AI Glossary
Clear explanations of 117+ artificial intelligence and machine learning terms. From basic concepts to advanced techniques.
A
A/B Testing
Comparing two versions of a model or system to determine which performs better.
Accuracy
Metric measuring the proportion of correct predictions made by a model out of all predictions.
Activation Function
Mathematical function in neural networks that determines whether a neuron should be activated, introducing non-linearity.
AGI (Artificial General Intelligence)
Hypothetical AI with human-like ability to understand, learn, and apply intelligence across diverse tasks.
AI Safety
Field focused on ensuring AI systems are reliable, secure, and beneficial to humanity.
AI Winter
Period of reduced funding and interest in AI research, historically occurring when expectations exceeded capabilities.
Algorithm
A set of rules or instructions given to an AI system to help it learn from data and make decisions or predictions.
Alignment Problem
Challenge of ensuring AI systems behave according to human values and intentions.
Annotation
Process of labeling data with correct answers for supervised learning.
API (Application Programming Interface)
A set of rules and protocols that allows different software applications to communicate with each other. AI APIs let developers integrate AI capabilities into their apps.
Artificial Intelligence (AI)
The simulation of human intelligence processes by machines, especially computer systems. These processes include learning, reasoning, and self-correction.
Attention Mechanism
Technique allowing models to focus on specific parts of input when producing output, crucial for modern NLP.
Autoencoder
Neural network that learns to compress data into a lower-dimensional representation and then reconstruct it.
AutoML
Automated Machine Learning - using AI to automate the process of applying ML to real-world problems.
B
Backpropagation
Algorithm for calculating gradients in neural networks, used to update weights during training.
Batch Normalization
Technique that normalizes inputs to each layer, accelerating training and improving stability.
Batch Size
Number of training examples used in one iteration of model training.
Bias (AI Bias)
Systematic errors in AI systems that create unfair outcomes, often reflecting prejudices in training data or algorithm design.
Bias (in Neural Networks)
Learnable parameter that allows shifting the activation function, helping the model fit data better.
C
Checkpoint
Saved state of a model during training, allowing resumption or rollback if needed.
Classification
ML task of predicting which category or class an input belongs to, such as spam detection or image recognition.
Clustering
Unsupervised learning technique that groups similar data points together without predefined labels.
Computer Vision
Field of AI that trains computers to interpret and understand visual information from the world, enabling applications like facial recognition and object detection.
Confusion Matrix
Table used to evaluate classification model performance, showing true positives, false positives, true negatives, and false negatives.
Context Window
Maximum amount of text an LLM can process at once, measured in tokens.
Convolutional Neural Network (CNN)
Type of deep learning network especially effective for image analysis, using convolutional layers to detect features.
Cost Function
Overall measure of model error across all training examples, often the average of the loss function.
Cross-Validation
Technique for assessing model performance by training and testing on different subsets of data.
D
Data Augmentation
Technique of creating additional training data by applying transformations to existing data (like rotating images).
Data Cleaning
Process of detecting and correcting errors, inconsistencies, and missing values in datasets.
Data Pipeline
Series of data processing steps that transform raw data into a format suitable for ML.
Data Preprocessing
Preparing raw data for ML by cleaning, transforming, and formatting it appropriately.
Decision Tree
ML model that makes predictions by following a tree-like structure of decisions based on features.
Deep Learning
A subset of machine learning based on artificial neural networks. It uses multiple layers to progressively extract higher-level features from raw input.
Dimensionality Reduction
Reducing the number of features in data while retaining important information, making data easier to work with.
Dropout
Regularization technique that randomly ignores neurons during training to prevent overfitting.
E
Early Stopping
Stopping model training when performance on validation data stops improving to prevent overfitting.
Edge AI
Running AI algorithms locally on devices (edge) rather than in the cloud, enabling faster processing and better privacy.
Embedding
A way of representing data (like words or images) as vectors of numbers that capture semantic meaning and relationships.
