🧠 How AI Works

Watch a neural network learn to recognize patterns — step by step

🔬 Step 1: Meet the Neural Network

A neural network is a computer program inspired by the human brain. It has layers of connected "neurons" that pass information to each other. Watch the network below — each dot is a neuron, and each line is a connection.

Input Layer (left): This is where data enters the network — like pixels of an image, words in a sentence, or numbers.
Hidden Layers (middle): These layers find patterns in the data. Each neuron looks for a specific feature. With more layers, the network can find more complex patterns.
Output Layer (right): This is the network's answer — its prediction or classification. For example: "This is a cat" or "This email is spam."

📊 Step 2: Train the AI with Data

AI learns from examples, just like you learn from practice. Click the items below to "feed" them to the neural network. The AI will try to learn which are fruits and which are NOT fruits.

Select at least 4 items to start training

🎯 AI Knowledge Quiz

Test what you learned about how AI works!

Understanding Neural Networks

Neural networks are the foundation of most modern artificial intelligence. They are computing systems loosely inspired by the biological neural networks in animal brains. A neural network consists of layers of interconnected nodes, or "neurons," that process information. Data enters through the input layer, passes through one or more hidden layers where patterns are detected, and produces a result in the output layer.

The "learning" in machine learning happens when the network adjusts the strength of connections between neurons. These connection strengths, called weights, start out random. When the network makes a wrong prediction, the weights are adjusted slightly so the network will do better next time. This process is repeated thousands or millions of times with different training examples until the network becomes accurate. This is called backpropagation, and it is the core algorithm behind deep learning.

How AI Learns from Data

The quality and quantity of training data is critical to how well an AI system works. If a neural network is trained on many diverse examples, it will generalize well and make accurate predictions on new data it has never seen. If it is trained on too few examples or biased data, it will make mistakes. This is why data scientists spend a large portion of their time collecting, cleaning, and organizing data before training begins.

There are three main types of machine learning. In supervised learning, the AI is given labeled examples (like photos tagged "cat" or "dog") and learns to predict the correct label. In unsupervised learning, the AI finds patterns in unlabeled data without being told what to look for. In reinforcement learning, the AI learns through trial and error, receiving rewards for correct actions — this is how AI learns to play games like chess and Go.

Real-World Applications

Neural networks power many technologies that children and adults use every day. Smartphone cameras use neural networks for face detection, portrait mode, and scene recognition. Language translation apps like Google Translate use neural networks to convert text between languages. Self-driving cars use neural networks to recognize pedestrians, traffic signs, and road markings. Medical AI uses neural networks to detect diseases in X-rays and MRI scans, sometimes with accuracy matching human doctors.

Limitations of AI

Despite impressive capabilities, current AI systems have significant limitations. They cannot truly "understand" information the way humans do — they recognize patterns without comprehension. They can be fooled by adversarial examples (inputs specifically designed to trick them). They require enormous amounts of data and computing power to train. And they can perpetuate or amplify biases present in their training data. Understanding these limitations is a key part of AI literacy.

Last reviewed: April 2026 · Aligned with AI4K12 Big Ideas 1 and 3