Artificial Intelligence Glossary: 40+ AI Terms Explained (2026)

Bookmark this page. Each term below is short and skimmable, and many link to a full guide if you want the deep version.

Core AI Concepts

Start here. These are the foundation words that every other AI term builds on.

• Artificial Intelligence (AI). Software that performs tasks we usually associate with human thinking, like recognizing speech, making predictions, or writing text. For the full picture, see our beginner guide to artificial intelligence.
• Algorithm. A set of step-by-step rules a computer follows to solve a problem. AI uses algorithms to learn patterns from data.
• Model. The trained "brain" an AI uses to make decisions. Think of it as the finished recipe after the system has studied lots of examples.
• Training data. The examples an AI learns from. Better, cleaner data usually means a smarter model.
• Inference. The moment an AI actually uses what it learned to answer your question or make a prediction.
• Narrow AI, General AI, and Superintelligence. Three levels of capability, from today's task-specific tools to a theoretical system smarter than humans. We break these down in types of AI: narrow, general, and super.
• Turing Test. A classic check from 1950: if you cannot tell whether you are chatting with a human or a machine, the machine passes. The Stanford Encyclopedia of Philosophy has the full history.

Machine Learning Terms

Machine learning is how most modern AI actually learns. These terms describe how that learning happens.

• Machine Learning (ML). A type of AI that learns patterns from data instead of being hand-coded with rules. See our machine learning beginner guide.
• Supervised learning. Training with labeled examples, like photos tagged "cat" or "dog," so the model learns the right answer.
• Unsupervised learning. Training with unlabeled data so the model finds hidden patterns on its own. Compare both in supervised vs unsupervised learning.
• Reinforcement learning. Learning by trial and error with rewards and penalties is the same method behind game-playing AIs. More in reinforcement learning basics.
• Feature. A single measurable input the model looks at, like price, age, or word count.
• Label. The correct answer is attached to a training example in supervised learning.
• Overfitting. When a model memorizes its training data too closely and then performs badly on new, real-world data.

Deep Learning and Neural Networks

This is the engine behind today's most powerful AI, from voice assistants to image generators.

• Deep learning. A branch of machine learning that uses many layers to learn complex patterns. Read what is deep learning for the full story.
• Neural network. A model loosely inspired by the human brain, built from connected "neurons" that pass signals along. We explain it simply in neural networks explained.
• Parameter (weight). An internal value the model adjusts during training. Big models have billions of them.
• Layer. A stage in a neural network. "Deep" learning just means many layers stacked together.
• GPU. A graphics chip that does the heavy math AI training needs. It is the reason modern AI has become fast.

Still mixing up AI, machine learning, and deep learning? Our breakdown of AI vs machine learning vs deep learning sorts out how they nest inside each other.

Generative AI and LLM Terms

These are the words behind tools like ChatGPT, Claude, and Midjourney. If you use AI day-to-day, learn this group well.

• Generative AI. AI that creates new content, such as text, images, or code, rather than just analyzing existing data. See generative AI vs traditional AI.
• Large Language Model (LLM). A huge AI trained on text that can write, summarize, and answer questions. ChatGPT runs on one.
• Token. A small chunk of text, roughly a word or part of a word. Models read and bill by tokens, not letters.
• Prompt. The instruction or question you give an AI. Better prompts get better answers.
• Prompt engineering. The skill of writing clear, specific prompts to get the result you want.
• Hallucination. When an AI states something false with full confidence. Always verify important facts.
• Fine-tuning. Further training a general model on your own data so it specializes in your topic or style.
• RAG (retrieval-augmented generation). A setup where the AI looks up real documents before answering, which cuts down on hallucinations.
• Diffusion model. The technique behind most AI image generators, which builds a picture by removing noise step by step.

Language and Vision Terms

How AI reads words and sees images.

• Natural Language Processing (NLP). The field that helps computers understand and produce human language. Full guide: what is natural language processing.
• Computer vision. Teaching machines to interpret photos and video, from face unlock to self-driving cars. See what is computer vision.
• Speech recognition. Turning spoken words into text, the tech behind voice assistants and live captions.
• Sentiment analysis. Detecting whether text feels positive, negative, or neutral is often used to read customer reviews.

AI Agents and Automation

The newest wave of AI does not just answer; it acts.

• AI agent. An AI that can plan and carry out multi-step tasks on its own, like booking a trip or running a workflow, without being re-prompted at every step.
• Chatbot. A program that holds a conversation in text or voice, from simple support bots to advanced assistants.
• API. A connection that lets one app use another app's AI, so developers can add AI features without building a model.
• Multimodal AI. A model that handles more than one type of input, such as text, images, and audio together.
• AGI (Artificial General Intelligence). A still-theoretical AI that could match human ability across almost any task.

Ethics, Safety and Governance

The terms shaping how AI is used responsibly. Expect to hear these more every year. Government bodies like the U.S. National Institute of Standards and Technology now publish frameworks around them.

• AI bias. When a model produces unfair results because its training data was skewed. We cover real cases in bias in AI.
• Alignment. The effort to make sure an AI's goals match human values and intentions.
• Explainability (black box). How clearly we can understand why an AI made a decision. Many deep models are hard to explain, hence "black box."
• Deepfake. Realistic but fake images, audio, or video created by AI, often used to mislead.
• Data privacy. Protecting the personal information AI systems collect and use. More in AI and data privacy concerns.

Frequently Asked Questions

Conclusion

You do not need a computer science degree to follow an AI conversation. Once these artificial intelligence terms click, headlines make sense, tools feel less mysterious, and you can judge the hype for yourself.

Keep this glossary handy and revisit it whenever a new word trips you up. The vocabulary is the doorway, and you just walked through it.

Which term still feels fuzzy? Tell us in the comments, and we will explain it and share this glossary with anyone who is just getting started with AI.