Encoder-Decoder
Architecture where an encoder processes input into a representation, and a decoder generates output from that representation.
Ensemble Learning
Combining multiple models to produce better predictions than any single model.
Epoch
One complete pass through the entire training dataset during the training process of a machine learning model.
Ethical AI
Development and deployment of AI systems that align with human values and ethical principles.
Explainable AI (XAI)
AI systems designed to make their decisions and processes understandable to humans.
F
F1 Score
Harmonic mean of precision and recall, providing a single metric that balances both measures.
Feature
An individual measurable property or characteristic of a phenomenon being observed and used as input for ML models.
Feature Engineering
The process of selecting, creating, and transforming data features to improve machine learning model performance.
Federated Learning
ML approach where models are trained across multiple decentralized devices without exchanging raw data.
Few-Shot Learning
ML approach where models learn to perform tasks with only a few examples.
Fine-Tuning
Process of adjusting a pre-trained model on specific data to improve its performance for particular tasks or domains.
G
GANs (Generative Adversarial Networks)
Two neural networks (generator and discriminator) that compete to create realistic synthetic data.
Generative AI
AI systems that can create new content including text, images, audio, and video based on prompts or training data.
Gradient Descent
Optimization algorithm used to minimize the error of a model by iteratively adjusting parameters in the direction of steepest descent.
Ground Truth
The actual, correct answer used to train or evaluate a model.
H
Hallucination (AI)
When AI models generate false or nonsensical information presented as fact.
Hyperparameter
Configuration settings used to control the learning process of a machine learning algorithm, set before training begins.
I
Image Recognition
AI ability to identify objects, people, places, and actions in images.
Inference
The process of using a trained model to make predictions or decisions on new, unseen data.
K
K-Means Clustering
Algorithm that groups data into K clusters based on feature similarity.
L
Label
The target or answer in supervised learning - what the model is trying to predict.
Large Language Model (LLM)
AI models trained on vast amounts of text data that can understand and generate human-like text. Examples include GPT-4, Claude, and Gemini.
Learning Rate
Hyperparameter that controls how much model weights are adjusted during training. Too high can cause instability, too low makes training slow.
Loss Function
Mathematical function that measures how wrong a model's predictions are, guiding the training process.
M
Machine Learning (ML)
A subset of AI that enables systems to learn and improve from experience without being explicitly programmed. It focuses on developing algorithms that can analyze data and make predictions.
Machine Translation
Automated translation of text from one language to another using AI.
Meta-Learning
Learning to learn - training models to quickly adapt to new tasks with minimal data.
MLOps
Set of practices combining Machine Learning, DevOps, and Data Engineering to deploy and maintain ML models in production.
Model
The output of a machine learning algorithm trained on data. It's used to make predictions or decisions without being explicitly programmed.
Model Architecture
The structure and design of a neural network, including layers, connections, and parameters.
Model Drift
When a deployed model's performance degrades over time as the real-world data differs from training data.
Multimodal AI
AI systems that can process and understand multiple types of data (text, images, audio) simultaneously.
N
Named Entity Recognition (NER)
NLP task of identifying and classifying named entities (people, organizations, locations) in text.
Narrow AI (Weak AI)
AI designed for specific tasks (like chess or image recognition), as opposed to general intelligence.
Natural Language Processing (NLP)
A branch of AI that helps computers understand, interpret, and manipulate human language. Powers chatbots, translation, and text analysis.
Neural Network
A computing system inspired by biological neural networks that constitute animal brains. It consists of interconnected nodes (neurons) that process information.
Normalization
Scaling features to a standard range (often 0-1) to ensure they contribute equally to model training.
O
Object Detection
Computer vision task that identifies and locates objects within an image or video.
One-Shot Learning
Learning from a single example, common in facial recognition and signature verification.
Optimizer
Algorithm that adjusts model parameters to minimize the loss function (e.g., Adam, SGD).
