--- source: 'https://howaiworks.ai/glossary' section: glossary count: 256 --- # Glossary > All 256 pages in the Glossary section. Each link is the Markdown twin; drop the `.md` for the HTML page. - [A/B Testing](https://howaiworks.ai/glossary/ab-testing.md) — updated 2026-07-25 A/B testing is the randomized experiment that decides whether a model, prompt, or feature change truly improves a live metric — or just moved on noise. - [Accountability](https://howaiworks.ai/glossary/accountability.md) — updated 2026-07-24 Who answers when an AI system causes harm — how the EU AI Act assigns duties by role, and the audit, disclosure and reporting mechanisms that enforce them. - [Activation Functions](https://howaiworks.ai/glossary/activation-functions.md) — updated 2026-07-21 Without a non-linearity, stacked layers collapse into one matrix. How activation functions fix that, and why ReLU, GELU and SwiGLU are the defaults. - [Active Learning](https://howaiworks.ai/glossary/active-learning.md) — updated 2026-07-25 Active learning lets a model choose which unlabelled examples to get labelled next, hitting target accuracy with far fewer labels than random labelling. - [Advanced Packaging](https://howaiworks.ai/glossary/advanced-packaging.md) — updated 2026-07-13 CoWoS, chiplets and 3D stacking — how compute dies and memory become one accelerator, and the assembly step that has actually gated AI chip supply. - [Agent Communication Protocol (ACP)](https://howaiworks.ai/glossary/agent-communication-protocol.md) — updated 2026-07-21 ACP is the horizontal link between AI agents, where MCP is the vertical link to tools. IBM's REST spec, merged into A2A in August 2025. - [Agent Memory](https://howaiworks.ai/glossary/agent-memory.md) — updated 2026-07-24 Agent memory is how an AI agent stores, retrieves, and summarizes information so it works across long tasks and sessions without overflowing its context window. - [Agent2Agent Protocol (A2A)](https://howaiworks.ai/glossary/agent2agent-protocol.md) — updated 2026-07-20 An open standard, governed by the Linux Foundation, that lets independent AI agents discover each other and delegate tasks across vendors and frameworks. - [Agentic Commerce](https://howaiworks.ai/glossary/agentic-commerce.md) — updated 2026-08-02 Buying where an AI agent handles discovery, comparison and checkout. Two rival protocols split the market: ACP behind ChatGPT, UCP behind Google. - [Agentic Workflow](https://howaiworks.ai/glossary/agentic-workflow.md) — updated 2026-07-24 An agentic workflow is a design pattern where an LLM plans, uses tools, and iterates in a loop to solve a task, instead of answering in a single shot. - [AI Agent](https://howaiworks.ai/glossary/ai-agent.md) — updated 2026-07-24 An AI agent is an LLM-powered system that plans, calls tools, and acts autonomously in a loop to finish multi-step tasks — not just answer questions. - [AI Architecture](https://howaiworks.ai/glossary/ai-architecture.md) — updated 2026-07-24 How a production AI system is assembled: retrieval, context assembly, the model call, tool execution, validation, logging and evaluation. - [AI Data Center](https://howaiworks.ai/glossary/ai-data-center.md) — updated 2026-07-24 Why AI facilities are measured in megawatts rather than servers — power density, liquid cooling, and the grid connection that has become the real constraint. - [AI Drug Discovery](https://howaiworks.ai/glossary/ai-drug-discovery.md) — updated 2026-07-21 What AI actually does at each stage of drug development, why it speeds up the cheap part, and where AI-designed molecules really stand in clinical trials. - [AI and Employment](https://howaiworks.ai/glossary/ai-employment.md) — updated 2026-07-22 Will AI take your job? Exposure concentrates in tasks, not whole occupations — what Frey-Osborne, the OECD and measured productivity studies found. - [AI Energy Consumption](https://howaiworks.ai/glossary/ai-energy-consumption.md) — updated 2026-07-24 How much energy an AI query really uses, how training compares with serving, and how to read the numbers without believing either the hype or the panic. - [AI in Finance](https://howaiworks.ai/glossary/ai-finance.md) — updated 2026-07-24 Machine learning that approves loans and blocks card fraud — where you ship the most accurate model you can defend, not the best one you can build. - [AI for Good](https://howaiworks.ai/glossary/ai-for-good.md) — updated 2026-07-22 AI for Good is an aspiration, not a technique, and nothing certifies the label. Here it is tested against named projects with published, measured results. - [AI Governance (AIG)](https://howaiworks.ai/glossary/ai-governance.md) — updated 2026-07-21 Which AI rules apply to an organisation — by role, risk tier and jurisdiction — and what compliance requires: inventories, documentation, release gates. - [AI Healthcare](https://howaiworks.ai/glossary/ai-healthcare.md) — updated 2026-07-24 Where AI is actually deployed in medicine — imaging, triage, ambient notes — and why 'beats doctors' headlines rarely survive contact with a clinic. - [AI Infrastructure](https://howaiworks.ai/glossary/ai-infrastructure.md) — updated 2026-07-24 Why AI clusters are designed around memory bandwidth and interconnect rather than FLOPs, and what that changes when you size a training run or a serving fleet. - [AI Research](https://howaiworks.ai/glossary/ai-research.md) — updated 2026-07-21 AI research is the systematic investigation into the development of algorithms, models, and systems that exhibit intelligent behavior. - [AI Safety](https://howaiworks.ai/glossary/ai-safety.md) — updated 2026-07-24 AI safety is three fields, not one: present-day reliability, alignment, and catastrophic risk. What each actually claims, and why they get confused. - [AI in Science](https://howaiworks.ai/glossary/ai-science.md) — updated 2026-07-21 Three different things get called AI for science. Two have produced verified results, the third mostly has not — and what got revised down. - [Ambient Clinical Documentation](https://howaiworks.ai/glossary/ambient-clinical-documentation.md) — updated 2026-07-21 AI scribes draft the clinical note from consultation audio. What the largest deployment and the randomised trials measured — and who stays liable. - [Anomaly Detection (AD)](https://howaiworks.ai/glossary/anomaly-detection.md) — updated 2026-07-21 A 99%-accurate detector can still hand you a queue that is 99% false alarms. How anomaly detection scores, thresholds and ranks rare events. - [API](https://howaiworks.ai/glossary/api.md) — updated 2026-07-24 An API (Application Programming Interface) is a set of rules and protocols that allows different software applications to communicate and share data seamlessly. - [Application-Specific Integrated Circuit (ASIC)](https://howaiworks.ai/glossary/application-specific-integrated-circuit.md) — updated 2026-07-21 A chip wired for one job. Why Google builds TPUs and bitcoin abandoned GPUs — the volume-and-time bet behind custom silicon, with the break-even maths. - [Artificial General Intelligence (AGI)](https://howaiworks.ai/glossary/artificial-general-intelligence.md) — updated 2026-07-24 AGI is AI matching human ability across most cognitive work — and there is no agreed test for it, so the definition you pick decides the answer. - [Artificial Intelligence (AI)](https://howaiworks.ai/glossary/artificial-intelligence.md) — updated 2026-09-06 AI is software whose behaviour is fitted from data instead of written rule by rule: what that means, why the label keeps moving, and where you already use it. - [Artificial Superintelligence (ASI)](https://howaiworks.ai/glossary/artificial-superintelligence.md) — updated 2026-07-21 Artificial superintelligence: what it would mean, why the intelligence explosion argument is contested, and what labs and regulators actually do about it. - [Attention