What AI Actually Is

Most explanations of AI either assume you already know, or drift into metaphors about brains that leave you no better off. Here is the plain version.

It predicts what comes next

The tools you have used — ChatGPT, Claude, Gemini, the assistant inside your email — are built on large language models. A model like this was trained on an enormous quantity of writing until it got very good at one narrow skill: predicting what text should come next.

That is genuinely all it does. Everything else — answering, summarising, translating, drafting — falls out of doing that prediction extremely well.

When you type a question, the model is not looking anything up. It is producing the words that most plausibly follow your question, one chunk at a time, based on patterns it absorbed during training.

Why that matters immediately

Once you hold that idea, most of AI's odd behaviour stops being mysterious.

It writes fluently because fluency is exactly what it learned. It sounds confident because confident text is what it mostly saw. It sometimes invents things because inventing something plausible and recalling something true are, from its point of view, the same operation.

It is not lying when it gets something wrong. It has no concept of true or false. It is producing likely text, and most of the time likely text happens to be accurate.

What it learned from

Everything the model can do comes from its training data — the material it was shown. That sets a hard boundary:

• If a subject was written about extensively, it will handle it well.
• If a subject is niche, local, or very recent, it has thin patterns to draw on.
• If the material carried a bias, the model carries it too.

Training also stops at a date, called the knowledge cutoff. Anything after that is invisible unless the tool can search the web at the moment you ask.

What it is not

It is worth being clear about three things AI is not, because a lot of marketing implies otherwise.

It is not a database. It does not contain documents you can retrieve. It contains patterns.

It does not understand you in the way a colleague does. It models language about the world, not the world.

It is not a person. It has no goals, no memory of you between conversations unless the product adds one, and no stake in whether you succeed.

None of that makes it less useful. It makes it predictably useful, which is better.

The one sentence to keep

If you remember nothing else from this chapter: AI predicts plausible text; it does not look things up.

Everything in the rest of this course follows from that.