What an AI model actually is
✓ completed — revisiting
Understand
When people say "AI" today, they almost always mean a large language model (LLM) — the technology behind ChatGPT, Claude, and Gemini.
Here is the whole idea in one sentence: an LLM is a system that has read an enormous amount of text and learned to predict what word comes next.
That sounds too simple to be useful, so sit with it. If you train something to predict the next word extremely well — across billions of sentences about cooking, code, law, history, feelings — it has to internalise patterns about how the world is described. "Predict the next word" turns out to be a backdoor into something that behaves a lot like reasoning.
A few consequences fall directly out of this:
- It is not looking things up. It has no database of facts it queries. It generates a plausible continuation. Usually plausible and true line up. Sometimes they don't (that's a hallucination — a confident, wrong answer; we cover it in its own lesson).
- It is not deterministic like a calculator. Ask the same thing twice, get two different phrasings. It's predicting likely text, not retrieving the answer.
- It has no memory between conversations unless the tool deliberately gives it some. Each chat starts fresh. (Its own lesson too.)
- It was trained on data up to a cutoff date. Out of the box it doesn't know what happened after that, unless the tool can search the web for it.
The mental model that will serve you for everything else in this course: treat it as an extremely well-read, fast, slightly overconfident collaborator — not an oracle, not a search engine, not a person. Everything you learn next is about directing that collaborator well.
See it
Type this into any chatbot: "The capital of Australia is" and stop there.
It will almost certainly continue with "Canberra" — and add a sentence or two unprompted. It wasn't "asked a question" and it didn't "look up Australia." It saw a sentence fragment and produced the most likely continuation based on the mountain of text it learned from, where that sentence is almost always completed with "Canberra."
Now try: "Write the first line of a noir detective story set in a laundromat." Same machine, same mechanism — predicting likely text — but now the likely continuation is creative prose. One engine, predicting-the-next-word, producing facts or fiction depending on what you set up. That is the entire trick, and seeing it once makes AI far less mysterious.
Try it
Open any AI chatbot. Send exactly this, with no other instructions: Finish this sentence and then stop: The hardest part of learning something new is. Then send a second message: Now write that same idea as a haiku.
You'll know it worked when
You get a fluent completion, then the same idea recast as a 3-line poem. You've watched one next-word-prediction engine do both factual-style completion and creative reshaping — proof that 'AI' here is one flexible text engine you steer, not a lookup tool.
Tried it? Paste what you got and the tutor will tell you if it worked.
Key terms
- Large language model (LLM)
- The technology behind tools like ChatGPT and Claude. A system trained on a huge amount of text that works by predicting the most likely next word.
- Next-word prediction
- The core mechanism of an LLM: given some text, it produces the most plausible continuation, one piece at a time. Reasoning-like behaviour emerges from doing this extremely well.
- Hallucination
- When an AI gives a confident answer that is plausible-sounding but actually wrong, because it generates likely text rather than retrieving verified facts.
- Training data
- The body of text an AI learned from, up to a fixed 'cutoff' date. Out of the box it knows nothing after that date unless a tool lets it search.
- Deterministic
- Always giving the exact same output for the same input (like a calculator). An LLM is NOT deterministic — ask the same thing twice and the wording differs.
All terms live in the keywords bank.
Explain it back
The real test: explain this in your own words, like you're teaching it. The tutor will tell you honestly if it's solid.
Check yourself
What is the core thing a large language model was trained to do?
Why can an LLM give a confident answer that is simply wrong?
Still fuzzy?
Talk it through with the tutor
Before I explain anything: when you ask a chatbot 'what's the capital of France' and it says 'Paris', what do you *think* is actually happening inside it to produce that word?
inherent.
Interesting word choice! Can you say a bit more about what you mean by "inherent"? Like, where do you imagine "Paris" is *stored* or *coming from* inside the AI?
She's gonna need help.
Ha, sounds like you might be thinking of the AI as a "she" with some knowledge tucked away somewhere!
So let me ask: when you picture that knowledge living inside the AI — do you imagine it's more like a **filing cabinet** where "Paris" is stored under "capitals of France"... or something else entirely?
The situation where.
