AI Glossary

The simulation of human intelligence by machines — enabling computers to learn, reason, and make decisions.

An AI system that can perceive its environment, make decisions, use tools, and take autonomous actions to achieve a goal.

A neural network technique that lets a model focus on the most relevant parts of an input when producing an output.

A mathematical function applied inside neural networks that introduces nonlinearity and lets models learn complex patterns.

The effort to make AI systems behave in ways that are helpful, safe, and consistent with human goals and values.

The algorithm used to compute how each model parameter contributed to error so the network can update itself during training.

The maximum amount of text (measured in tokens) that an AI model can process in a single interaction.

A vision-language model that learns shared representations of images and text so they can be compared in the same embedding space.

The field of AI focused on enabling machines to interpret and understand images and video.

A subset of machine learning using multi-layered neural networks to learn complex patterns from large datasets.

An AI model that generates images or other media by learning to reverse a gradual noise-adding process.

A numerical vector representation of text, images, or other data that captures semantic meaning in a high-dimensional space.

Training a pre-trained model further on a smaller, task-specific dataset to adapt it for a particular use case.

A capability that lets language models request structured tool or API calls instead of only generating plain text.

A prompting or training approach where a model is shown a small number of examples before handling a new task.

AI that can create new content — text, images, audio, video, and code — rather than just classifying or predicting.

An optimization method that updates model parameters in the direction that most reduces prediction error.

When an AI model generates confident-sounding but factually incorrect or fabricated information.

The process of using a trained AI model to generate predictions, classifications, or responses on new input data.

The ability of a model to learn patterns from instructions and examples provided inside the current prompt without updating its weights.

A fine-tuning approach where a model is trained on many instruction-and-response examples to improve its ability to follow user requests.

A structured representation of entities and the relationships between them, used to organize and reason over information.

A massive AI model trained on text data that can generate, summarize, translate, and reason about language.

A mathematical measure of how wrong a model’s predictions are during training.

Low-Rank Adaptation, a parameter-efficient fine-tuning method that updates a small set of low-rank matrices instead of the full model.

A subset of AI where systems learn from data to improve performance without being explicitly programmed.

AI systems that can process and generate multiple types of data — such as text, images, audio, and video — in a unified model.

A transformer technique that runs multiple attention operations in parallel so the model can capture different kinds of relationships at once.

The learned parameter values in a neural network that determine how input signals are transformed into outputs.

A computational model loosely inspired by the brain, made of interconnected nodes (neurons) that process information in layers.

When a model learns the training data too specifically and performs poorly on new, unseen data.

The practice of crafting inputs to AI models to elicit better, more accurate, or more useful outputs.

The learned numerical values inside a neural network that store what the model has learned from training data.

A workflow pattern where multiple prompts are linked together so the output of one step becomes the input to the next.

A model that has already been trained on broad data and can then be adapted or used for downstream tasks.

A LoRA-based fine-tuning method that combines low-rank adapters with quantized base models to reduce memory requirements even further.

A technique that reduces model size and inference cost by storing weights and activations with lower numerical precision.

A technique that enhances LLM outputs by first retrieving relevant documents from an external knowledge base before generating a response.

Reinforcement Learning from Human Feedback — a training technique that aligns AI models with human preferences using human ratings.

A form of attention where each token in a sequence looks at every other token in the same sequence to build context-aware representations.

A high-priority instruction that sets the role, behavior, constraints, and goals for an AI model within an application.

A search approach that finds results based on meaning and intent rather than exact keyword matches.

The neural network architecture behind most modern AI — uses attention mechanisms to process sequences in parallel.

The process of splitting text into smaller units (tokens) that a language model can process.

A parameter that controls the randomness of an AI model's outputs — lower values are more deterministic, higher values are more creative.

The basic unit of text processed by a language model, often representing a word, subword, punctuation mark, or symbol.

AI generation that creates images from natural language prompts.

AI generation that creates video clips from natural language prompts.

When a model is too simple or insufficiently trained to capture meaningful patterns in the data.

A database optimized for storing and querying high-dimensional vector embeddings at scale.