What people mean by "AI"
"AI" stands for artificial intelligence. The term covers a lot, from chess programs to spam filters. But when people say "AI" today, they usually mean a chatbot or AI assistant: a tool you type to, and it writes back in plain language.
Most of these tools run on a large language model, or LLM. A language model is a computer program trained on huge amounts of text so it can produce text of its own. "Large" means it was built from a lot of data. LLMs are one kind of machine learning, which means software that learns patterns from examples instead of following rules a person wrote out.
Learned patterns, not written rules
Think about how regular software works. A tax calculator follows rules a programmer typed in: if income is above this amount, apply that rate. Every step is spelled out.
Machine learning works differently. Nobody writes the rules. Instead, the program is shown a huge number of examples and slowly adjusts itself until it gets good at spotting patterns in them. A spam filter learns what spam looks like from piles of emails people marked as spam. A language model learns how words tend to follow each other from a vast amount of written text.
There's a catch. Because the patterns were learned, not written, even the people who build these systems can't point to one line of code that explains a given answer.
What it is and isn't
A few common mix-ups are worth clearing up:
- There's no good evidence it is sentient. A chatbot can sound warm or upset, but that's a pattern in its writing, not a sign that it has feelings or wants anything.
- It is not a search engine. A plain language model doesn't look things up. It writes an answer based on patterns it learned. Some apps add a search step, which you'll meet later in this branch.
- It is not always right. It can state something false in the same sure tone it uses for facts.
Good at, bad at
In broad strokes, these tools are good at working with language. They can draft an email, sum up a long article, explain an idea in simpler words, brainstorm names, or rewrite something in a new tone.
They are weaker at things that need exact facts, careful math, or fresh news, unless the app gives them extra help. They also don't know your situation unless you tell them. A good way to think of one is as a fast, well-read helper whose work you still need to check.
Try it
Ask a chatbot: "How are you different from a search engine? Answer in three sentences." Then type the same question into a search engine. Notice how one writes you a new answer, while the other points you to pages that already exist.