Session 4: The Brain of AI
Your brain has about 86 billion neurons - cells connected together. When you learn something, the connections between neurons get stronger.
Key idea: Learning is about patterns. Both brains and AI learn patterns.
Artificial Neural Networks copy how brains work. They have "neurons" connected by "synapses" (connections) with adjustable "weights" (connection strengths).
Information travels through the network like electricity through circuits.
Example: Recognizing a cat → pixels → edges → shapes → cat features → answer "cat!"
The "strength" of connections between neurons are called weights. These change as the AI learns.
Think: Like tuning a guitar by turning pegs slightly, over and over
Each neuron decides whether to "fire" (activate) based on its inputs. This is done with activation functions.
Why it matters: Without activation, neural networks would just be math (no learning power)
Classic example: Can a neural network recognize handwritten numbers 0-9?
Backpropagation is how neural networks learn. It's like tracing backwards to find which weights were "wrong" and need adjustment.
Result: After thousands of examples, network learns!
Modern AI uses deep neural networks with many hidden layers. More layers = more complex patterns!
Sometimes networks memorize training data instead of learning general patterns. This is called overfitting.
Now you understand how AI learns like a brain
Next Session: Types of Machine Learning