AI Glossary
Master the language of artificial intelligence with our comprehensive glossary. Each term includes clear definitions and real-world examples.
💡 Use search above to quickly find terms
A
A/B Testing
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
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
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
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
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)
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
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)
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
Buying where an AI agent handles discovery, comparison and checkout. Two rival protocols split the market: ACP behind ChatGPT, UCP behind Google.
Agentic Workflow
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
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 and Employment
Will AI take your job? Exposure concentrates in tasks, not whole occupations — what Frey-Osborne, the OECD and measured productivity studies found.
AI Architecture
How a production AI system is assembled: retrieval, context assembly, the model call, tool execution, validation, logging and evaluation.
AI Data Center
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
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 Energy Consumption
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 for Good
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)
Which AI rules apply to an organisation — by role, risk tier and jurisdiction — and what compliance requires: inventories, documentation, release gates.
AI Hallucinations
Models hallucinate because training and benchmarks reward guessing over admitting ignorance: a guess sometimes scores, "I don't know" scores zero every time.
AI Healthcare
Where AI is actually deployed in medicine — imaging, triage, ambient notes — and why 'beats doctors' headlines rarely survive contact with a clinic.
AI in Finance
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 in Science
Three different things get called AI for science. Two have produced verified results, the third mostly has not — and what got revised down.
AI Infrastructure
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
AI research is the systematic investigation into the development of algorithms, models, and systems that exhibit intelligent behavior.
AI Safety
AI safety is three fields, not one: present-day reliability, alignment, and catastrophic risk. What each actually claims, and why they get confused.
Ambient Clinical Documentation
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)
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
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)
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)
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)
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)
Artificial superintelligence: what it would mean, why the intelligence explosion argument is contested, and what labs and regulators actually do about it.
Attention Mechanism
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)
How computers turn sound into numbers: sampling, the Nyquist limit, spectrograms and mel filterbanks — the front end of every speech and audio model.
Autoencoder
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
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
Autonomous vehicle safety technologies, standards, and testing protocols designed to ensure self-driving cars can operate safely without human intervention.
B
Backpropagation
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
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)
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
Statistical bias, not fairness. Error splits into bias squared, variance and irreducible noise — derived, worked on numbers, and honest about where it breaks.
C
Calibration
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
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
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)
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
Chunking splits documents into smaller passages before embedding them for retrieval. Chunk size, overlap, and splitting strategy quietly decide RAG quality.
Class Imbalance
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)
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)
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
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
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)
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)
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
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
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
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
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
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)
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
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
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)
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
Conversational AI holds a multi-turn dialogue in natural language. Unlike an intent-based chatbot, it generates each reply rather than picking one.
Convolution
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)
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)
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
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.
D
Data Analysis
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
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
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
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
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)
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
What makes deep learning deep: each layer builds on the one below, so capacity grows multiplicatively with depth and only additively with width.
Deepfake
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)
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
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)
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)
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)
Disaggregated serving runs LLM prefill and decode on separate GPU pools because the two phases hit opposite limits: compute versus memory bandwidth.
Distributed Computing
Distributed computing splits work across separate machines that talk only by messages. Why partial failure, not speed, is the defining problem.
Distributed Training
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.
E
Early Stopping
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
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
AI that tutors, adapts practice and grades work — and what forty years of effect-size research actually shows about whether it improves learning.
Embedding
What an embedding vector actually is, why cosine similarity beats Euclidean distance, what dimensions buy you, and where the geometry misleads you.
Embodied AI
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
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
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
AI ethics is what happens when good principles provably conflict: fairness against fairness, privacy against accuracy, and whose values get to decide.
EUV Lithography
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)
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.
F
Feature Scaling
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)
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
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)
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)
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
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)
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
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)
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.
G
General Problem Solver (GPS)
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
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)
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
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)
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
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
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
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)
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
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)
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)
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
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.
H
High Bandwidth Memory (HBM)
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
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)
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)
Hybrid search runs keyword (BM25) and vector retrieval together, then fuses their ranked lists with Reciprocal Rank Fusion (RRF), a score-agnostic method.
Hyperparameter
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.
I
Image Generation
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
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
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
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)
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)
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.
K
k-Nearest Neighbors (kNN)
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
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)
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
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
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.
