The AI world loves acronyms and technical terms. But you don't need a PhD to understand them. Here's a plain-English glossary of the 40 most important AI terms you'll encounter.
Core Concepts (A-F)
- AGI (Artificial General Intelligence) — A hypothetical AI that can do any intellectual task a human can. Doesn't exist yet.
- Algorithm — A set of rules or instructions a computer follows. AI algorithms learn patterns from data instead of following fixed rules.
- Alignment — Making sure AI systems do what humans actually want. A major focus of AI safety research.
- API (Application Programming Interface) — A way for software to communicate with AI models. Developers use APIs to add AI to their apps.
- Bias — When AI produces unfair or skewed results because of patterns in its training data.
- Chatbot — An AI program designed for conversation. ChatGPT, Claude, and Gemini are chatbots.
- Context window — The maximum amount of text an AI can process at once. Measured in tokens. Larger is better for long documents.
- Deep learning — A type of machine learning using neural networks with many layers. Powers modern AI.
- Embedding — Converting text into numbers that capture meaning. Allows AI to understand relationships between words and concepts.
- Fine-tuning — Taking a pre-trained model and training it further on specific data for a particular task.
Core Concepts (G-P)
- Generative AI — AI that creates new content (text, images, code, music) rather than just analyzing existing data.
- GPT (Generative Pre-trained Transformer) — OpenAI's family of language models. GPT-4 is the latest version.
- Grounding — Connecting AI responses to factual sources to reduce hallucinations.
- Hallucination — When AI generates confident-sounding but factually incorrect information. A known limitation of all LLMs.
- Inference — When a trained AI model processes new input and generates output. This is what happens when you chat with AI.
- LLM (Large Language Model) — A massive AI model trained on text data. Powers ChatGPT, Claude, Gemini, and similar tools.
- Machine learning — AI that learns from data rather than following explicit rules. A subset of AI.
- Multimodal — AI that can process multiple types of input: text, images, audio, video.
- Neural network — A computing system inspired by the human brain. Made of interconnected nodes that process information in layers.
- NLP (Natural Language Processing) — AI's ability to understand and generate human language.
- Parameters — The adjustable values in a neural network that are learned during training. More parameters generally mean more capability.
- Prompt — The text you type into an AI chatbot. Prompt quality determines output quality.
Core Concepts (R-Z)
- RAG (Retrieval-Augmented Generation) — A technique where AI looks up relevant documents before answering, improving accuracy.
- RLHF (Reinforcement Learning from Human Feedback) — Training AI to prefer responses that humans rate highly.
- Semantic search — Search based on meaning rather than exact keyword matching. AI understands what you're looking for.
- System prompt — Hidden instructions that define how an AI assistant behaves. Sets the personality and rules.
- Temperature — A setting that controls how creative or predictable AI responses are. Higher = more creative, lower = more focused.
- Token — The unit AI uses to process text. Roughly ¾ of a word. "Hello world" ≈ 2 tokens.
- Training data — The text, images, or other data used to teach an AI model.
- Transformer — The neural network architecture behind all modern LLMs. Introduced in the famous "Attention Is All You Need" paper.
- Zero-shot — When AI performs a task it wasn't specifically trained for, based on its general knowledge.
- Few-shot — Giving AI a few examples of what you want before asking it to perform the task.
Keep This Handy
Bookmark this page and refer back whenever you encounter unfamiliar AI terms. As you build your AI skills, these concepts will become second nature. Start putting them into practice with our free prompt library.