🚀 Zero → Master Learning Experience

AI · ML · DL · GenAI
Explained

From "what even is AI?" to deep technical mastery — interactive, animated, made for every learning level.

🤖 Artificial Intelligence

📊 Machine Learning

🧠 Deep Learning

✨ Generative AI

Your Progress0% Complete

🧸 Level 0 — Intuition

What Are These Things?

Let's start with a story. No jargon — just pure understanding with real-life analogies.

🤖

ARTIFICIAL INTELLIGENCE

The Smart Friend

Imagine a friend who can do almost anything a human can do — play chess, recognize your face, drive a car, even write stories. That's AI! It's the big umbrella idea: making computers act smart.

📊

MACHINE LEARNING

Learning from Examples

Instead of programming every rule, you show the computer thousands of examples and it figures out the rules itself. Like teaching a child — show them 100 cats and dogs and they learn to tell them apart!

🧠

DEEP LEARNING

The Brain-Inspired Turbo

Deep Learning is a type of ML that copies how your brain works — with billions of tiny connected switches (neurons). It's why AI can recognize faces, understand speech, and generate images.

✨

GENERATIVE AI

The Creative Superpower

GenAI doesn't just classify things — it creates brand-new things! Text, images, code, music, videos. ChatGPT, DALL-E, Midjourney — all powered by Generative AI.

🎨 Analogy:

GenAI is like a student who reads every book ever written, watches every movie, studies every painting — and then creates something completely new. It's not copying; it's learned the essence of human creativity.

🪆 How They Nest Inside Each Other

🤖 Artificial Intelligence (everything that makes machines smart)

📊 Machine Learning (learning from data)

🧠 Deep Learning (brain-like neural nets)

✨ Generative AI (creates new content)

👀 Level 1 — Basic Understanding

Visual Step-by-Step

See exactly how each concept works, with interactive examples you can click through.

How Machine Learning Actually Works

3 Types of Machine Learning

How a Neural Network Thinks

Click a neuron to fire it and watch the signal cascade through the network!

🎮 Level 1–2 — Interactive

Interactive Playground

Experiment with real AI concepts — no coding required.

🐱 Train a Mini Image Classifier

Add labeled examples, then hit "Train" to make an AI that recognizes things!

Add Training Examples

Test the Model

📉 Training a Model — Watch the Loss Drop

Drag the "Epochs" slider to simulate training. See how the model gets better over time!

Epochs 0

Start (untrained, random guesses)End (well-trained)

⚖️ Underfitting vs Overfitting

This is a critical concept. Adjust model complexity and see what happens!

Complexity Just Right

🧠 Level 2 — Intermediate

Guided Deep Dive

Real mechanics, actual math concepts (visualized!), and how everything connects.

The ML Formula

// The core idea of ML:
function train(data, labels) {
  // Start with random weights
  weights = random();

  for epoch in 1000 {
    // Make prediction
    pred = predict(data, weights);
    // Measure error
    loss = compare(pred, labels);
    // Adjust weights to reduce error
    weights = update(weights, loss);
  }
}

Key Concepts

Features

The inputs. E.g. pixel values in an image

Labels

The answers. E.g. "cat" or "dog"

Loss Function

Measures how wrong the model is

Gradient Descent

The math trick to reduce loss step by step

Hyperparameters

Settings you tune: learning rate, layers, etc.

🏔️ Gradient Descent — Rolling Down a Hill

Click "Step" to watch the model roll toward the minimum loss — like a ball rolling down a hill!

Learning rate: 0.3

Layers in a Neural Network

Input Layer

Raw data comes in — pixels, numbers, words

Hidden Layers

Where learning happens. Each layer learns patterns (edges → shapes → objects)

Output Layer

Final answer — a class, a number, a word

Activation Function

Decides if a neuron fires (ReLU, Sigmoid, etc.)

Backpropagation

Error signal travels backward to fix weights

Architecture Families

CNN

Convolutional Neural Net — great for images. Slides filters across pixels to detect patterns.

RNN / LSTM

Recurrent — great for sequences, text, time series. Has "memory" of previous tokens.

Transformer

Self-attention mechanism. Powers GPT, BERT, DALL-E. Processes all tokens in parallel.

GAN

Two networks compete. Generator creates fakes, Discriminator spots fakes. Both get better.

🔢 What Are "Weights" Really?

Each connection between neurons has a weight — a number that says how important that signal is. Drag to change a weight and see the output change!

Weighted sum = 0.0 → Output: 0.5

How GPT-style Models Work

1. Pre-training on massive text

Trained on billions of web pages, books, code. Learns language patterns.

2. Tokenization

Text is split into tokens (roughly word-pieces). "Hello" → [15496]

3. Self-Attention

Each token "looks at" all other tokens to understand context. This is the transformer's superpower.

4. Next-token prediction

Predicts the most likely next word, one at a time. "The cat sat on the ___" → "mat"

5. RLHF Fine-tuning

Human feedback makes the model helpful, harmless, honest.

🔤 See Tokenization Live

Type something and watch your text get split into tokens (like how GPT sees it)

0 tokens

Property	🤖 AI	📊 ML	🧠 DL	✨ GenAI

⚙️ Level 3 — Advanced

Advanced Concepts

Trade-offs, architectures, attention mechanisms, and the real engineering challenges.

🔍 Attention — How Transformers Read

Click any word. See which other words it "attends to" (pays attention to) when deciding its meaning.

🏗 Architecture Trade-offs

⚖️ Bias & Fairness in AI

AI inherits the biases in its training data. See what happens when training data is imbalanced.

Training Data Distribution

Group A80%

Group B20%

🏗 Level 3 — Applications

Real-World AI/ML/DL/GenAI

Where does each technology actually live in the real world?

🏢 AI Stack at Real Companies

🔮 Emerging Trends in 2024–2025

🧪 Level 4 — Mastery

Experiment Zone

Build your own mini neural network. Toggle switches, watch signals fire, break things on purpose!

🏗 Build Your Own Neural Network

Configure the architecture, then click "Run Forward Pass" to see data flow through!

Hidden Layers2

Neurons/Layer4

Activation

💥 "Try Breaking It" Scenarios

What goes wrong when you do these things? Click to find out!

❓ Quiz — All Levels

Test Your Knowledge

5 questions across Beginner → Expert. Get instant explanations for every answer.

🧗 Level 4 — Mastery Challenges

Master Challenges

Open-ended, thinking-required challenges. No multiple choice — you must reason!

🟢 Easy

The Sorting Hat Problem

You want to build an AI that categorizes emails as "Work", "Personal", or "Spam". Describe which TYPE of ML (supervised/unsupervised/reinforcement) you would use and WHY.

Hint: Think about: do you already have labeled examples of email categories?

🟡 Medium

The Black Box Problem

A hospital uses a Deep Learning model to diagnose cancer. The model is 95% accurate but no one understands HOW it makes decisions. What are the risks? What should be done?

🔴 Hard

The Architecture Design Challenge

You need to build an AI system that: (1) watches live CCTV video, (2) detects if someone falls, (3) immediately alerts hospital staff with a text description of what happened. Design the full AI pipeline — which models, which architectures, in which order?

🏁 Summary & Key Takeaways

You Made It! Here's Everything.

Your complete AI/ML/DL/GenAI reference — all on one page.

📋 The Ultimate Cheatsheet

🧠 5 Mental Models to Never Forget

🎓

You're Now an AI Expert!

You went from zero to master. The AI revolution needs people who understand it deeply.

AI · ML · DL · GenAIExplained