Neuronales Netz

A Neural Network is an AI model inspired by the human brain: artificial neurons are arranged in layers and connected via weighted connections. Through training on data, the network learns to recognize patterns — from simple image classification to complex language generation. Neural networks are the foundation of virtually all modern AI.

Key architectures: Feedforward networks (simplest form), Convolutional Neural Networks/CNNs (for Computer Vision), Recurrent Neural Networks/RNNs (for sequences, largely replaced by Transformers), and Transformers (the architecture behind LLMs and modern image generators). Deep Learning refers to neural networks with many layers.

Training works through backpropagation: the network makes a prediction, the error is calculated, and weights are adjusted — millions of times until the network masters the task. Modern LLMs like GPT-5.2 have hundreds of billions of parameters (weights) trained on trillions of tokens.

For practitioners: you don't need to build a neural network yourself — APIs from OpenAI, Anthropic, and Google make the technology accessible via API calls. But understanding the basics helps you better assess the strengths and limitations of AI tools and make informed decisions.