Reinforcement Learning
Training by trial, error and reward rather than by examples. It is how models are tuned to be helpful and to refuse things they should refuse.
Related terms
Embedding
A way of turning text into numbers so a computer can tell that “web hosting” and “website server space” are about the same thing. It is what makes search by meaning possible rather than search by exact word.
Prompt Engineering
The craft of writing instructions that get reliable results from an AI — being specific, giving examples, saying what to avoid. Less mysterious than it sounds, and mostly the same skill as briefing a new colleague well.
Watermarking
Marking AI-generated content so it can be identified later, often invisibly. Part of the same effort as provenance, and equally early.
Product Schema
Markup for an item you sell: name, price, currency, availability, condition. Without it an assistant reading your shop has to guess the price from the page, and sometimes guesses wrong.
Token
The unit an AI reads and writes in — roughly three quarters of a word. Models are priced and limited by tokens, which is why very long documents cost more and sometimes get truncated.
AI Referral Traffic
Visitors who arrive from an AI assistant rather than a search engine. Usually small in number and unusually well-qualified, because they clicked after reading a recommendation.