Overfitting
When a model learns the training data too well, including its noise and outliers, leading to poor performance on new data.
P
Precision
Metric measuring the proportion of true positive predictions out of all positive predictions made by a model.
Principal Component Analysis (PCA)
Technique for dimensionality reduction by transforming data to a new coordinate system.
Prompt Engineering
The practice of crafting effective instructions (prompts) to get desired outputs from AI models, especially large language models.
R
Random Forest
Ensemble method using multiple decision trees to improve prediction accuracy and control overfitting.
Recall
Metric measuring the proportion of true positives out of all actual positive cases in the dataset.
Recurrent Neural Network (RNN)
Neural network designed to work with sequential data by maintaining a 'memory' of previous inputs.
Regression
ML task of predicting continuous numerical values, such as house prices or temperature forecasts.
Regularization
Techniques to prevent overfitting by adding constraints or penalties to the model.
Reinforcement Learning
ML technique where an agent learns to make decisions by receiving rewards or penalties for its actions in an environment.
Responsible AI
Framework for developing AI that is fair, transparent, accountable, and respects privacy.
Retrieval-Augmented Generation (RAG)
Technique where LLMs retrieve relevant information from databases before generating responses, improving accuracy.
ROC Curve
Receiver Operating Characteristic curve that illustrates the diagnostic ability of a binary classifier as its discrimination threshold varies.
S
Semantic Segmentation
Computer vision task that classifies each pixel in an image to a specific category.
Sentiment Analysis
NLP technique that determines the emotional tone or opinion expressed in text (positive, negative, or neutral).
Singularity
Hypothetical future point when AI advancement becomes uncontrollable and irreversible, leading to unforeseeable changes.
Speech Recognition
Technology that enables computers to identify and process human speech into text.
Standardization
Transforming features to have zero mean and unit variance for better model performance.
Strong AI
Theoretical AI with consciousness and self-awareness equivalent to human intelligence.
Supervised Learning
Machine learning approach where models are trained on labeled data, with known input-output pairs to learn patterns.
Support Vector Machine (SVM)
ML algorithm that finds the optimal boundary (hyperplane) to separate different classes of data.
Synthetic Data
Artificially generated data that mimics real-world data, used when real data is scarce or sensitive.
T
Temperature (in AI)
Parameter controlling randomness in AI text generation. Higher values make output more creative, lower values more deterministic.
Text Classification
Task of assigning predefined categories to text documents, like spam filtering or topic categorization.
Text-to-Speech (TTS)
AI technology that converts written text into spoken audio.
Token
In NLP, a piece of text broken down by the model (can be a word, part of a word, or character). Models process text as sequences of tokens.
Top-k Sampling
Text generation technique that considers only the k most likely next tokens.
Top-p Sampling (Nucleus Sampling)
Text generation method that samples from the smallest set of tokens whose cumulative probability exceeds p.
Train-Test Split
Dividing data into separate sets for training the model and evaluating its performance.
Training Data
The dataset used to teach a machine learning model. The quality and quantity of training data significantly impacts model performance.
Transfer Learning
Technique where a model developed for one task is reused as the starting point for a model on a second task, saving time and resources.
Transformer
Neural network architecture that uses attention mechanisms, powering modern LLMs like GPT and BERT.
Turing Test
Test of a machine's ability to exhibit intelligent behavior indistinguishable from a human.
U
Underfitting
When a model is too simple to capture the underlying patterns in the data, resulting in poor performance on both training and new data.
Unsupervised Learning
ML approach where models find patterns in unlabeled data without pre-existing tags or categories.
V
Validation Set
Subset of data used during training to tune model hyperparameters and prevent overfitting.
W
Weights
Learned parameters in a neural network that determine the strength of connections between neurons.
Z
Zero-Shot Learning
Model's ability to perform tasks it wasn't explicitly trained on, using knowledge from related tasks.
Ready to Put These Terms into Practice?
Now that you understand the terminology, try our free AI tools or explore our comprehensive blog.