Neural Network Laboratory
Client-side cinematic ML engineering. Train neural networks directly in your browser, visualize activation flow, and explore Convolutional layers.
How did AI get inspired by the human brain?
Neural Networks are mathematically designed to mimic the way biological neurons process information. Let's break down the connection.
1. The Neuron
From biological chemistry to mathematical weights.
Biological Neuron
Receives chemical signals through dendrites. If the total charge passes a threshold, it fires an electrical impulse down the axon.
Artificial Neuron
Receives numerical inputs. Multiplies them by weights (synaptic strength). If the sum passes an activation function, it outputs a number.
Interactive
y = 0.5597
Try it Yourself: Train a Neuron
You are a single Artificial Neuron. Your job is to be a "T" shape detector. Adjust the 9 pixel weights below so that your neuron FIRES (>85%) when shown a 'T', but STAYS QUIET (<15%)when shown an 'L'.
2. The Visual Cortex (CNNs)
Convolutional Neural Networks (CNNs) were directly inspired by the cat visual cortex experiment (Hubel & Wiesel, 1959). They discovered that specific neurons only fire when they see specific edges or angles.
The Scale of Complexity
Neural networks are inspired by the brain, but the difference in scale is astronomical.
Early Vision Network
1,000 Neurons
Early Vision Network: Can recognize handwritten digits (MNIST).
Brain vs Neural Network: The Truth
Inspired by biology Artificial neurons are mathematically inspired by the firing thresholds of biological neurons.
Synaptic weights Learning in an AI adjusts "weights", which loosely resembles how biological synapses grow stronger or weaker.
AI does not "think" Neural networks are just highly optimized matrix multiplications. They do not possess consciousness or reasoning.
Vastly simpler A real neuron is a complex chemical machine with its own DNA, proteins, and non-linear dynamics. An AI neuron is just `y = activation(Wx + b)`.