AI Basics · Part 5
30 Confusing AI Terms, Explained
Here are 30 AI terms that keep popping up in articles, YouTube videos, and office meetings, each explained in a single line. Read straight through, or come back and use it like a dictionary whenever you need it. (Ctrl+F is your friend.)
Core Concepts
| Term | Plain-English meaning |
|---|---|
| AI (artificial intelligence) | The broadest umbrella term for any technology that behaves intelligently, like a human |
| Machine learning | Letting a system learn patterns from data on its own, instead of hand-coding rules |
| Deep learning | Machine learning built on deeply stacked, brain-inspired neural networks. The engine of today's AI boom |
| Generative AI | AI that creates new content — text, images, video, and more |
| LLM (large language model) | A gigantic AI model trained on vast amounts of text to work with language. The heart of ChatGPT |
| Model | The finished, trained AI program itself. "GPT-5" and "Claude" are model names |
| Algorithm | A procedure or method for solving a problem — the recipe used to build a model |
| AGI (artificial general intelligence) | A hypothetical AI that could handle any intellectual task a human can, not just one specific job |
Terms You'll Meet While Using AI
| Term | Plain-English meaning |
|---|---|
| Prompt | The request or question you give an AI. The single biggest factor in output quality |
| Token | The smallest unit an AI reads and writes text in. The basis for pricing and processing limits |
| Context window | How much an AI can keep in mind within a single conversation |
| Hallucination | When an AI confidently makes up content that isn't true |
| Multimodal | The ability to understand and generate not just text but also images, audio, and video |
| Reasoning | A model working through its thinking step by step before answering. A key ability of the latest models |
| System prompt | Standing instructions given to an AI before the conversation begins |
| Few-shot prompting | Showing a few examples of the style you want and having the AI follow the pattern |
Terms From the Tech News
| Term | Plain-English meaning |
|---|---|
| Parameters | The number of adjustable dials inside a model. More usually means smarter, but heavier |
| Pre-training | The first phase of training, where massive data builds a model's core abilities |
| Fine-tuning | A second phase that further shapes a trained model for a specific purpose |
| RLHF | A training method that uses human preference ratings to improve answer quality and safety |
| RAG | Searching external documents before answering and using them as evidence. Reduces hallucination |
| Open-source model | A model released for anyone to download and use (Llama and others) |
| On-device AI | AI that runs directly on your phone or PC instead of on a server |
| GPU | The parallel-computing chip used to train and run AI. The hardware hero of the AI boom |
| Benchmark | A standardized set of test problems for comparing model performance |
One Step Further
| Term | Plain-English meaning |
|---|---|
| AI agent | An AI that plans and carries out multi-step tasks on its own from a single instruction (e.g., booking, research, coding) |
| MCP | A standard for connecting AI to external tools and data. Often called "the USB port for AI" |
| API | The connection channel that lets your own program call an AI like a building block |
| Vibe coding | Building software by describing what you want to an AI in plain words instead of writing the code yourself |
| Deepfake | AI-generated fake faces or voices. A technology with serious potential for abuse |
| AI alignment | The research field focused on making AI behave in line with human intentions and values |
Know just these 30 and you'll breeze through 90% of AI news articles without getting stuck. I'll keep adding to this page as new terms emerge.
That wraps up the basics series. Now that the theory is covered, head over to the Reviews, where I put real services through their paces.