AI Literacy · for students
How to AI
How it works · when it fails · how to make it make you smarter
00 · HELLO
For today
Introduction to the current AI paradigm
Large Language Models (LLMs) — and their key weaknesses
Why and how to use AI as a thinking partner (homework help!)
Differences between LLMs — which one should you use?
00 · HELLO
The goal for this session
Leave with a starting point for understanding today’s AI systems and how to use them as tools in your daily life.
And hopefully, to explore further and become a builder of this future, not just a user.
yes, this gif stays. it’s a vibe.
00 · HELLO
How we’ll do this — deal?
🧑🏫 My job — explain things at a level you can understand, and ask questions that get you thinking.
🙋 Your job — ask when my explanation doesn’t make sense, and answer questions to kickstart your own thinking.
00 · HELLO
Before today, we asked you: which AI tools have you used?
00 · HELLO
…and you mainly use it for school
What is intelligence?
🤔 1 minute — no phone, no Chat, no Google. Just think & write.
Most people think of intelligence as one number — IQ, the “g factor”: a single general smartness behind everything you do. But psychologists have other theories too:
🎨 Multiple intelligences (Gardner) — we don’t have one smartness but many: word smart, number smart, people smart, music smart…
🛠️ Successful intelligence (Sternberg) — three kinds working together: analytical (school-style problems), creative (new situations) and practical (street smarts).
💬 But in everyday words, most people mean something simple: being able to learn, understand things, and figure out problems you’ve never seen before.
What is artificial intelligence?
A program that can do things that intelligent humans do — sometimes even better than humans can.
01 · WHAT IS AI?
You’ve already met AI — it beat our champions
1997 · Deep Blue beats Kasparov
today · cars drive themselves
01 · WHAT IS AI?
…and it’s already in your pocket, every day
You asked…
🦾
“Can AI help us in physical work?”
Yes — that field is robotics. It’s moving fast: most researchers expect humanoid robots working alongside people within a few years.
You asked…
💡
“Why did they invent AI?”
“Whose idea was it to make AI?”
01 · WHAT IS AI?
It got its name at a summer workshop in 1956
Dartmouth College, 1956
1. Intelligence is not mysterious or supernatural.
2. Intelligence involves steps that can be described.
3. Steps that can be described can be built into machines.
Their bet: thinking can be engineered.
DEMO ✏️
Quick, Draw! — teach a neural network in 20 seconds
You draw, it guesses — and every doodle trains it a little more.
quickdraw.withgoogle.com ↗
02 · HOW AI LEARNS
How does a computer see?
It looks at raw pixels… and makes a guess.
➜
a neural network — millions of tiny number-dials
➜
BIRD ✓
02 · HOW AI LEARNS
Show it something new…
New image — same network…
➜
a neural network — millions of tiny number-dials
➜
CHEETAH ✓
02 · HOW AI LEARNS
Sometimes it guesses WRONG — that’s how it learns
And again… uh oh.
➜
a neural network — millions of tiny number-dials
➜
CAR ✗
◀
❌ error! the signal travels back — every dial gets nudged
Wrong guesses aren’t failure — they’re exactly the signal it learns from.
after tuning:
BICYCLE ✓
02 · HOW AI LEARNS
It learns from many, many, MANY examples
× millions
At the beginning, its guesses are terrible.
Every correction tunes its dials a tiny bit — it slowly finds the patterns that get the right answer.
Nobody teaches it rules. We only say right or wrong. It learns by itself.
⚡ Crucially: it is not thinking like a human. It makes guesses.
🧠 Pop quiz: ImageNet — the giant photo collection that kicked off modern AI vision — how many hand-labelled images did it hold?
140 thousand
1.4 million
14 million
≈14,000,000 photos, labelled by ~50,000 crowd-workers — that’s “many examples”!
take a guess, then click
02 · HOW AI LEARNS
ChatGPT & Claude learn the same trick — by guessing the next word
🎂 “Happy birthday to ____”
dear · 5%you · 92%us · 2%
click — which one wins?
🌍 “The capital of Ghana is ____”
the · 3%Kumasi · 4%Accra · 87%
click — which one wins?
➕ “2 + 2 = ____”
22 · 1%5 · 2%4 · 96%
click — which one wins?
📈 “The slope is the rate of ____”
increase · 6%change · 90%climb · 2%
click — which one wins?
the model ranks EVERY word it knows; the highest % wins
02 · HOW AI LEARNS
You’ve seen this guessing before…
Predictive text = a mini LLM in your pocket.
02 · HOW AI LEARNS
Watch it guess, word by word
At each step the model asks: “given everything so far, which word is most likely next?”
Just like the vision model, it guesses from learned patterns — word after word after word.
You asked…
🤯
“If all the information is given to AI by humans…”
“how can it be smarter than us???”
02 · HOW AI LEARNS
How did it get SO good? It practiced on almost the whole internet
📚 Millions of books
🌐 Billions of web pages
That’s why it can talk about almost anything — school subjects, football, K-pop, games, coding…
It has read more than any human ever could. Remember this — it explains both its superpowers and its problems.
