In generative AI, a generator is the component responsible for producing new content, whether text, images, audio, or code, based on patterns it learned from training data rather than by retrieving or copying existing examples. ChatGPT and other large language models act as generators for text: given a prompt, they predict and assemble a sequence of words that forms a coherent, original response.
The term carries more specific weight in certain architectures. In a Generative Adversarial Network (GAN), the generator is one half of a two-network system, paired with a discriminator that tries to tell generated content apart from real examples; the generator improves by repeatedly trying to fool the discriminator, and the discriminator improves by getting harder to fool, a competitive setup that pushes both networks toward higher-quality output over training.
