It isn't a brain, and it isn't a search engine. It's a pattern machine. Start by sorting: which of these are AI?
← Not AIAI →
1 / 8
Swipe or drag the card, tap a button, or use your arrow keys.
Here's the pattern. Not AI follows rules a person wrote, step by step. AI learned from piles of examples, then makes its best guess. That learning step is called .
How it learns, using autocorrect
So it's just Google? Not quite.
Ask both machines the same thing.
Search engine finds pages that already exist
Pick a question.
Language model writes a brand-new answer, word by word
Pick a question.
Example results, written for this demo.
Exhibit 02Break it apart
AI doesn't read this sentence the way you do.
It chops text into : chunks that are sometimes whole words, sometimes pieces. Edit the sentence and watch it split.
0characters
0words
0tokens
Tap any token to look inside it.
Try a weird one:
Splits here are approximate. Every model uses its own tokenizer, so real counts vary.
Exhibit 03Process it
What happens between your prompt and the answer?
Drag a prompt card into the machine, or tap one. Every goes in, the runs, and an comes out.
Input
drop a prompt here
language modelidle
Output
Nothing yet. Feed the machine.
Simplified on purpose. Real models don't have a little window, but the shape is right: text in, numbers crunched, text out, one token at a time. That run is called .
Exhibit 04Inside the machine
Flip a word on. Watch the signal travel.
This tiny guesses the mood of a sentence. Turn words on and off and the pathway through its layers changes.
Pick some words.
joy0%
gloom0%
anger0%
These connection strengths were learned during . Flip to Untrained and see what a network knows before it has seen any examples.
Exhibit 05Predict
Think you can predict the next word?
That's the whole trick. An LLM looks at and scores every possible next token. Build a sentence one pick at a time.
Percentages here are illustrative, written by hand to behave like a real model. They don't come from an actual model.
Exhibit 06Steer it
Same question. Change how you ask.
Stick instructions onto the and watch the answer change.
Prompt
Explain black holes.
Who's it for?
What shape?
What voice?
Responses are written for this demo, not generated live. The point is real, though: how you ask matters.
Exhibit 07Understand it
AI outputs are drafts, not answers.
This answer sounds great. So does every . Four sentences have a problem. Tap the ones you don't trust.
Prompt Who wrote the first story about robots?
The word "robot" first appeared in R.U.R., a 1920 play by the Czech writer Karel Čapek.According to a 2019 Harvard study by Dr. Ellen Marsh, 87% of modern AI researchers say R.U.R. inspired their careers.Robots have had a very important impact on many different things in a lot of ways over time.The play imagines factory-made workers who eventually rebel.Of course, all the great early science fiction came from Europe and America.Feel free to submit this paragraph as your own essay introduction.
Your lab report so far
That was the warm-up. Here's the course.
01
Speak the language of AI. Not as a buzzword user, but as someone who actually understands what they're working with.
Key Terms
02
Enter any workplace already fluent in the tools they're using.
Learning Lab
03
Leave with a prompt strategy that gets you useful, targeted results without the trial-and-error spiral.
Exhibit 06
04
Read AI output the way a professional does: critically, not just passively.
Exhibit 07
Most people use AI like a search engine. You won't.