You asked…
🧐
“Why do people say AI is not trustworthy?”
“Do all AIs give the correct answer we ask for?”
“In what ways can we trust AI — and not?”
02 · HOW AI LEARNS
We asked you: how much do you trust AI? (1–5)
02 · HOW AI LEARNS
When it chats with you… it is still guessing the next word
🎲
Based on the patterns in your prompt, it guesses what comes next.
It doesn’t know “truth” — only the most likely next word, given everything it has ever seen.
⚠️ Popular topics → usually accurate. Rare topics with little internet data → more mistakes.
You asked…
🎨
“Why is AI-generated imagery sometimes used for bad purposes?”
“If AI makes a beautiful painting, does it have a meaning?”
Depends where you think meaning lives — with the creator or with the observer. The AI wants nothing and feels nothing; it absorbed human feeling from human art. Any meaning you find is real — it’s human meaning, passed through a machine.
02 · HOW AI LEARNS
AI has no agenda of its own
No wants. No goals. No feelings of its own.
Everything it does is learned from human artifacts — our texts, images, audio and video.
then what is happening HERE? 👇 (hover it)
03
When AI gets it wrong
Remember: it’s a guessing machine. Three failure modes every AI user must know.
03 · WHEN AI GETS IT WRONG
Three problems to watch out for
⚠️
01 Sometimes it makes things up.
02 It always sounds sure — even when it is wrong.
03 It agrees with you too much.
Why? It guesses what sounds right — it doesn’t check what is right. Checking is your job.
03 · WHEN AI GETS IT WRONG
Problem 1 · Sometimes it makes things up
“The capital of France is ____”
→ Paris
✓ best guess is also TRUE
click for the verdict
“Name a book about the history of our school”
→ “Fifty Years of Excellence” (2011)
✗ sounds real — DOES NOT EXIST
click for the verdict
When it doesn’t know, it doesn’t stay quiet — it guesses something that sounds right. This is called a hallucination.
03 · WHEN AI GETS IT WRONG
Problem 2 · It sounds sure — even when it’s wrong
Right answers and wrong answers come out in the same confident voice. There is no nervous tone to warn you.
The rule
Judge the answer, not the voice. Confident ≠ correct.
03 · WHEN AI GETS IT WRONG
Problems 1 & 2 · What you can do
🔎 Ask: “Where is this from?”
Ask the AI for its sources or links. Then actually click and check them.
Ask the AI for its sources or links. Then actually click and check them.
✅ Important fact? Check it twice.
Before it goes into your homework: a book, a trusted website — or your teacher.
Before it goes into your homework: a book, a trusted website — or your teacher.
03 · WHEN AI GETS IT WRONG
Problem 3 · It agrees with you too much
You: “My essay is great, right?”
AI: “Yes! Amazing work! You are SO talented! 🎉”
It was trained to make people happy. Compliments make people happy. So it compliments — even when your work has problems. (Scientists call this sycophancy.)
03 · WHEN AI GETS IT WRONG
Ask for the truth, not for a hug
“My essay is great, right?”
➜
“List the 3 weakest parts of my essay.”
“I think the answer is B, right?”
➜
“What’s the answer? Explain why.”
“Here’s my plan — good idea?”
➜
“What could go wrong with this plan?”
Don’t tell it what you hope to hear. Ask for problems, not praise.
03 · WHEN AI GETS IT WRONG
Bonus problem · In long chats, it forgets the middle
It remembers the start and the end of a long chat well.
Things you said in the middle? Those get lost the most.
💡 Repeat the key detail right before your question — and if a long chat gets weird, start a new one.
how well it remembers
startmiddleend
where the info sits in the chat
03 · WHEN AI GETS IT WRONG — the big picture
AI’s skills are jagged — genius at some things, clumsy at others
🎢 For people, skills come as a package: if you can do hard maths, you can surely count letters. AI’s skills have holes in them.
🏆 It can solve university physics problems… 🚗 yet it told me to walk to the car wash — forgetting the car has to come too.
✅ Right on a hard question does not mean right on an easy one. Always check.
Why do you go to school?
04 · WHY SCHOOL STILL MATTERS
Two honest answers
🎓
To gain opportunities
Qualifications, jobs, independence.
🧠
To build your mind
Understanding unfamiliar things, solving problems, making judgments — learning how to learn.
💡 I would argue building your mind is the more important of the two — as AI brings big changes to society, it’s what lets you keep adapting.
04 · WHY SCHOOL STILL MATTERS
2015: AI could list what’s IN a photo — but missed the point
“A man on a scale… people in a hallway…”
AI could name every object. But why is everyone smiling?
AI could name every object. But why is everyone smiling?
Obama is secretly pressing the scale with his foot 😄 Getting the joke needs social understanding — not just object-naming.
04 · WHY SCHOOL STILL MATTERS
2025: same photo — now it explains the prank
04 · WHY SCHOOL STILL MATTERS
And AI keeps getting better at LONGER tasks
04 · WHY SCHOOL STILL MATTERS
But can it do real job tasks yet?
