Glossary
Words for the new machines
Plain-language definitions of the ideas behind modern AI, from tokens to agents. Each term links to a longer explanation.
- Agentic AI AI systems that can take autonomous actions — writing code, browsing the web, using tools — not just generating text.
- AI Agent A system where an LLM operates in a loop — perceiving its environment, deciding on actions, executing them, and observing results.
- AI Alignment The challenge of ensuring AI systems behave in accordance with human values and intentions.
- Attention Mechanism A technique that allows a model to focus on relevant parts of the input when producing output, weighting different tokens differently.
- Chain-of-Thought (CoT) A prompting technique where the model is asked to show its reasoning steps before giving a final answer.
- Context Window The maximum amount of text (measured in tokens) that a model can process in a single interaction.
- Embedding A dense vector representation of text that captures semantic meaning in numerical form.
- Few-Shot Learning Providing a model with a small number of examples to demonstrate a task, rather than extensive training data.
- Fine-Tuning Adapting a pre-trained model to a specific task or domain by training it further on specialized data.
- Function Calling A model capability where the LLM generates structured JSON to invoke predefined functions, enabling tool integration.
- Hallucination When an AI model generates plausible-sounding but factually incorrect or fabricated information.
- Large Language Model (LLM) A neural network trained on massive text datasets that can generate, understand, and manipulate human language.
- LoRA (Low-Rank Adaptation) An efficient fine-tuning method that trains small adapter layers instead of updating all model parameters.
- Multi-Agent Systems Architectures where multiple AI agents collaborate, each with specialized roles, to accomplish complex tasks.
- Pre-Training The initial training phase where a model learns language patterns from a massive, general-purpose text corpus.
- Prompt Engineering The practice of crafting effective inputs to AI models to get better, more reliable outputs.
- RAG (Retrieval-Augmented Generation) A technique that enhances LLM responses by retrieving relevant documents from a knowledge base before generating.
- Reasoning (in AI) The ability of AI models to perform logical deduction, planning, and multi-step problem solving.
- RLHF (Reinforcement Learning from Human Feedback) A training technique that uses human preferences to fine-tune models, making outputs more helpful, harmless, and honest.
- Semantic Search Search based on meaning rather than keyword matching, using embeddings to find conceptually similar content.
- Token The basic unit of text that a language model processes — typically a word, subword, or character.
- Tokenization The process of converting text into tokens that a model can process, using algorithms like BPE or SentencePiece.
- Tool Use / Function Calling The ability of an LLM to invoke external tools (code execution, web search, APIs) as part of generating a response.
- Transformer A neural network architecture based on self-attention, introduced in the 2017 paper 'Attention Is All You Need.'
- Vector Database A database optimized for storing and querying high-dimensional vector embeddings, enabling similarity search.