Mechanism](https://howaiworks.ai/glossary/attention-mechanism.md) — updated 2026-07-21 Each token emits a query, key and value; dot-product scores scaled by √d become softmax weights over every other token. The cost grows as n². - [Audio Processing (AP)](https://howaiworks.ai/glossary/audio-processing.md) — updated 2026-07-21 How computers turn sound into numbers: sampling, the Nyquist limit, spectrograms and mel filterbanks — the front end of every speech and audio model. - [Autoencoder](https://howaiworks.ai/glossary/autoencoder.md) — updated 2026-07-24 A neural network trained to reproduce its own input through a bottleneck too narrow to carry it — so the compressed code, not the copy, is the product. - [Autonomous Systems](https://howaiworks.ai/glossary/autonomous-systems.md) — updated 2026-07-22 Automation runs a procedure someone wrote down; autonomy picks the procedure at run time. Sense-plan-act loops, operational design domains and autonomy levels. - [Autonomous Vehicle Safety](https://howaiworks.ai/glossary/autonomous-vehicle-safety.md) — updated 2026-07-24 Autonomous vehicle safety technologies, standards, and testing protocols designed to ensure self-driving cars can operate safely without human intervention. - [Backpropagation](https://howaiworks.ai/glossary/backpropagation.md) — updated 2026-07-21 Backpropagation computes the gradient of the loss for every parameter at once, for about the cost of one extra forward pass. How the chain rule does it. - [Benchmark](https://howaiworks.ai/glossary/benchmark.md) — updated 2026-07-24 An AI benchmark is a standardized test or dataset used to evaluate and compare the performance of different AI models across tasks like reasoning and coding. - [Bias in AI (Algorithmic Bias)](https://howaiworks.ai/glossary/bias.md) — updated 2026-07-24 Why AI systems produce unfair outcomes: how bias enters through data and proxy labels, and why no model can satisfy every definition of fairness at once. - [Bias-Variance Tradeoff](https://howaiworks.ai/glossary/bias-variance-tradeoff.md) — updated 2026-07-24 Statistical bias, not fairness. Error splits into bias squared, variance and irreducible noise — derived, worked on numbers, and honest about where it breaks. - [Calibration](https://howaiworks.ai/glossary/calibration.md) — updated 2026-07-25 Calibration means a model's confidence matches reality: of everything it calls 80% likely, about 80% should happen — a property separate from raw accuracy. - [Catastrophic Forgetting](https://howaiworks.ai/glossary/catastrophic-forgetting.md) — updated 2026-07-21 Why fine-tuning makes a model worse at everything it used to do well — the mechanism inside shared weights, and what each fix actually costs. - [Causal Reasoning](https://howaiworks.ai/glossary/causal-reasoning.md) — updated 2026-07-22 Why a model trained on observational data cannot tell cause from correlation, what that costs when you act on it, and what actually identifies one. - [Chain-of-Thought (CoT)](https://howaiworks.ai/glossary/chain-of-thought.md) — updated 2026-09-06 A model writing out its reasoning before answering. Once a prompting trick, now trained in by RL, billed as output tokens, and frequently unfaithful. - [Chunking](https://howaiworks.ai/glossary/chunking.md) — updated 2026-07-24 Chunking splits documents into smaller passages before embedding them for retrieval. Chunk size, overlap, and splitting strategy quietly decide RAG quality. - [Class Imbalance](https://howaiworks.ai/glossary/class-imbalance.md) — updated 2026-07-25 Why a fraud model can score 99% accuracy and catch nothing — the imbalance trap, why precision/recall/PR-AUC replace accuracy, and how to fix it. - [Classification (CLF)](https://howaiworks.ai/glossary/classification.md) — updated 2026-07-24 Accuracy is usually the wrong metric. What a confusion matrix, precision, recall and the decision threshold really tell you about a classifier. - [Classifier-Free Guidance (CFG)](https://howaiworks.ai/glossary/classifier-free-guidance.md) — updated 2026-07-25 Classifier-free guidance (CFG) is how a diffusion model follows a prompt: each step runs the model with and without the prompt and extrapolates the gap. - [Cloud Computing](https://howaiworks.ai/glossary/cloud-computing.md) — updated 2026-07-22 Cloud computing is the on-demand delivery of computing power, database storage, applications, and other IT resources via the internet with pay-as-you-go. - [Clustering](https://howaiworks.ai/glossary/clustering.md) — updated 2026-07-21 Grouping unlabeled data into clusters: which algorithm to pick, why each one imposes a shape, how to choose k, and how to tell if the clusters are real. - [Computer Use (GUI Agents)](https://howaiworks.ai/glossary/computer-use.md) — updated 2026-07-24 How AI operates software by reading screenshots and issuing clicks and keystrokes: the perception-action loop, why grounding is hard, and what OSWorld measures. - [Computer Vision (CV)](https://howaiworks.ai/glossary/computer-vision.md) — updated 2026-07-21 How computers turn a grid of pixel values into labels, boxes and masks — the task ladder, the 2012 hinge, and where vision is reliable versus brittle. - [Concept Drift](https://howaiworks.ai/glossary/concept-drift.md) — updated 2026-07-25 Concept drift is when the relationship a deployed model learned changes over time, so accuracy silently decays until you detect the drift and retrain. - [Concurrency](https://howaiworks.ai/glossary/concurrency.md) — updated 2026-07-22 Concurrency is structuring a program so independent tasks interleave. How it differs from parallelism, and why serving a model is a queueing problem. - [Conformal Prediction](https://howaiworks.ai/glossary/conformal-prediction.md) — updated 2026-07-25 A model-agnostic, distribution-free method that turns any prediction into a set guaranteed to contain the true answer at a chosen rate, such as 90%. - [Consciousness](https://howaiworks.ai/glossary/consciousness.md) — updated 2026-07-21 Could an AI have subjective experience, and how would we tell? The hard problem, the four leading theories, and why a model's self-report proves nothing. - [Consensus Algorithm](https://howaiworks.ai/glossary/consensus-algorithm.md) — updated 2026-07-22 A consensus algorithm makes separate machines agree on one value even when some fail. Why majorities always overlap, and why agreement costs a round trip. - [Content Provenance (C2PA)](https://howaiworks.ai/glossary/content-provenance.md) — updated 2026-07-24 A cryptographically signed record of where a file came from — proving what device or model made it, instead of guessing from pixels whether it is AI-generated. - [Context Engineering](https://howaiworks.ai/glossary/context-engineering.md) — updated 2026-07-24 Deciding what enters a model's context window each turn and what gets evicted — the discipline that took over from prompt engineering once agents arrived. - [Context Window](https://howaiworks.ai/glossary/context-window.md) — updated 2026-09-06 Everything a model can see at once — why the advertised million tokens is not the usable number, and why you pay for the whole window on every turn. - [Continuous Learning (CL)](https://howaiworks.ai/glossary/continuous-learning.md) — updated 2026-07-24 Keeping a deployed model current as the world moves — why teams schedule batch retraining instead of true online learning, and what breaks if they don't. - [Conversational AI](https://howaiworks.ai/glossary/conversational-ai.md) — updated 2026-07-24 Conversational AI holds a multi-turn dialogue in natural language. Unlike an intent-based chatbot, it generates each reply rather than picking one. - [Convolution](https://howaiworks.ai/glossary/convolution.md) — updated 2026-07-22 Convolution slides a small kernel over an image, multiplying and summing each patch into one output number. The arithmetic of stride, padding and kernels. - [Convolutional Neural Network (CNN)](https://howaiworks.ai/glossary/convolutional-neural-network.md) — updated 2026-07-22 The architecture that assumes image features are local and position-independent — how the stack builds a hierarchy, and where CNNs still beat transformers. - [Cross-Validation (CV)](https://howaiworks.ai/glossary/cross-validation.md) — updated 2026-07-24 How cross-validation turns one lucky train/test split into a stable estimate, why 5 or 10 folds, and the leakage that quietly inflates every score. - [CUDA](https://howaiworks.ai/glossary/cuda.md) — updated 2026-07-24 NVIDIA's software layer for GPUs — the programming model that turned graphics cards into AI hardware, and the real reason rivals with comparable silicon lose. - [Data Analysis](https://howaiworks.ai/glossary/data-analysis.md) — updated 2026-07-22 What the work consists of: framing a question, looking before aggregating, choosing a summary that does not lie, and turning a number into a decision. - [Data Augmentation](https://howaiworks.ai/glossary/data-augmentation.md) — updated 2026-07-22 Making new training examples by transforming existing ones. Valid only when the label is indifferent to the transform — and it adds variety, not information. - [Data Leakage](https://howaiworks.ai/glossary/data-leakage.md) — updated 2026-07-24 Data leakage is training a model on information it will not have at prediction time. It pushes scores up, not down, which is why review never catches it. - [Data Poisoning](https://howaiworks.ai/glossary/data-poisoning.md) — updated 2026-07-21 Data poisoning is a cyberattack where malicious actors insert corrupted or misleading data into an AI model's training set to manipulate its future. - [Data Processing](https://howaiworks.ai/glossary/data-processing.md) — updated 2026-07-22 What happens to raw data before a model can use it: ingestion, validation, transformation, ETL vs ELT, batch vs streaming, and what silently breaks. - [Decision Trees (DT)](https://howaiworks.ai/glossary/decision-trees.md) — updated 2026-07-22 How a decision tree actually chooses each split: Gini impurity worked by hand, why depth drives overfitting, and what a single tree cannot represent. - [Deep Learning](https://howaiworks.ai/glossary/deep-learning.md) — updated 2026-07-22 What makes deep learning deep: each layer builds on the one below, so capacity grows multiplicatively with depth and only additively with width. - [Deepfake](https://howaiworks.ai/glossary/deepfake.md) — updated 2026-07-24 AI-synthesized video, image or audio that makes a real person appear to say or do something they never did — how deepfakes are made, and how they are caught. - [Diffusion Language Models (DLMs)](https://howaiworks.ai/glossary/diffusion-language-models.md) — updated 2026-07-24 Text models that unmask many tokens per pass instead of writing one at a time — what that actually buys in speed, and what it costs in quality. - [Diffusion Model](https://howaiworks.ai/glossary/diffusion-model.md) — updated 2026-07-24 A generative model that learns to undo a fixed noising process — what the network really predicts, why the step count is the cost, and why it displaced GANs. - [Dimensionality Reduction (DR)](https://howaiworks.ai/glossary/dimensionality-reduction.md) — updated 2026-07-22 Why high-dimensional data has to be compressed, how PCA picks its axes, and why cluster sizes and gaps in a t-SNE or UMAP plot mean nothing. - [Direct Preference Optimization (DPO)](https://howaiworks.ai/glossary/direct-preference-optimization.md) — updated 2026-07-21 DPO trains a model on human preferences without a reward model or reinforcement learning: the derivation, the memory it saves, and where it loses to PPO. - [Disaggregated Serving (Prefill/Decode)](https://howaiworks.ai/glossary/disaggregated-serving.md) — updated 2026-07-25 Disaggregated serving runs LLM prefill and decode on separate GPU pools because the two phases hit opposite limits: compute versus memory bandwidth. - [Distributed Computing](https://howaiworks.ai/glossary/distributed-computing.md) — updated 2026-07-22 Distributed computing splits work across separate machines that talk only by messages. Why partial failure, not speed, is the defining problem. - [Distributed Training](https://howaiworks.ai/glossary/distributed-training.md) — updated 2026-07-24 How one model is trained across thousands of GPUs — data, tensor and pipeline parallelism, and why the network rather than the chip becomes the bottleneck. - [Early Stopping](https://howaiworks.ai/glossary/early-stopping.md) — updated 2026-07-25 Early stopping halts training once validation loss stops improving, so the model generalises instead of memorising. How patience and best-weight restore work. - [Edge AI](https://howaiworks.ai/glossary/edge-ai.md) — updated 2026-07-22 Edge AI runs the model on the device that captured the data instead of in the cloud — what that buys in latency, bandwidth and privacy, and what it costs. - [Educational AI](https://howaiworks.ai/glossary/educational-ai.md) — updated 2026-07-22 AI that tutors, adapts practice and grades work — and what forty years of effect-size research actually shows about whether it improves learning. - [Embedding](https://howaiworks.ai/glossary/embedding.md) — updated 2026-07-22 What an embedding vector actually is, why cosine similarity beats Euclidean distance, what dimensions buy you, and where the geometry misleads you. - [Embodied AI](https://howaiworks.ai/glossary/embodied-ai.md) — updated 2026-07-22 AI that learns by acting on the physical world: why robot data cannot be scraped, why a 76 ms model cannot run a 20 ms loop, and what sim-to-real costs. - [Ensemble Methods](https://howaiworks.ai/glossary/ensemble-methods.md) — updated 2026-07-22 Why averaging models reduces error and where it stops: the variance of a mean of m correlated predictors, the correlation floor, and what bagging cannot fix. - [Error Handling in AI Systems](https://howaiworks.ai/glossary/error-handling.md) — updated 2026-07-24 Ordinary code fails by raising an exception. A model fails by returning a confident, well-formed, wrong answer that nothing catches. What to do about it. - [Ethics in AI](https://howaiworks.ai/glossary/ethics-in-ai.md) — updated 2026-07-23 AI ethics is what happens when good principles provably conflict: fairness against fairness, privacy against accuracy, and whose values get to decide. - [EUV Lithography](https://howaiworks.ai/glossary/euv-lithography.md) — updated 2026-07-18 The machines that print advanced chips with 13.5 nm light — how they work, why only one company makes them, and what that chokepoint means. - [Explainable AI (XAI)](https://howaiworks.ai/glossary/explainable-ai.md) — updated 2026-07-22 What being explainable actually requires: a model you can read versus a post-hoc guess about one, and the evidence that the guess is often wrong. - [Feature Scaling](https://howaiworks.ai/glossary/feature-scaling.md) — updated 2026-07-25 Feature scaling puts numeric features on comparable ranges so no variable dominates by its units alone — why distance- and gradient-based models need it. - [Feature Selection (FS)](https://howaiworks.ai/glossary/feature-selection.md) — updated 2026-07-22 Which columns to keep and how to decide: filter vs wrapper vs embedded, why 2^n subsets rules out brute force, and the CV mistake that fakes accuracy. - [Federated Learning](https://howaiworks.ai/glossary/federated-learning.md) — updated 2026-07-24 Federated learning trains a shared model across many devices without moving their raw data: the local-train, send-updates, aggregate loop. - [Few-shot Learning (FSL)](https://howaiworks.ai/glossary/few-shot-learning.md) — updated 2026-07-22 Teaching a model a task by showing it a handful of worked examples. In its modern form, few-shot prompting, no weights are updated at all. - [Fine-tuning (FT)](https://howaiworks.ai/glossary/fine-tuning.md) — updated 2026-07-22 