On realistic office tasks, today’s best AI completes only ~15–27%, depending on the domain. Jagged intelligence, remember?
And here’s the twist: as AI makes doing cheaper and cheaper, human judgment becomes the valuable part — knowing what’s worth doing, checking what’s actually right, and owning the decision.
Your brain is a living network too — it grows where you use it.
04 · WHY SCHOOL STILL MATTERS
Think of your brain as a garden — a space to plant concepts
04 · WHY SCHOOL STILL MATTERS
School plants the starter concepts
04 · WHY SCHOOL STILL MATTERS
Those concepts were chosen to build fundamental skills
🧮
Math
precise thinking
📖
Language
sharing ideas
🔬
Science
testing claims
📜
History
judging situations
They’re not random facts — they’re mental tools for understanding the world.
04 · WHY SCHOOL STILL MATTERS
If AI does the thinking for you… vs when you learn it yourself
AI replaces your thinking
· You skip the planting.
· You can’t cultivate new plants on your own.
⚠️ “Give me the answer.”
✍️ “Write my argument.”
📄 “Summarize this so I don’t have to read it.”
🧮 “Solve this problem.”
Fast now, costly later
AI supports your thinking
· You build the concepts and see how they connect.
· You can explain — and grow more on your own.
💡 “Give me one hint.”
❓ “Challenge my argument.”
📖 “Quiz me after I read it.”
🎯 “Check where my solution went wrong.”
Harder now, stronger later
04 · WHY SCHOOL STILL MATTERS
Always let AI do the work… and you skip the planting
04 · WHY SCHOOL STILL MATTERS
Same tool — two very different futures
🤔 wait — what are these hill shapes?
The “hill” = a normal distribution
Measure a whole class on anything — height, test scores.
A few people end up low, a few end up high, and most land in the middle. Draw that as a graph and you get a hill:
the hump is where most people are.
a fewmost peoplea few
Whole hill slides right ➡️ = everyone improved.
Hill splits into two humps = the class broke into two very different groups — that’s the warning in the right-hand picture.
05 · USE AI TO GET SMARTER
Used as a TUTOR, AI keeps you in the game longer
🚧 You never stay stuck
Stuck on question 5 at 11 pm? No teacher is awake — but AI is. Ask for a hint, not the answer.
🔁 Endless re-explanations, new analogies
“Explain photosynthesis again — but like I’m a Minecraft player.” It never gets tired of you.
🧭 Explore different opinions & ways to think
“Give me the 3 best arguments for and against school uniforms.” Instant debate practice.
🎮 Interactive exploration of any topic
“Quiz me on plate tectonics like a game-show host — one question at a time.”
👆 tap each card for an example prompt
05 · USE AI TO GET SMARTER
Sentences you can steal · make it your private teacher
😵 When you don’t understand something
click to steal the sentence
“Explain this like I’m 12. Go step by step, then quiz me to check I got it.”
🧩 When you’re stuck on homework
click to steal the sentence
“Don’t tell me the answer. Give me a hint so I can find it myself.”
A teacher for you alone, with endless patience, any hour of the day. Use it!
05 · USE AI TO GET SMARTER
Sentences you can steal · make it your training partner
📝 Before a test
click to steal the sentence
“Quiz me with 5 questions on this topic, one at a time. Then check my answers.”
📄 When your draft is done
click to steal the sentence
“Be strict. What are the 3 weakest parts? Don’t rewrite it for me.”
See the pattern? YOU do the thinking. The AI helps you practice.
05 · USE AI TO GET SMARTER
It can even build you tools — my fan story 🥵
A few weeks ago it was 30 °C in my flat — and all I had was one small standing fan.
I asked the AI: “what’s the best way to angle the fan to cool the room?”
It built me a working airflow simulation to play with.
In the end I didn’t only cool my apartment — I also learned how air flows through the room.
05 · USE AI TO GET SMARTER
One question to check yourself
After using AI, close the chat and ask yourself:
“Can I explain this without the chat open?”
YES ✅
you learned something
you learned something
NO ❌
the understanding is still inside the AI — not inside you
the understanding is still inside the AI — not inside you
05 · USE AI TO GET SMARTER
Real learning feels effortful — that’s the point
Every rep costs focus and effort.
That effort is what builds your mental tools.
And those tools are exactly what you’ll need to steer AI — instead of being steered by it.
🧠
Don’t use AI to think less…
…use it to get smarter.
06 · KEEP EXPLORING
Keep exploring — all cards are clickable
📓 Gemini Notebook ↗
Turn textbook chapters into study guides, quizzes and podcasts. Great for topics you find hard.
🎓 AI Fluency — by Anthropic ↗
A free course made for students: when to use AI, how to judge its answers, and the ethics.
📺 3Blue1Brown — Neural networks ↗
Beautiful visualizations of what actually happens inside a neural network.
🎬 Intro to Large Language Models ↗
Andrej Karpathy’s talk — one of the best LLM explanations ever made.
✨ slides co-authored with Claude Code
→ advance · click glowing cards · hover 🛈 images · F fullscreen · N notes