Jonáš SvětlíkA no-nonsense AI glossary
A lot of jargon flies around AI. Here it is in plain English - come back whenever you hit a term you do not know.
- AI (artificial intelligence) - an umbrella term for programs that handle things that used to need a person. These days it usually means language models.
- LLM (large language model) - the kind of AI behind ChatGPT and friends. It learned from a huge amount of text and can write and answer.
- Prompt - the instruction you write for the AI. The better the prompt, the better the answer.
- Token - a piece of a word, the unit AI counts in. Both text length and price are measured in tokens (roughly 4 tokens = 3 English words).
- Context / context window - how much text the AI "remembers" at once in a single conversation. If it gets too long, the older part falls away.
- Hallucination - when AI confidently invents something that is not true. The main reason to check facts.
- Model - the particular "brain" (GPT-4, Claude and so on). Newer and bigger models tend to be smarter, and pricier.
- Training - the process by which the model learned from data. It happens up front, not with every question you ask.
- RAG - a technique where AI answers only from your own documents and cites the source. Excellent against hallucinations.
- Agent - AI that does not just answer but takes steps of its own (sends an email, writes into a spreadsheet).
- API / API key - a way to plug AI into another program. An API key is like the password you sign in with.
- Prompt engineering - the skill of writing instructions so the AI does exactly what you want.
- Multimodal - a model that handles not just text but images, audio or video too.
- Automation / workflow - linking steps together so they run on their own. Often in a tool like n8n, quite possibly with AI inside.
- Fine-tuning - teaching a model on your specific data or style. Most companies do not need it - a good prompt or RAG is enough.
- Open-source model - a model you can download and run yourself. More privacy, more work to deploy.
You have the basics - what next?
That was the entry course. Carry on with the prompting course to get far more out of AI. And if you want AI put to work in your company, you know where to find me.