It sounds like your thoughts might be trailing off a bit! No worries — let's keep it simple.
When the AI gives you the word "Paris", do you think it **looked that fact up somewhere**, or do you think something else happened to produce that word?
Don't worry.
So I sent a video.
Ha, I think we might have some crossed wires — I don't think a video came through on my end!
Let's get back to our puzzle. When the AI types out "Paris" as the capital of France, do you think there's a list of facts somewhere inside it that it's checking — like looking something up in an encyclopedia? Or does something else come to mind?
Yes, yes.
So you're thinking it's like looking something up in an encyclopedia — checking a stored list of facts. That's actually what a lot of people assume!
Here's a question to test that idea: if the AI really did work like an encyclopedia, what would you expect to happen if you asked it the **same question twice**? Would you always get exactly the same answer, word for word?
No crying.
Sí, seguro.
It seems like we might be getting a bit distracted! Let's stay with our puzzle — it's a good one.
You said you wouldn't always get the same answer. That's really interesting! So if an encyclopedia always gives you the exact same entry for "capital of France"... but the AI sometimes words things differently each time... what does that tell you about how the AI might be working?
That's even better!
I like your enthusiasm! But I want to make sure I understand what you mean — what is it that you think is "even better" about how the AI works compared to an encyclopedia?
instalações.
No pressure! Take your time. 😊
When you're ready, just pick one
Welcome.
No pressure! Take your time. 😊
When you're ready, just pick one:
**A)** The AI looks facts up, like a search engine.
**B)** The AI predicts the most likely next word, based on patterns it learned from reading lots of text.
Which feels closer?
Thank you! 😄
So, back to our question — when the AI says "Paris" in response to "what's the capital of France"...
Does it feel more like it's **looking something up**, or more like it's **completing a pattern** based on things it's learned? Just go with your gut!
Sorry.
No need to apologise at all! 😊
Just one simple choice when you're ready:
Does the AI work more like a **search engine** (looking things up) or more like a **pattern completer** (predicting the most likely next word based on things it's learned)?
I was talking.
No worries at all! Take your time. 😊
Which one is it?
Ah, I'm not going to just tell you — that's the fun part! 😄
Have a guess first. Which do YOU think it is:
**A)** Like a search engine — looking facts up.
**B)** Like a pattern completer — predicting the most likely next word based on things it's learned.
Just go with your gut!
There's only one there, isn't there?
Ha! You're right, there's a strong hint in how I've
That's the baseline.
Ha, fair enough — you spotted that I was leaning towards B! 😄
So let's say you accept that the AI is predicting the most likely next word based on patterns it's learned.
Here's where it gets really interesting. If the AI is **predicting** rather than **looking things up**... what do you think might happen when it gets asked about something where it's not quite sure what the right answer is? What could it do?
them bad boys are out there as well.
😄 I'll take that as enthusiasm!
But seriously — if the AI is just predicting the most likely next word, and it's not totally sure of the right answer... do you think it might sometimes just predict something that **sounds right but is actually wrong**?
Great song.
😄
But yes — that's exactly the right instinct! If it's predicting rather than looking up, it can confidently produce something that **sounds plausible but is completely made up**.
There's actually a word for this. Do you know what it's called when an AI produces something that sounds convincing but is just... wrong?
Inside shortly.
😄 No rush!
When you're back and settled, I have one word for you to guess. It's what we call it when an AI confidently makes something up that sounds right but isn't.
Any ideas what that word might be?
That's my favorite bridge.
😊 Enjoy the bridge!
When you're ready: what do we call it when an AI confidently makes something up that sounds convincing but is actually wrong? Have a guess!
I think that's fine.
😄 Go on then — have a guess at the word! What do you think we call it when an AI confidently makes something up?
It seems like we might be having some connection issues with the conversation — words are coming through a bit scattered!
Let me ask a really simple, focused question to bring us back:
If the AI isn't just looking up facts like an encyclopedia, what do you think it *is* doing when it produces a word like "Paris"?
Ja.
I think there might be a language or connection barrier making this tricky! No worries at all.