L
Large Language Model (LLM)
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
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
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
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)
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.
Loss Function
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
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.
Low-Rank Adaptation (LoRA)
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.
M
Machine Learning (ML)
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 Learning Operations (MLOps)
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.
Machine Unlearning
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)
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
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
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
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)
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.
Model Card
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
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)
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
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
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
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
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)
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
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.
N
Naive Bayes
A probabilistic classifier applying Bayes' theorem under a deliberately false feature-independence assumption — fast, data-light, a strong text baseline.
Natural Language Processing (NLP)
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
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)
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
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
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)
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
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.
O
One-shot Learning
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
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
In machine learning, optimization means minimizing a loss function to fit a model's parameters — the iterative, gradient-driven mechanism behind training.
Overfitting
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.
P
Parallel Processing
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
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)
Pattern recognition is the automatic discovery and classification of regularities in data — the older name for what became statistical machine learning.
Performance
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
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
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)
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
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
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
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)
Principal Component Analysis (PCA) reduces dimensions by projecting data onto orthogonal directions of maximum variance, ranked by the variance each keeps.
Privacy
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
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
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
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
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
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.
Q
Quantization
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)
Quantum computing uses qubits — held in superposition and entangled — for exponential speedups on a narrow set of problems, not general computing.
R
Random Forest (RF)
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
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
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)
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
The supervised-learning task of predicting a continuous number — a price, a temperature, a duration — from input features, in contrast to classification.
Regularization
Regularization deliberately constrains a model so it fits training data less perfectly, trading a little training accuracy for much better generalization.
Reinforcement Learning (RL)
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)
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)
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
Representation learning is machine learning that discovers the useful features of raw data automatically, instead of humans hand-engineering them.
Reranking (Cross-Encoder)
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)
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)
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
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)
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
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
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.
S
Scalable AI
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
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
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)
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)
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
Semantic search finds results by meaning, not exact keywords: it embeds the query and documents as vectors and returns the nearest ones.
Semantic Understanding
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
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)
Method for turning a pretrained autoregressive LLM into a diffusion language model by continuing pretraining with a bidirectional attention mask.
Small Language Model (SLM)
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
Social AI is artificial intelligence built to perceive, interpret, and take part in human social interaction — reading emotion, tone, intent, and context.
Speculative Decoding
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
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)
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)
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)
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)
Supervised fine-tuning trains a pretrained language model on curated prompt-response pairs with next-token loss, teaching it to follow instructions.
Supervised Learning
Training a model on labeled input-output pairs so it learns a mapping it can apply to new, unlabeled inputs.
Support Vector Machines (SVM)
A classifier that separates two classes with the widest possible margin, using support vectors and the kernel trick to handle non-linear data.
Sycophancy
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
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
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.
T
Temperature
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
The array manipulations — elementwise math, matrix multiplication, reshaping, broadcasting, reduction — that every neural network is built out of.
Tensor Processing Unit (TPU)
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
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
Turning unstructured text into structured data — the concrete NLP task catalog: classification, sentiment analysis, entity extraction and topic modeling.
Text Generation
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)
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
Data points indexed in time order, where the sequence matters and rows aren't independent — analyzed to forecast future values or flag anomalies.
Token
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
How text becomes the integers a language model reads: BPE, WordPiece and Unigram, why " strawberry" is one token, and what that breaks.
Training
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)
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
The 2017 neural-network architecture that processes a whole sequence in parallel using self-attention instead of recurrence — the basis of modern LLMs.
Transparency
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
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.
U
Underfitting
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
Finding structure in unlabeled data — no correct answers are given, so the model discovers groupings, patterns, or a compressed representation on its own.
V
Value Learning
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
A vector database stores embeddings with metadata, builds an approximate-nearest-neighbour index over them, and serves filtered similarity queries at scale.
Vector Search
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
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
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
Video generation is an AI technology that creates moving images and scenes from text descriptions, images, or other video clips.
Vision Transformer (ViT)
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
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
How AI turns spoken words into text (speech recognition / ASR) — and how that differs from identifying who is speaking (speaker recognition).
W
Watermarking (SynthID)
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
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
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.
Can't Find a Term?
We're constantly expanding our glossary. Let us know what AI terms you'd like us to explain!