Overview
AI systems look like magic until you separate three layers: data, model, and interface. This lab keeps vocabulary plain — enough to use tools wisely, not enough to pretend you train GPUs.
Foundation
18 min
3 checks
How AI Actually Works (No Hype)
Foundations module — models, tokens, and limits in plain English.
A “token” is closest to…
Check answerLog in to lock in your answer and track score.
Models, briefly
A model is a giant pattern machine. It predicts the next useful chunk of text from what you (and the system) fed it. That prediction feels like understanding because fluent language is the output format — not because it “knows” your assignment the way a tutor does.
Why do long chat histories get expensive or hit limits?
Check answerLog in to lock in your answer and track score.
Tokens and limits
Text is chopped into tokens (words or pieces of words). Context windows, pricing, and many errors track tokens. Long chat histories raise cost and truncation risk: every turn often re-sends older messages.
Best first move when an answer looks confidently wrong?
Check answerLog in to lock in your answer and track score.
Limits that matter
- Fluency ≠ correctness
- Training knowledge has a cutoff; tools/search may not
- Confidential data does not belong in random chats
Read each beat, answer the check that follows, then continue. When every check is done, mark the lab complete.
Lab wrap
Finish the checks to clear this lab
0/3
Correct
0%
Lab score
1
2
3