Activation Function
Activation Function is a function applied to a neuron's output that lets neural networks model non-linear relationships.
Active Learning
Active Learning is a strategy where the model chooses the most useful examples for humans to label next.
Agent
Agent is an AI system that can take actions, use tools, and pursue goals across multiple steps with limited human input.
AGI
AGI is artificial general intelligence, a hypothetical AI able to perform any intellectual task a human can.
Alignment
Alignment is the effort to make AI systems behave in line with human values and intentions.
Anomaly Detection
Anomaly Detection is identifying data points that differ significantly from the norm, often used for fraud and fault detection.
API
API is an application programming interface that lets software send input to an AI model and receive its output.
Artificial Intelligence
Artificial Intelligence is the field of building computer systems that can perform tasks normally requiring human intelligence, such as understanding la...
Attention Mechanism
Attention Mechanism is a technique that lets a model weigh which parts of the input matter most when producing each part of…
Backpropagation
Backpropagation is the algorithm that calculates how to adjust each parameter in a neural network to reduce error.
Batch Size
Batch Size is the number of training examples processed before the model updates its parameters.
Benchmark
Benchmark is a standard dataset and metric used to compare the performance of different models.
Bias
Bias is systematic errors in AI outputs, often reflecting imbalances or prejudices present in the training data.
Bias Term
Bias Term is an extra learnable value added in a neuron that shifts the activation, helping models fit data better.
Catastrophic Forgetting
Catastrophic Forgetting is when a model trained on new data loses what it previously learned.
Chain of Thought
Chain of Thought is a prompting technique that asks a model to reason step by step, which improves performance on complex problems.
Classification
Classification is assigning input data to one of several predefined categories.
Closed Model
Closed Model is a proprietary AI model accessible only through an API, with weights kept private.
Clustering
Clustering is grouping similar data points together without predefined labels.
Computer Vision
Computer Vision is the field of AI that enables machines to interpret and understand visual information from images and video.
Confusion Matrix
Confusion Matrix is a table showing correct and incorrect predictions per class to evaluate a classifier.
Context Window
Context Window is the maximum amount of text a language model can consider at once when generating a response.
Convolutional Neural Network
Convolutional Neural Network is a neural network design especially effective at processing images by detecting local visual patterns.
Cross-Validation
Cross-Validation is a method of testing a model on multiple data splits to estimate how well it generalises.
Data Augmentation
Data Augmentation is expanding a dataset by creating modified copies of existing examples to improve model robustness.
Data Labelling
Data Labelling is the process of annotating data with the correct answers so it can be used for supervised learning.
Data Pipeline
Data Pipeline is the automated flow that collects, cleans, and delivers data to models for training or inference.
Dataset
Dataset is a structured collection of data used to train, validate, or test AI models.
Decision Tree
Decision Tree is a model that makes predictions by following a branching series of yes/no questions.
Deep Learning
Deep Learning is a subset of machine learning that uses multi-layered neural networks to learn complex patterns from large amounts of data.
Deepfake
Deepfake is synthetic media that convincingly replaces a person's likeness or voice, raising misuse concerns.
Diffusion Model
Diffusion Model is a generative model that creates images by gradually removing noise, used by many text-to-image tools.
Distillation
Distillation is training a smaller model to mimic a larger one so it runs faster while keeping much of the performance.
Dropout
Dropout is a regularisation method that randomly ignores some neurons during training to improve generalisation.
Edge AI
Edge AI is running AI models directly on local devices rather than in the cloud, for speed and privacy.
Embedding
Embedding is a numerical vector representation of text, images, or other data that captures meaning so similar items sit close together.
Ensemble Learning
Ensemble Learning is combining multiple models to produce better predictions than any single model alone.
Epoch
Epoch is one full pass of the training algorithm over the entire training dataset.
Explainability
Explainability is the degree to which a human can understand why an AI model made a particular decision.
F1 Score
F1 Score is a metric that balances precision and recall to evaluate classification performance.
Feature
Feature is an individual measurable property of the data that a model uses to make predictions.
Feature Engineering
Feature Engineering is the process of selecting and transforming raw data into features that improve model performance.
Federated Learning
Federated Learning is training models across many devices without moving raw data off them, preserving privacy.
Few-Shot Learning
Few-Shot Learning is giving a model a handful of examples in the prompt to guide it on a new task.
Fine-Tuning
Fine-Tuning is further training a pre-trained model on a specific dataset so it performs better on a particular task or domain.
Foundation Model
Foundation Model is a large model trained on broad data that can be adapted to many downstream tasks.
Function Calling
Function Calling is a capability that lets a language model trigger external tools or code in a structured way.
GAN
GAN is a generative adversarial network, where two neural networks compete to produce realistic synthetic data.
Generative AI
Generative AI is AI that creates new content — text, images, audio, video, or code — rather than only analysing or classifying…
GPU
GPU is a graphics processing unit, hardware widely used to train and run AI models because it handles parallel maths efficiently.
Gradient Boosting
Gradient Boosting is an ensemble technique that builds models sequentially, each correcting the errors of the last.