What fine-tuning changes inside a model, what it costs in GPU memory, and why it reliably buys format and style but not new facts. - [FlashAttention](https://howaiworks.ai/glossary/flashattention.md) — updated 2026-07-25 The exact-attention algorithm that made long context affordable: it never writes the n x n score matrix to GPU memory, so it runs faster and changes no output. - [FLOPs (Floating Point Operations)](https://howaiworks.ai/glossary/flops.md) — updated 2026-07-24 The unit AI compute is measured in — how a training run's FLOP budget is counted, why FLOP and FLOPS are not the same, and why regulators use the number. - [Foundation Models](https://howaiworks.ai/glossary/foundation-models.md) — updated 2026-07-24 Not a marketing word for a big AI model: a foundation model is trained once on broad data and reused as the base for many downstream tasks. - [Function Calling (Tool Calling)](https://howaiworks.ai/glossary/function-calling.md) — updated 2026-07-24 The model never runs your function. It emits a structured request; your code decides whether to honour it, executes it, and hands the result back. - [General Problem Solver (GPS)](https://howaiworks.ai/glossary/general-problem-solving.md) — updated 2026-07-22 The 1957 Newell-Shaw-Simon program that solved problems by means-ends analysis, and the b^d arithmetic that stopped it from scaling past puzzles. - [Generalization](https://howaiworks.ai/glossary/generalization.md) — updated 2026-07-22 Why a model that scores 99% in training fails in the world: the generalization gap, the test-set size needed to measure it, and why classical theory is wrong. - [Generative Adversarial Network (GAN)](https://howaiworks.ai/glossary/generative-adversarial-network.md) — updated 2026-07-24 Two networks trained against each other: one invents data, the other judges it. The generator never sees a real example — only the judge's verdict. - [Generative AI](https://howaiworks.ai/glossary/generative-ai.md) — updated 2026-07-24 Generative AI learns the probability distribution its training data came from, well enough to draw new samples from it — not just sort inputs into classes. - [Generative Engine Optimization (GEO)](https://howaiworks.ai/glossary/generative-engine-optimization.md) — updated 2026-07-22 GEO — also called AEO — is optimizing content so AI answer engines cite it. From a 2024 KDD paper that found keyword stuffing backfires. - [GPU Computing](https://howaiworks.ai/glossary/gpu-computing.md) — updated 2026-07-22 Why a graphics chip runs AI: a GPU spends its transistors on arithmetic instead of on making one thread fast — and what that trade wins and costs. - [Gradient Boosting](https://howaiworks.ai/glossary/gradient-boosting.md) — updated 2026-07-24 Each tree is fitted to the negative gradient of the loss, not to the residuals — a boosting round worked by hand, and what changes under log loss. - [Gradient Descent](https://howaiworks.ai/glossary/gradient-descent.md) — updated 2026-07-22 The rule that turns a gradient into a weight update. Worked on f(x)=x²: why the step size alone decides between converging, oscillating and blowing up. - [Graph Neural Networks (GNN)](https://howaiworks.ai/glossary/graph-neural-networks.md) — updated 2026-07-22 A GNN takes adjacency as data, not as an architectural assumption. Message passing worked by hand, why k layers reach d^k nodes, and why deep GNNs collapse. - [GraphRAG](https://howaiworks.ai/glossary/graphrag.md) — updated 2026-07-25 GraphRAG runs RAG over a knowledge graph and community summaries built from a corpus, answering global questions that plain vector RAG cannot. - [Group Relative Policy Optimization (GRPO)](https://howaiworks.ai/glossary/group-relative-policy-optimization.md) — updated 2026-07-24 GRPO is PPO without the critic: it scores a group of sampled answers against each other. How it works, what it saves, and where it biases training. - [Grouped-Query Attention (GQA)](https://howaiworks.ai/glossary/grouped-query-attention.md) — updated 2026-07-25 Grouped-query attention lets query heads share key/value heads, shrinking the KV cache that dominates long-context inference at near multi-head quality. - [Guardrails](https://howaiworks.ai/glossary/guardrails.md) — updated 2026-07-24 A guardrail is a deterministic check around a model that can block, rewrite, or force a retry — external code you can test, unlike prompting a model to behave. - [AI Hallucinations](https://howaiworks.ai/glossary/hallucinations.md) — updated 2026-07-24 Models hallucinate because training and benchmarks reward guessing over admitting ignorance: a guess sometimes scores, "I don't know" scores zero every time. - [High Bandwidth Memory (HBM)](https://howaiworks.ai/glossary/high-bandwidth-memory.md) — updated 2026-07-24 Stacked DRAM bonded beside the GPU die — the scarce component that gates AI accelerator supply, and why bandwidth rather than capacity is the constraint. - [High Bias](https://howaiworks.ai/glossary/high-bias.md) — updated 2026-07-22 Statistical bias, not AI fairness: the error a model class cannot escape even with infinite data — worked in closed form on a line fitted to a parabola. - [Human-AI Collaboration (HAC)](https://howaiworks.ai/glossary/human-ai-collaboration.md) — updated 2026-07-24 Splitting one task between a person and a model. A meta-analysis of 106 experiments found the pair usually loses to whichever of the two is better alone. - [Hybrid Search (RRF)](https://howaiworks.ai/glossary/hybrid-search.md) — updated 2026-07-25 Hybrid search runs keyword (BM25) and vector retrieval together, then fuses their ranked lists with Reciprocal Rank Fusion (RRF), a score-agnostic method. - [Hyperparameter](https://howaiworks.ai/glossary/hyperparameter.md) — updated 2026-07-24 A setting you choose before training that gradient descent never updates — learning rate, batch size, layer count — and why a bad one caps the model. - [Image Generation](https://howaiworks.ai/glossary/image-generation.md) — updated 2026-07-24 How AI turns noise into a picture: the diffusion loop, the noise schedule on numbers, why it runs in a 48x-compressed latent space, and what guidance costs. - [Inference](https://howaiworks.ai/glossary/inference.md) — updated 2026-07-21 Running a trained model to get an answer. Training is paid once; inference is paid on every request — and it is where most AI money goes. - [Inference Optimization](https://howaiworks.ai/glossary/inference-optimization.md) — updated 2026-07-13 How a served LLM is made cheap: continuous batching, the prefill/decode split, and the throughput-versus-latency tradeoff behind every API price. - [Information Gain](https://howaiworks.ai/glossary/information-gain.md) — updated 2026-07-22 Entropy worked from scratch, one split computed by hand in bits, and why a unique ID column scores the maximum possible information gain. - [Information Retrieval (IR)](https://howaiworks.ai/glossary/information-retrieval.md) — updated 2026-07-24 Information retrieval ranks documents against a query. How inverted indexes, TF-IDF and BM25 score, and how precision, recall, MAP and nDCG measure it. - [Interconnect (NVLink, InfiniBand)](https://howaiworks.ai/glossary/interconnect.md) — updated 2026-07-25 The wiring that moves data between AI chips: on-package memory, NVLink inside a server, InfiniBand between them — and why each tier is far slower than the last. - [Jailbreak (Red-Teaming)](https://howaiworks.ai/glossary/jailbreak.md) — updated 2026-07-24 A jailbreak is an input that steers an AI past its safety training to produce content it was built to refuse. How they work, and how they are defended. - [k-Nearest Neighbors (kNN)](https://howaiworks.ai/glossary/k-nearest-neighbors.md) — updated 2026-07-25 How k-Nearest Neighbors classifies a point by majority vote of its k closest neighbours: choosing k, the distance metric, and the curse of dimensionality. - [Knowledge