Lesson 1 of 6 · 6 min read
What AI is
A plain account of the chatbots and similar tools people actually meet — what they are doing when they answer, and what they are not.
A tool, not a teammate
In this course, “AI” means software that learned patterns from a huge number of examples and now uses those patterns to draft text, label images, sort items, or suggest a next step. The chat box on a phone is the example most people know. It is a product, with a vendor, a design, and limits.
It is not a person. It does not know you, owe you honesty, or hold a license. A smooth paragraph is not a sign that someone checked the facts.
What a model is doing
A model is the part that was trained. Training means showing the software many examples until it gets good at a narrow job: often, guessing a likely next word, pixel, or label. After training, you give it a new input — a question, a photo, a spreadsheet column — and it produces an output that resembles what followed that kind of input in the past.
Some products add tools: a web search, a calculator, your company files. If the product says it used a tool, that is a different and usually better situation than a free-form guess. You still read the result. The tool can search the wrong page or do the wrong math.
Narrow jobs and open chat
A spam filter, a map route, and a chatbot are all called AI in ads. They are not the same. A spam filter has one job and a known kind of mistake. A chatbot will attempt almost any sentence you type, including jobs it is bad at: the law in your state, a dose of medicine, the quote you half remember.
Treat a wide-open chat as a first draft machine. Treat a narrow tool as a helper for that one job, and learn what mistakes that job usually makes.
Keep these
- A chatbot produces a likely reply. It does not understand the way a person does.
- A model was trained on examples. It is not looking up a single official book unless the product says it is using a source.
- Polished language is not the same thing as a checked fact.