Gradient Descent
Gradient Descent is an optimisation method that iteratively adjusts model parameters to reduce prediction error.
Ground Truth
Ground Truth is the verified correct answers used to train and evaluate a model.
Grounding
Grounding is connecting a model's answers to verified sources or data to reduce errors.
Guardrails
Guardrails is rules and filters that constrain what an AI system is allowed to say or do.
Large Language Model
Large Language Model is a neural network trained on huge amounts of text that can understand and generate human-like language.
Latency
Latency is the time it takes for an AI system to return a result after receiving input.
Learning Rate
Learning Rate is a setting that controls how big each step is when a model updates its parameters during training.
Loss Function
Loss Function is a measure of how far a model's predictions are from the correct answers, which training tries to minimise.
Machine Learning
Machine Learning is a branch of AI where systems learn patterns from data and improve at a task without being explicitly programmed…
MLOps
MLOps is the practices and tools for deploying, monitoring, and maintaining machine learning models in production.
Model
Model is the trained mathematical system that takes input data and produces predictions or generated output.
Model Drift
Model Drift is the gradual decline in a model's accuracy as real-world data changes over time.
Multimodal AI
Multimodal AI is AI that can understand and generate more than one type of data, such as text and images together.
Named Entity Recognition
Named Entity Recognition is identifying and classifying key items in text, such as names, places, dates, and organisations.
Natural Language Processing
Natural Language Processing is the area of AI focused on enabling computers to understand, interpret, and generate human language.
Neural Network
Neural Network is a model loosely inspired by the brain, made of layers of connected nodes that transform input data into useful…
Object Detection
Object Detection is locating and labelling specific objects within an image or video.
Open Source Model
Open Source Model is an AI model whose weights and code are publicly available to use, study, and modify.
Optical Character Recognition
Optical Character Recognition is extracting readable text from images or scanned documents.
Overfitting
Overfitting is when a model learns the training data too closely, including its noise, and performs poorly on new data.
Overhead
Overhead is the extra compute, memory, or time an AI system needs beyond the core task.
Overparameterization
Overparameterization is using a model with far more parameters than strictly needed, common in modern deep learning.
Parameter
Parameter is an internal value a model learns during training; large models have billions of them.
Pre-training
Pre-training is the initial, large-scale training of a model on broad data before it is adapted to specific tasks.
Precision
Precision is of the items a model flagged as positive, the fraction that were actually correct.
Prompt
Prompt is the instruction or input you give a generative AI model to tell it what you want it to produce.
Prompt Engineering
Prompt Engineering is the practice of designing and refining prompts to get more accurate and useful results from AI models.
Random Forest
Random Forest is an ensemble of many decision trees whose combined vote produces more robust predictions.
Recurrent Neural Network
Recurrent Neural Network is a neural network designed for sequential data, where each step's output depends on previous steps.
Red Teaming
Red Teaming is deliberately stress-testing an AI system to find harmful or unsafe behaviours before release.
Regression
Regression is predicting a continuous numeric value, such as a price or temperature.
Regularization
Regularization is techniques that discourage overly complex models to reduce overfitting.
Reinforcement Learning
Reinforcement Learning is training an agent to make decisions by rewarding good actions and penalising bad ones over many trials.
Retrieval-Augmented Generation
Retrieval-Augmented Generation is a technique that lets a model pull in relevant external documents at query time so its answers are grounded…
RLHF
RLHF is reinforcement learning from human feedback, a method that aligns AI outputs with human preferences using human ratings.
Self-Supervised Learning
Self-Supervised Learning is learning from unlabelled data by creating its own training signals, key to modern LLMs.
Semantic Search
Semantic Search is searching by meaning rather than exact keywords, powered by embeddings.
Sentiment Analysis
Sentiment Analysis is automatically determining whether a piece of text expresses positive, negative, or neutral feeling.
Softmax
Softmax is a function that turns a set of numbers into probabilities that sum to one, used for classification.
Speech Recognition
Speech Recognition is converting spoken audio into written text, also known as speech-to-text.
Supervised Learning
Supervised Learning is training a model on labelled examples so it learns to map inputs to known correct outputs.
Synthetic Data
Synthetic Data is artificially generated data used to train or test models when real data is scarce or sensitive.
System Prompt
System Prompt is a hidden instruction that sets a model's role, tone, and rules before a conversation begins.
Temperature
Temperature is a setting that controls randomness in a model's output — lower is more focused, higher is more creative.
Text-to-Speech
Text-to-Speech is converting written text into natural-sounding spoken audio.
Token
Token is a chunk of text — often a word or part of a word — that a language model reads and…
Tokenization
Tokenization is the process of breaking text into tokens that a model can process.
TPU
TPU is a tensor processing unit, custom hardware designed by Google to accelerate machine learning workloads.
Training Data
Training Data is the dataset used to teach a machine learning model the patterns it needs to make predictions.
Transfer Learning
Transfer Learning is reusing a model trained on one task as the starting point for a related task, saving data and compute.
Transformer
Transformer is a neural network architecture based on attention that powers most modern language and multimodal models.
Turing Test
Turing Test is a test of whether a machine's conversation is indistinguishable from a human's.