Distillation](https://howaiworks.ai/glossary/knowledge-distillation.md) — updated 2026-07-22 How a small model learns what a big one knows: the teacher's full probability distribution carries structure the hard labels do not — Hinton's dark knowledge. - [Knowledge Graphs (KG)](https://howaiworks.ai/glossary/knowledge-graphs.md) — updated 2026-07-22 A knowledge graph stores facts as typed links between named things so one query can join across several. The structure is easy; the curation is what costs. - [Knowledge Representation](https://howaiworks.ai/glossary/knowledge-representation.md) — updated 2026-07-22 How you encode what a system knows fixes which conclusions are cheap, which are exponential and which are impossible. The expressiveness-tractability trade-off. - [KV Cache](https://howaiworks.ai/glossary/kv-cache.md) — updated 2026-07-24 The memory an LLM keeps so it never recomputes the past — why a long context costs money, and how many users one GPU can actually serve at once. - [Large Language Model (LLM)](https://howaiworks.ai/glossary/large-language-model.md) — updated 2026-09-06 A large language model predicts the next token of text. How that produces answers, where its knowledge is stored, and why an LLM is not a chatbot. - [Layers in Neural Networks](https://howaiworks.ai/glossary/layers.md) — updated 2026-07-22 A layer is one weight matrix plus a nonlinearity. How it differs from a neuron and a parameter, why stacking helps, and how many layers real models have. - [LLM-as-a-Judge](https://howaiworks.ai/glossary/llm-as-a-judge.md) — updated 2026-07-24 Using one model to grade another's output against a rubric: how it works, the biases that break it, and the human agreement rate you can actually reach. - [llms.txt](https://howaiworks.ai/glossary/llms-txt.md) — updated 2026-08-02 A proposed Markdown file at /llms.txt that gives AI systems a clean map of a site. Google ignores it; coding agents and doc tools use it heavily. - [Logistic Regression (LR)](https://howaiworks.ai/glossary/logistic-regression.md) — updated 2026-07-22 Logistic regression turns weighted evidence into a probability with the sigmoid: why the name says regression, how to read coefficients, and when to use it. - [Low-Rank Adaptation (LoRA)](https://howaiworks.ai/glossary/lora.md) — updated 2026-07-24 LoRA freezes a model's weights and trains two thin matrices beside each one: 98,304 numbers instead of 151 million, and no added inference latency. - [Loss Function](https://howaiworks.ai/glossary/loss-function.md) — updated 2026-07-22 The number a model is trained to make smaller. Its slope, not its value, moves every weight — so the loss you pick decides what the model learns. - [Low Variance](https://howaiworks.ai/glossary/low-variance.md) — updated 2026-07-22 Low variance means values sit close to their average — and for a model, that retraining it on fresh data barely moves its predictions. Worked on numbers. - [Machine Learning (ML)](https://howaiworks.ai/glossary/machine-learning.md) — updated 2026-07-22 How a machine finds a rule nobody wrote: a line fitted to four points by hand, loss falling from 433.5 to 6.38 in three steps, and why 98% accuracy lies. - [Machine Unlearning](https://howaiworks.ai/glossary/machine-unlearning.md) — updated 2026-07-25 Removing the influence of specific training data from an already-trained model without full retraining, driven by erasure law, copyright, and safety. - [Matrix Multiplication (GEMM)](https://howaiworks.ai/glossary/matrix-multiplication.md) — updated 2026-07-13 The one operation that consumes most of an AI model's compute — and the reason AI chips are built the way they are, from tensor cores to systolic arrays. - [Membership Inference](https://howaiworks.ai/glossary/membership-inference.md) — updated 2026-07-25 A privacy attack that decides whether a specific record was in a model's training set, exploiting that models are more confident on data they memorized. - [Memory Wall](https://howaiworks.ai/glossary/memory-wall.md) — updated 2026-07-24 Why LLM inference waits on memory rather than maths — arithmetic intensity, the roofline, and the one fact that explains HBM scarcity and the price of a token. - [Meta-Learning](https://howaiworks.ai/glossary/meta-learning.md) — updated 2026-07-22 Learning to learn: training a model across thousands of tiny tasks so it adapts from a handful of examples — and how that differs from fine-tuning. - [Mixture-of-Experts (MoE)](https://howaiworks.ai/glossary/mixture-of-experts.md) — updated 2026-07-22 The architecture behind "671B total, 37B active": hundreds of expert sub-networks, a router that picks a few per token, and a memory bill that never shrinks. - [Machine Learning Operations (MLOps)](https://howaiworks.ai/glossary/mlops.md) — updated 2026-07-22 Why shipping a model is not like shipping software: the artifact has three inputs — code, data and trained weights — and only one of them lives in git. - [Model Card](https://howaiworks.ai/glossary/model-card.md) — updated 2026-07-25 A model card is a short document shipped with a released AI model stating its intended use, training data, and evaluation broken down by group. - [Model Compression](https://howaiworks.ai/glossary/model-compression.md) — updated 2026-07-21 Making a trained model smaller and cheaper to run — quantization, pruning, distillation and low-rank methods, and which one your problem actually needs. - [Model Context Protocol (MCP)](https://howaiworks.ai/glossary/model-context-protocol.md) — updated 2026-07-24 An open standard that connects any AI app to any tool over one JSON-RPC interface, turning M clients x N tools from M x N custom connectors into M + N. - [Model Deployment](https://howaiworks.ai/glossary/model-deployment.md) — updated 2026-07-22 What has to happen between a model that trains well and users calling it: replica sizing, latency budgets, cold starts, and the rollback nobody built. - [Model Routing](https://howaiworks.ai/glossary/model-routing.md) — updated 2026-07-25 Model routing sends each request to the cheapest capable model — a small model handles the easy majority and hard queries escalate to a frontier model. - [Model Size](https://howaiworks.ai/glossary/model-size.md) — updated 2026-07-24 Model size is a model's parameter count. What that one number predicts — memory, cost per token, capability — worked on real models from 110M to trillions. - [Monitoring](https://howaiworks.ai/glossary/monitoring.md) — updated 2026-07-22 Why a deployed model rots while every dashboard stays green — and what you can actually measure when the true labels arrive weeks late, or never. - [Multi-Agent Systems (MAS)](https://howaiworks.ai/glossary/multi-agent-systems.md) — updated 2026-07-24 A multi-agent system coordinates two or more AI agents, each with its own role and tools, to solve tasks a single agent would handle worse or not at all. - [Multimodal AI](https://howaiworks.ai/glossary/multimodal-ai.md) — updated 2026-07-22 How one model reads an image and text at once: the picture is cut into fixed patches, each patch becomes a token, and those tokens cost context. - [Naive Bayes](https://howaiworks.ai/glossary/naive-bayes.md) — updated 2026-07-25 A probabilistic classifier applying Bayes' theorem under a deliberately false feature-independence assumption — fast, data-light, a strong text baseline. - [Natural Language Processing (NLP)](https://howaiworks.ai/glossary/natural-language-processing.md) — updated 2026-07-22 NLP is the field; large language models are now the answer to most of it. What sixty years of work was for, and which parts are still unsolved. - [Neural Network](https://howaiworks.ai/glossary/neural-network.md) — updated 2026-07-22 Why a stack of layers does what one cannot: two linear layers collapse into a single matrix, and a hidden layer of two units solves XOR. - [Neural Processing Unit (NPU)](https://howaiworks.ai/glossary/neural-processing-unit.md) — updated 2026-07-24 An NPU is a fixed-function matrix engine inside a phone or laptop chip that runs AI models at a few watts — and why TOPS is the wrong number to judge it by. - [Neurons](https://howaiworks.ai/glossary/neurons.md) — updated 2026-07-22 One neuron multiplies its inputs by weights, adds a bias and applies a function. The arithmetic worked by hand, and where the brain analogy breaks. - [No-Code Tools](https://howaiworks.ai/glossary/no-code-tools.md) — updated 2026-07-22 Visual builders over a fixed execution model. What they cost per run at real volume, where the envelope ends, and why leaving one is a rewrite. - [Normalization (LayerNorm, RMSNorm, BatchNorm)](https://howaiworks.ai/glossary/normalization.md) — updated 2026-07-25 Normalization rescales a layer's activations to a stable mean and scale so deep networks train faster. How BatchNorm, LayerNorm and RMSNorm differ. - [NVIDIA GPU for AI](https://howaiworks.ai/glossary/nvidia-gpu-ai.md) — updated 2026-07-24 Why AI training runs on NVIDIA rather than cheaper silicon — tensor cores, NVLink, and the software that decides how much of the datasheet you actually get. - [One-shot Learning](https://howaiworks.ai/glossary/one-shot-learning.md) — updated 2026-07-24 Learning a new class from exactly one example — and the split that matters: one-shot prompting updates no weights; classical one-shot training does. - [Ontologies](https://howaiworks.ai/glossary/ontologies.md) — updated 2026-07-23 An ontology is a formal, machine-readable model of a domain — classes, properties and logical axioms a computer can reason over, not just a taxonomy of labels. - [Optimization](https://howaiworks.ai/glossary/optimization.md) — updated 2026-07-24 In machine learning, optimization means minimizing a loss function to fit a model's parameters — the iterative, gradient-driven mechanism behind training. - [Overfitting](https://howaiworks.ai/glossary/overfitting.md) — updated 2026-07-24 Overfitting is when a model memorizes its training data, including the noise, so it scores high on training data but fails on new, unseen data. - [Parallel Processing](https://howaiworks.ai/glossary/parallel-processing.md) — updated 2026-07-24 Running many computations at the same instant across many processing units. In AI it makes matrix multiplication fast, which is why GPUs train models. - [Parameters](https://howaiworks.ai/glossary/parameters.md) — updated 2026-07-24 The learned numbers inside a model — what '70 billion parameters' actually means, and how the count converts directly into the memory a GPU must hold. - [Pattern Recognition (PR)](https://howaiworks.ai/glossary/pattern-recognition.md) — updated 2026-07-23 Pattern recognition is the automatic discovery and classification of regularities in data — the older name for what became statistical machine learning. - [Performance](https://howaiworks.ai/glossary/performance.md) — updated 2026-07-23 AI performance is not one number but a frontier of quality, speed, and cost—improving one axis (via quantization or batching) usually trades off another. - [Policy](https://howaiworks.ai/glossary/policy.md) — updated 2026-07-24 In reinforcement learning, a policy is the function that maps a state to an action — it is the agent's behavior, the thing training optimizes. - [Pooling](https://howaiworks.ai/glossary/pooling.md) — updated 2026-07-24 A downsampling operation in CNNs that shrinks a feature map by summarizing each small region into one value, cutting resolution and compute. - [Positional Encoding (RoPE)](https://howaiworks.ai/glossary/positional-encoding.md) — updated 2026-07-25 Positional encoding tells a transformer where each token sits in a sequence. Why self-attention needs it, and how rotary embeddings (RoPE) work. - [Pre-trained Models](https://howaiworks.ai/glossary/pre-trained-models.md) — updated 2026-07-24 A neural network someone else already trained on a huge general dataset, which you download and adapt to your task instead of training from scratch. - [Precision and Recall](https://howaiworks.ai/glossary/precision-and-recall.md) — updated 2026-07-24 Precision, recall, F1 and the confusion matrix worked through one imbalanced example — and why the model with the best accuracy can catch nothing at all. - [Precision Medicine](https://howaiworks.ai/glossary/precision-medicine.md) — updated 2026-07-23 How precision medicine tailors care to a person's genome and clinical data, and what AI does in it: variant interpretation, risk, and drug response. - [Principal Component Analysis (PCA)](https://howaiworks.ai/glossary/principal-component-analysis.md) — updated 2026-07-25 Principal Component Analysis (PCA) reduces dimensions by projecting data onto orthogonal directions of maximum variance, ranked by the variance each keeps. - [Privacy](https://howaiworks.ai/glossary/privacy.md) — updated 2026-07-23 In AI, privacy is about what a model can memorize and leak about people, and the techniques—differential privacy, federated learning—that limit it. - [Production Systems](https://howaiworks.ai/glossary/production-systems.md) — updated 2026-07-23 A model serving live users under latency and uptime guarantees — judged at the 99th percentile under real load, not by the median on a laptop. - [Prompt Caching](https://howaiworks.ai/glossary/prompt-caching.md) — updated 2026-07-24 Repeated prompt prefixes bill at a fraction of the input price. How prefix matching works, what silently breaks it, and where the break-even sits. - [Prompt Engineering](https://howaiworks.ai/glossary/prompt-engineering.md) — updated 2026-07-24 Writing the input to a fixed, already-trained language model so it reliably produces the output you want — steering the model, not retraining it. - [Prompt Injection](https://howaiworks.ai/glossary/prompt-injection.md) — updated 2026-07-24 Prompt injection is when text an AI was meant to read gets obeyed as an instruction. Direct vs indirect, why it is unfixable, and what actually contains it. - [Protein Folding](https://howaiworks.ai/glossary/protein-folding.md) — updated 2026-07-24 How a chain of amino acids collapses into the 3D shape that sets a protein's function - and how predicting that shape from sequence became AlphaFold's win. - [Quantization](https://howaiworks.ai/glossary/quantization.md) — updated 2026-07-24 Running a model in fewer bits — FP16, FP8, INT4 — to cut memory and cost, what accuracy it actually costs, and why a memory-bound model also gets faster. - [Quantum Computing (QC)](https://howaiworks.ai/glossary/quantum-computing.md) — updated 2026-07-24 Quantum computing uses qubits — held in superposition and entangled — for exponential speedups on a narrow set of problems, not general computing. - [Random Forest (RF)](https://howaiworks.ai/glossary/random-forest.md) — updated 2026-07-23 A random forest averages many decision trees, each grown on a bootstrap sample and split on a random subset of features, to cut variance without adding bias. - [Reasoning Model](https://howaiworks.ai/glossary/reasoning-model.md) — updated 2026-07-24 A model trained to think at length before answering. How it differs from a chat model, what the effort dial does, and the tasks where it earns its price. - [Recommendation Systems](https://howaiworks.ai/glossary/recommendation-systems.md) — updated 2026-07-23 A recommendation system predicts what a user will want — a film, product, or song — from patterns in past behavior, and surfaces it automatically. - [Recurrent Neural Network (RNN)](https://howaiworks.ai/glossary/recurrent-neural-network.md) — updated 2026-07-23 A recurrent neural network reads a sequence one element at a time, carrying a hidden state that summarizes everything seen so far to model order and context. - [Regression](https://howaiworks.ai/glossary/regression.md) — updated 2026-07-23 The supervised-learning task