is the set of policies, roles and processes defining who can use which data, how, and how its quality, security and compliance are kept.Glossary · AI & DataData LabelingData labeling is the process of tagging raw data such as images, text or audio with correct answers so supervised models can learn from it.Glossary · AI & DataData LakeA data lake is a storage repository that holds large volumes of raw data in its original format, structured or unstructured, until it is needed.Glossary · AI & DataData PipelineA data pipeline is an automated series of steps that moves data from sources through processing to a destination, such as a database, warehouse, or model.Glossary · AI & DataData ScienceData science is the discipline of extracting insight and predictions from data by combining statistics, programming, and domain knowledge.Glossary · AI & DataData WarehouseA data warehouse is a central database optimized for analytics that stores cleaned, structured data from many sources for reporting and querying.Glossary · AI & DataDatasetA dataset is an organized collection of related data, usually arranged as rows and columns, used for analysis, reporting, or training machine learning models.Glossary · AI & DataDecision TreeA decision tree is a model that predicts outcomes by asking a sequence of yes or no questions about features, branching until it reaches a final answer.Glossary · AI & DataDeep LearningDeep learning is a type of machine learning that uses neural networks with many layers to learn complex patterns from large amounts of data.Glossary · AI & DataDiffusion ModelA diffusion model generates data such as images by learning to reverse a gradual noising process, turning random noise into a coherent output step by step.Glossary · AI & DataEmbeddingAn embedding is a list of numbers that represents the meaning of text, an image, or other data so that similar items sit close together in vector space.Glossary · AI & DataEpochAn epoch is one complete pass of the entire training dataset through a machine learning model during training.Glossary · AI & DataETL (Extract, Transform, Load)ETL is a data integration process that extracts data from sources, transforms it into a consistent form, and loads it into a target such as a warehouse.Glossary · AI & DataFew-Shot PromptingFew-shot prompting puts a handful of worked input and output examples in the prompt so a language model can infer the task and format without retraining.Glossary · AI & DataFine-TuningFine-tuning is the process of further training a pre-trained model on a smaller, task-specific dataset so it adopts a particular style, format, or skill.Glossary · AI & DataFunction Calling (Tool Use)Function calling, or tool use, lets a language model return a structured request to run a named function, so an application can fetch data or take actions.Glossary · AI & DataGAN (Generative Adversarial Network)A GAN is a generative model in which a generator and a discriminator network compete, pushing the generator to produce increasingly realistic synthetic data.Glossary · AI & DataGenerative AIGenerative AI refers to models that create new content such as text, images, audio, video, or code by learning patterns from existing data.Glossary · AI & DataGPU (Graphics Processing Unit)A GPU is a processor with thousands of small cores built for parallel math, now the standard hardware for training and running AI models.Glossary · AI & DataGradient DescentGradient descent is an optimisation algorithm that repeatedly adjusts model parameters in the direction that lowers the loss, using the loss function's slope.Glossary · AI & DataGuardrails (AI)AI guardrails are checks and rules placed around a model's inputs and outputs to keep behaviour safe, on-topic and compliant with policy.Glossary · AI & DataHistogramA histogram is a chart that shows the distribution of a numeric variable by counting how many values fall into consecutive ranges called bins.Glossary · AI & DataInference (AI)Inference is the stage where a trained AI model is run on new input to produce a prediction or generated output, as opposed to the training stage.Glossary · AI & DataKnowledge GraphA knowledge graph stores facts as entities and the typed relationships between them, so software can query connections and reason over linked data.Glossary · AI & DataLLM (Large Language Model)A large language model is a neural network trained on vast text to predict the next token, so it can write, summarize, translate, and answer questions.Glossary · AI & DataLoss FunctionA loss function is a formula that scores how wrong a model's predictions are, giving the training process a single number to minimise.Glossary · AI & DataMachine Learning (ML)Machine learning is a method where algorithms learn patterns from data to make predictions or decisions without being explicitly programmed with rules.Glossary · AI & DataMean vs MedianThe mean is the arithmetic average of values, while the median is the middle value when sorted; the median resists outliers and skew far better.Glossary · AI & DataMetadataMetadata is data that describes other data, such as a file's author, creation date, size, or format, helping systems and people find and interpret it.Glossary · AI & DataModel DistillationModel distillation trains a smaller student model to imitate the outputs of a larger teacher model, keeping much of its quality at lower cost and latency.Glossary · AI & DataModel ParametersModel parameters are the numeric values, mainly weights and biases, that a machine learning model learns during training and uses to turn inputs into outputs.Glossary · AI & DataMultimodal AIMultimodal AI refers to models that can process and sometimes generate more than one type of data, such as text, images, audio, and video.Glossary · AI & DataNeural NetworkA neural network is a machine learning model made of layers of connected nodes whose numeric weights are adjusted during training to map inputs to outputs.Glossary · AI & DataNLP (Natural Language Processing)Natural language processing is the field of AI concerned with enabling computers to read, interpret, and generate human language in text or speech.Glossary · AI & DataOpen-Weight ModelAn open-weight model has its trained parameters published for download so anyone can run or adapt it, though training data and code may stay closed.Glossary · AI & DataOutlierAn outlier is a data point that lies unusually far from the rest of the values, caused by error, rare events or genuine extreme behaviour.Glossary · AI & DataOverfittingOverfitting happens when a model memorizes noise and quirks of its training data, scoring well on that data but poorly on new, unseen examples.Glossary · AI & DataP-ValueA p-value is the probability of results at least as extreme as those observed if the null hypothesis were true; it is not the chance the hypothesis is true.Glossary · AI & DataPromptA prompt is the input text or instruction given to an AI model, such as a question, command, or example, that tells it what output to produce.Glossary · AI & DataPrompt EngineeringPrompt engineering is the practice of designing and refining instructions, context, and examples so an AI model produces reliable, useful output.Glossary · AI & DataQuantizationQuantization reduces a model's memory and compute needs by storing its weights, and sometimes activations, with fewer bits, such as 8-bit or 4-bit numbers.Glossary · AI & DataRAG (Retrieval-Augmented Generation)Retrieval-augmented generation is a technique that fetches relevant documents and adds them to a prompt so a language model answers from supplied sources.Glossary · AI & DataRandom ForestA random forest is an ensemble model that trains many decision trees on random samples of data and features, then combines their votes or averages.Glossary · AI & DataRecommendation SystemA recommendation system predicts which items a user is likely to want, such as products, videos or articles, using behaviour, content and similarity signals.Glossary · AI & DataRegression (Machine Learning)Regression is a supervised learning task where a model predicts a continuous numeric value, such as a price or response time, from input features.Glossary · AI & DataRLHF (Reinforcement Learning from Human Feedback)RLHF is a training method that tunes a language model using human preference rankings of its outputs, so replies become more helpful and better aligned.Glossary · AI & DataSampling (Statistics)Sampling is selecting a subset of a population to study so that conclusions about the whole group can be drawn without measuring every member.Glossary · AI & DataSemantic SearchSemantic search finds results by meaning rather than exact keywords, typically by comparing vector embeddings of the query and the documents.Glossary · AI & DataSpeech RecognitionSpeech recognition, or speech-to-text, is technology that converts spoken audio into written text using acoustic and language models.Glossary · AI & DataStandard DeviationStandard deviation measures how spread out values are around their mean, in the same units as the data; a larger value means more variability.Glossary · AI & DataStructured vs Unstructured DataStructured data fits a fixed schema of rows and columns, while unstructured data such as text, images, and audio has no predefined tabular format.Glossary · AI & DataSupervised LearningSupervised learning trains a model on labeled examples, where each input has a known correct answer, so it can predict labels for new, unseen data.Glossary · AI & DataSynthetic DataSynthetic data is artificially generated information that mimics the statistical patterns of real data, used for testing, privacy protection and model training.Glossary · AI & DataSystem PromptA system prompt is a set of instructions given to a language model ahead of user messages that defines its role, tone, rules and constraints for a conversation.Glossary · AI & DataTemperature (Sampling)Temperature is a sampling setting for language models that scales how random the next-token choice is: low values are predictable, high values are more varied.Glossary · AI & DataText-to-Speech (TTS)Text-to-speech converts written text into spoken audio using synthesis models, used for assistants, accessibility, navigation and narration.Glossary · AI & DataToken (AI)In language models, a token is a small chunk of text, often a word piece, that the model reads and generates, and the unit used to measure length and cost.Glossary · AI & DataTokenizationTokenization is the process of splitting text into tokens, such as words, subwords or characters, that a language model converts into numbers and processes.Glossary · AI & DataTop-p (Nucleus Sampling)Top-p, or nucleus sampling, limits a language model's next-token choice to the smallest set of tokens whose combined probability reaches a threshold p.Glossary · AI & DataTraining DataTraining data is the collection of examples a machine learning model learns from, whose quality, coverage, and labeling largely determine the model's behavior.Glossary · AI & DataTransformer (AI Architecture)The transformer is a neural network architecture using self-attention to weigh relationships across an input, and it underlies most modern language models.Glossary · AI & DataTuring TestThe Turing test, proposed by Alan Turing in 1950, judges a machine intelligent if a person cannot reliably tell it from a human in conversation.Glossary · AI & DataUnsupervised LearningUnsupervised learning finds hidden structure in unlabeled data, such as groups of similar items or unusual outliers, without being given correct answers.Glossary · AI & DataVector DatabaseA vector database stores embeddings and quickly finds the ones most similar to a query vector, enabling semantic search and retrieval for AI applications.Glossary · AI & DataZero-Shot PromptingZero-shot prompting asks a language model to perform a task using only instructions, with no worked examples included in the prompt.