AI Fundamentals
What exactly IS artificial intelligence? Let's build your mental model from the ground up — no jargon, no math.
1. The Big Picture
Let's kill the myth right away: AI is not magic, and it's not sentient. Artificial intelligence is software — specifically, software that makes decisions or predictions by finding patterns in data. That's it.
Think of a spam filter. It reads your email, notices certain patterns ("Buy now!", "Click here!", suspicious sender address), and decides: spam or not spam. That's AI. A very simple one, but still AI. Now scale that idea up enormously:
This is why AI can seem both incredibly impressive and surprisingly dumb. It's very good at pattern matching when given lots of relevant data, and very bad when asked to reason about things outside those patterns. There's no understanding happening — just very fast, very complex pattern recognition.
2. The Hierarchy: AI ⊃ ML ⊃ DL
You'll hear these three terms used interchangeably, but they're not the same thing. They're nested: each one is a subset of the previous.
The broadest term. Any software that appears "intelligent." A chess engine from 1997 is AI. A spam filter is AI. ChatGPT is AI. All use different techniques, but the goal is the same: make a computer behave intelligently.
Instead of writing explicit rules ("if the email contains 'buy now' AND sender is unknown, mark as spam"), you give the system thousands of examples and it figures out the rules itself. The more data, the better it gets.
A specific type of ML using artificial neural networks with many layers (hence "deep"). Each layer learns more abstract representations. This is what powers image recognition, voice assistants, and LLMs.
3. Interactive Glossary
Search any AI term below. Use the category buttons to filter by topic. Click a term to expand its full definition.
4. Visual Mindmap
This interactive mindmap shows how all the major AI concepts relate to each other. Zoom in to explore — the full AI landscape in one diagram.
mindmap
root((AI))
Machine Learning
Supervised Learning
Classification
Regression
Unsupervised Learning
Clustering
Dimensionality Reduction
Reinforcement Learning
Reward Functions
Policy
Deep Learning
Neural Networks
Layers
Weights and Biases
Activation Functions
Transformers
Attention
LLMs
GPT
Claude
Gemini
Applications
NLP
Chatbots
Translation
Summarization
Computer Vision
Image Recognition
Object Detection
Recommendation
Key Concepts
Training
Dataset
Loss Function
Backpropagation
Inference
Temperature
Context Window
Prompting
Zero-Shot
Few-Shot
Chain-of-Thought
5. How Models Learn
Modern AI models learn in three distinct phases. Understanding these phases demystifies most of the jargon you'll encounter.
Data Collection
Before a model can learn anything, it needs data — lots of it. For an LLM like Claude, this means hundreds of billions of words from books, websites, and code. The quality and diversity of this data directly shapes what the model knows and how it thinks.
Training
The model is shown data over and over. Each time it makes a prediction (e.g., "what word comes next?"), it compares its guess to the correct answer, calculates how wrong it was (the loss), and adjusts its billions of internal parameters slightly via backpropagation. Repeat millions of times.
Inference
Once trained, the model's parameters are frozen. When you send a prompt, the model runs your input through all its layers (a "forward pass") and generates a response token by token. The temperature setting controls how random/creative vs. deterministic/focused the outputs are.
6. Quick Knowledge Check
Three questions to test your understanding. Click an answer to see instant feedback.
7. What's Next
You've got the foundation. The next posts build on these concepts — each one goes deeper into a specific area.