of predicting a continuous number — a price, a temperature, a duration — from input features, in contrast to classification. - [Regularization](https://howaiworks.ai/glossary/regularization.md) — updated 2026-07-24 Regularization deliberately constrains a model so it fits training data less perfectly, trading a little training accuracy for much better generalization. - [Reinforcement Learning (RL)](https://howaiworks.ai/glossary/reinforcement-learning.md) — updated 2026-07-23 How an agent learns to act by trial and error: it takes actions in an environment, earns rewards or penalties, and adjusts to maximize long-term reward. - [Reinforcement Learning from Human Feedback (RLHF)](https://howaiworks.ai/glossary/reinforcement-learning-from-human-feedback.md) — updated 2026-07-24 How RLHF works, how it compares with DPO and GRPO, and why it makes language models verbose, agreeable and confidently wrong. - [Reinforcement Learning from Verifiable Rewards (RLVR)](https://howaiworks.ai/glossary/reinforcement-learning-from-verifiable-rewards.md) — updated 2026-07-24 RLVR replaces the learned reward model with a program that checks the answer. What that buys, where it stops working, and how it differs from RLHF. - [Representation Learning](https://howaiworks.ai/glossary/representation-learning.md) — updated 2026-07-23 Representation learning is machine learning that discovers the useful features of raw data automatically, instead of humans hand-engineering them. - [Reranking (Cross-Encoder)](https://howaiworks.ai/glossary/reranking.md) — updated 2026-07-24 Reranking rescores the top-N results from a fast retriever with a slow, accurate cross-encoder that reads the query and document together. - [Residual Connections (Skip Connections)](https://howaiworks.ai/glossary/residual-connections.md) — updated 2026-07-25 Residual connections add a block's input back to its output — F(x) + x — the trick that lets gradients flow through deep nets and made transformers trainable. - [Retrieval-Augmented Generation (RAG)](https://howaiworks.ai/glossary/retrieval-augmented-generation.md) — updated 2026-07-23 A technique that retrieves relevant documents at query time and feeds them into a language model's prompt, so it answers from current, external facts. - [Reward Hacking](https://howaiworks.ai/glossary/reward-hacking.md) — updated 2026-07-25 When an AI maximises its specified reward but defeats the designer's intent — optimising the proxy, not the goal. Goodhart's law, made mechanical. - [Reward Model (RM)](https://howaiworks.ai/glossary/reward-model.md) — updated 2026-07-24 The scorer that stands in for a human during RLHF: how reward models are trained, why their absolute scores mean nothing, and how they break. - [Robotics](https://howaiworks.ai/glossary/robotics.md) — updated 2026-07-24 Robotics builds machines that sense, plan, and act physically — where AI perception and control meet Moravec's paradox and the sim-to-real gap. - [Robustness](https://howaiworks.ai/glossary/robustness.md) — updated 2026-07-23 A model's ability to keep performing when inputs are noisy, out of distribution, or crafted by an attacker — not just accurate on clean test data. - [Scalable AI](https://howaiworks.ai/glossary/scalable-ai.md) — updated 2026-07-23 Scaling an AI system means three different things — bigger models, faster training, more concurrent users — and each axis hits a different hard limit. - [Scaling Laws](https://howaiworks.ai/glossary/scaling-laws.md) — updated 2026-07-13 The empirical curves that make AI capex rational: model loss falls predictably with compute, data and parameters — and what they do not promise. - [Self-Attention](https://howaiworks.ai/glossary/self-attention.md) — updated 2026-07-23 The transformer operation that lets every token in a sequence weigh every other token by relevance, computed as softmax(QKᵀ/√d_k)·V. - [Self-Improving AI (SIAI)](https://howaiworks.ai/glossary/self-improving-ai.md) — updated 2026-07-23 Self-improving AI (SIAI) is a system that upgrades its own capabilities. Real today only in narrow forms like self-play; the recursive version is speculative. - [Self-supervised Learning (SSL)](https://howaiworks.ai/glossary/self-supervised-learning.md) — updated 2026-07-24 Training a model on unlabeled data by hiding part of each input and making the model predict it — the paradigm behind next-token prediction and modern LLMs. - [Semantic Search](https://howaiworks.ai/glossary/semantic-search.md) — updated 2026-07-23 Semantic search finds results by meaning, not exact keywords: it embeds the query and documents as vectors and returns the nearest ones. - [Semantic Understanding](https://howaiworks.ai/glossary/semantic-understanding.md) — updated 2026-07-24 Whether an AI grasps meaning well enough to act correctly on novel inputs, or is pattern-matching that breaks on rephrasing — and why that line is contested. - [Semiconductor Manufacturing](https://howaiworks.ai/glossary/semiconductor-manufacturing.md) — updated 2026-07-13 How an AI chip is actually made — wafers, process nodes, yield and the foundry model — and why supply cannot simply respond to a spike in demand. - [Simple Continual Pretraining (SCP)](https://howaiworks.ai/glossary/simple-continual-pretraining.md) — updated 2026-07-23 Method for turning a pretrained autoregressive LLM into a diffusion language model by continuing pretraining with a bidirectional attention mask. - [Small Language Model (SLM)](https://howaiworks.ai/glossary/small-language-model.md) — updated 2026-08-02 A language model small enough to run on the device that uses it — roughly under 5B parameters. Trained far past compute-optimal to buy cheap inference. - [Social AI](https://howaiworks.ai/glossary/social-ai.md) — updated 2026-07-24 Social AI is artificial intelligence built to perceive, interpret, and take part in human social interaction — reading emotion, tone, intent, and context. - [Speculative Decoding](https://howaiworks.ai/glossary/speculative-decoding.md) — updated 2026-07-13 A small model drafts tokens and a large one verifies them in parallel — how LLM inference gets several times faster with mathematically identical output. - [Speech-to-Speech](https://howaiworks.ai/glossary/speech-to-speech.md) — updated 2026-07-25 Speech-to-speech AI takes audio in and returns audio out — conversation or translation — via a cascade (ASR to text to TTS) or one end-to-end model. - [State of the Art Model (SOTA)](https://howaiworks.ai/glossary/state-of-the-art-model.md) — updated 2026-07-21 There is no single SOTA model. How to read a benchmark claim — and why a 3-point lead on GPQA Diamond is six questions, well inside the noise. - [State-Space Models (Mamba)](https://howaiworks.ai/glossary/state-space-models.md) — updated 2026-07-25 State-space models are sequence models that scale linearly with length, not quadratically like attention. Mamba's input-dependent state made them competitive. - [Structured Outputs (Constrained Decoding)](https://howaiworks.ai/glossary/structured-outputs.md) — updated 2026-07-24 Constrained decoding masks a language model's token distribution against a JSON schema or grammar at each step, so the output is parseable by construction. - [Supervised Fine-Tuning (SFT)](https://howaiworks.ai/glossary/supervised-fine-tuning.md) — updated 2026-07-24 Supervised fine-tuning trains a pretrained language model on curated prompt-response pairs with next-token loss, teaching it to follow instructions. - [Supervised Learning](https://howaiworks.ai/glossary/supervised-learning.md) — updated 2026-07-23 Training a model on labeled input-output pairs so it learns a mapping it can apply to new, unlabeled inputs. - [Support Vector Machines (SVM)](https://howaiworks.ai/glossary/support-vector-machines.md) — updated 2026-07-23 A classifier that separates two classes with the widest possible margin, using support vectors and the kernel