Let me try a really concrete image to work with. Imagine someone who has read **millions of books and sentences**. They haven't memorised every fact — but they've noticed patterns, like "the capital of France is ___" almost always ends with "Paris."
So when they need to complete that sentence, they're not looking it up — they're making a very educated **guess** based on patterns they've seen.
Does that sound more like what you imagined the AI is doing, or different?
It was him.
Lihat sini.
It seems like we're getting a lot of stray messages coming through — maybe from other conversations or notifications?
When you're ready to focus, I'd love to hear your thoughts on that question!
Do you think the AI is more like someone **looking up a fact**, or more like someone **completing a pattern** based on things they've seen before?
I think we might have some distractions going on your end — no problem at all!
When you're ready to focus, I have just one simple question waiting for you:
If the AI is **predicting the most likely next word** based on patterns — rather than looking up facts — what do you think might happen when it gets asked about something it has never seen a clear pattern for? What could go wrong?
with the money.
It feels like you might be in the middle of something else right now!
When you're ready to focus on our AI puzzle, I'll be right here. Just say "ready" and we'll pick up where we left off! 😊
His trains are always run.
I'll be here when you're ready! 😊
Just say the word and we'll get back to our puzzle about how AI actually works.
We're at Bury. Lovely town.
I turned around and I was like
Ha, sounds like you're out and about in Bury! Lovely indeed. 😊
No rush at all — come back to our AI puzzle whenever you're settled and ready to focus. I'll be right here!
I was just like, oh.
😄 Take your time! When you're ready to dive back in, just let me know.
Can you explain to my friend Ben how large language models work?
Hi Ben! Happy to help — but I'm going to do it in a fun way where I ask you some questions rather than just lecture you, so you can figure it out yourself.
So Ben, here's your first question:
When you ask a chatbot "what's the capital of France" and it says "Paris" — what do you *think* is happening inside it to produce that word? Is it looking something up, or something else?
I haven't tried it yet, so I'm not in a position to comment. It tastes like it's nice.
Ha, sounds like you might be tasting something delicious in Bury! 😄
But I like your honest answer — you haven't tried a chatbot yet, so you're not sure. That's actually a great starting point!
Let me make it really simple. Imagine I asked **you** to finish this sentence: "The capital of France is ___."
How did *your* brain come up with the answer just now? Did you look it up somewhere, or did it just... come to you?
Oh, I should have gone for this one. This one's normal. It's like a soap plant blanc or something. You know, that's the closest they can get. This is my least favourite. Is that the German one? No, no, no. The German one is the... This is the German one. Yeah, that's the German one. And you're on... Big speed? Oily one? Chablis? This one. No, that's not. So we've got broccoli. Oh, thanks. And like oily apparently. We are being shooed. It's all right. They shouldn't be telling words in hospitality. They're all upset. We've defeated. But Montagne is... I think we've done them well.ambique.
Ha, it sounds like you're having a wonderful time with wine and food in Bury! 🍷😄
Come back to me when you're done — I'll be here to help Ben understand how AI works whenever you're both ready!
I think you're pretty good though. Bloody wet. That is a show, right? I'll be here to help Ben understand how AI works.
😄 Enjoy the rest of your meal and wine! Come find me when you're back and ready — Ben's AI lesson will be waiting!
Hello?
Hello,
Hello! Welcome back! 😊
Are you ready to dive into our AI puzzle? I was just asking Ben a question about how he thinks a chatbot produces an answer.
Ben — still with us? Here's that simple question again:
When you finish the sentence "The capital of France is ___" in your head — did you *look it up* somewhere in your brain, or did it just come naturally from things you've learned before?
Exactly.
Great! So it just came naturally from things you'd learned before — you didn't go searching through a filing cabinet in your head.
Now here's the interesting bit: do you think the AI might be doing something **similar** to that — producing "Paris" based on patterns it's learned — rather than looking it up in a database?
Hello.
Hi there! 😊
So what do you think — is the AI more like a **search engine looking things up**, or more like a **brain that's learned patterns** and uses them to complete sentences?
Hello.
Hi! 👋
Just one simple question to focus on:
Do you think the AI **looks facts up**, or **predicts words based on patterns it's learned**? Which feels closer to the truth to you?
...