trick to handle non-linear data. - [Sycophancy](https://howaiworks.ai/glossary/sycophancy.md) — updated 2026-07-25 Sycophancy is when an AI model tells you what you want to hear — agreeing, flattering, caving when challenged — because human-preference training rewards it. - [Symbolic AI](https://howaiworks.ai/glossary/symbolic-ai.md) — updated 2026-07-23 Symbolic AI (GOFAI) builds intelligence from human-written symbols and logic rules rather than learned data - the paradigm behind expert systems like MYCIN. - [Synthetic Data](https://howaiworks.ai/glossary/synthetic-data.md) — updated 2026-07-24 Data generated by a model instead of collected from the world: what synthetic data is, where it helps, and why recursive use triggers model collapse. - [Temperature](https://howaiworks.ai/glossary/temperature.md) — updated 2026-07-24 Temperature is the number that controls how random a language model's output is — it divides the logits before softmax, from near-deterministic to diverse. - [Tensor Operations](https://howaiworks.ai/glossary/tensor-operations.md) — updated 2026-07-23 The array manipulations — elementwise math, matrix multiplication, reshaping, broadcasting, reduction — that every neural network is built out of. - [Tensor Processing Unit (TPU)](https://howaiworks.ai/glossary/tensor-processing-unit.md) — updated 2026-07-24 A TPU is Google's custom ASIC for the matrix multiplications in neural networks: a systolic array that reuses each value instead of refetching it. - [Test-Time Compute](https://howaiworks.ai/glossary/test-time-compute.md) — updated 2026-07-24 Spending compute at inference instead of training — how reasoning models buy accuracy with thinking time, and why it moved chip demand into serving. - [Text Analysis](https://howaiworks.ai/glossary/text-analysis.md) — updated 2026-07-24 Turning unstructured text into structured data — the concrete NLP task catalog: classification, sentiment analysis, entity extraction and topic modeling. - [Text Generation](https://howaiworks.ai/glossary/text-generation.md) — updated 2026-07-23 Text generation is a language model producing text one token at a time: at each step it predicts a distribution over its vocabulary, picks a token, repeats. - [Text-to-Speech (TTS)](https://howaiworks.ai/glossary/text-to-speech.md) — updated 2026-07-24 Text-to-Speech (TTS) is an AI technology that converts written text into natural-sounding human speech. Modern TTS uses deep learning to capture emotion. - [Time Series](https://howaiworks.ai/glossary/time-series.md) — updated 2026-07-24 Data points indexed in time order, where the sequence matters and rows aren't independent — analyzed to forecast future values or flag anomalies. - [Token](https://howaiworks.ai/glossary/token.md) — updated 2026-09-06 The unit an AI model reads, writes and bills by — roughly 1.3 tokens per English word. Input, output and cached tokens are priced up to 10x apart. - [Tokenization](https://howaiworks.ai/glossary/tokenization.md) — updated 2026-07-21 How text becomes the integers a language model reads: BPE, WordPiece and Unigram, why " strawberry" is one token, and what that breaks. - [Training](https://howaiworks.ai/glossary/training.md) — updated 2026-07-24 Training fits a model's parameters to data: forward pass, measure the loss, backward pass, nudge every weight downhill, repeat until the loss stops falling. - [Transfer Learning (TL)](https://howaiworks.ai/glossary/transfer-learning.md) — updated 2026-07-23 Transfer learning reuses a model trained on one large task as the starting point for a related task, so you need far less labeled data and compute. - [Transformer](https://howaiworks.ai/glossary/transformer.md) — updated 2026-07-23 The 2017 neural-network architecture that processes a whole sequence in parallel using self-attention instead of recurrence — the basis of modern LLMs. - [Transparency](https://howaiworks.ai/glossary/transparency.md) — updated 2026-07-24 Transparency in AI is the degree to which a system's data, workings, and decisions are visible and disclosed to users, auditors, and regulators. - [Trust](https://howaiworks.ai/glossary/trust.md) — updated 2026-07-24 Trust in AI is a person's willingness to rely on a system's output; the aim is calibration — trusting it as much as its real reliability deserves. - [Underfitting](https://howaiworks.ai/glossary/underfitting.md) — updated 2026-07-24 Underfitting is when a model is too simple to capture the real pattern, so it does poorly on both training and new data — the mirror of overfitting. - [Unsupervised Learning](https://howaiworks.ai/glossary/unsupervised-learning.md) — updated 2026-07-24 Finding structure in unlabeled data — no correct answers are given, so the model discovers groupings, patterns, or a compressed representation on its own. - [Value Learning](https://howaiworks.ai/glossary/value-learning.md) — updated 2026-07-23 The AI-alignment problem of teaching a system to infer and act on human values, so a capable optimizer does what people want, not a misspecified proxy. - [Vector Database](https://howaiworks.ai/glossary/vector-database.md) — updated 2026-07-24 A vector database stores embeddings with metadata, builds an approximate-nearest-neighbour index over them, and serves filtered similarity queries at scale. - [Vector Search](https://howaiworks.ai/glossary/vector-search.md) — updated 2026-07-23 Vector search returns the stored vectors nearest to a query vector by a distance metric — the retrieval primitive under semantic search, RAG and recommenders. - [Vectorization](https://howaiworks.ai/glossary/vectorization.md) — updated 2026-07-23 Replacing element-by-element loops with operations on whole arrays at once, so the work runs as SIMD instructions on a CPU or in parallel on a GPU. - [Vibe Coding](https://howaiworks.ai/glossary/vibe-coding.md) — updated 2026-07-24 Vibe coding is building software by describing what you want to an AI and running the code it generates without reading or fully understanding it. - [Video Generation](https://howaiworks.ai/glossary/video-generation.md) — updated 2026-08-02 Video generation is an AI technology that creates moving images and scenes from text descriptions, images, or other video clips. - [Vision Transformer (ViT)](https://howaiworks.ai/glossary/vision-transformer.md) — updated 2026-07-25 A Vision Transformer (ViT) cuts an image into fixed patches, embeds each as a token, and runs a transformer encoder — trading CNN inductive bias for scale. - [Voice Cloning](https://howaiworks.ai/glossary/voice-cloning.md) — updated 2026-07-25 Voice cloning uses AI to copy a specific person's voice from seconds of reference audio. How zero-shot systems work, plus the fraud and consent risks. - [Voice Recognition](https://howaiworks.ai/glossary/voice-recognition.md) — updated 2026-07-23 How AI turns spoken words into text (speech recognition / ASR) — and how that differs from identifying who is speaking (speaker recognition). - [Watermarking (SynthID)](https://howaiworks.ai/glossary/watermarking.md) — updated 2026-07-25 AI watermarking hides a detectable signal inside model-generated text and images. How SynthID works, its robustness limits, and the EU AI Act rules. - [Weights](https://howaiworks.ai/glossary/weights.md) — updated 2026-07-23 A weight is a single learned number setting the strength of a connection between two neurons; together, the weights are the trained model. - [World Models](https://howaiworks.ai/glossary/world-models.md) — updated 2026-07-25 A world model learns an environment's dynamics — given a state and an action, predict the next state — letting an agent plan or train in imagination. - [Zero-shot Learning](https://howaiworks.ai/glossary/zero-shot-learning.md) — updated 2026-07-23 Performing a task with zero examples, on prior knowledge alone — and the split that matters: zero-shot prompting versus classical zero-shot learning.