# Generative AI Generative AI refers to systems that *produce* new content — text, images, audio, video, code — rather than merely classifying or predicting. Where a discriminative model answers "is this a cat?", a generative model answers "draw me a cat." The current wave is built on a few key techniques: [[Large Language Model]]s for text (and now code and reasoning), Diffusion Models for images and video (Midjourney, DALL-E, Stable Diffusion, Sora), and earlier Generative Adversarial Networks. The 2022 arrival of ChatGPT and high-quality image generators made generative AI the public face of the entire field and triggered both an investment frenzy and a cultural reckoning. ## Why it matters here Generative AI is arguably the AI subfield with the most direct and immediate political consequences in this vault, because it industrializes the production of persuasive media. **Deepfakes and disinformation.** Generative models can fabricate convincing fake images, audio, and video of real people — [[Deep Fakes]] — and unlimited fluent text. This is a force multiplier for [[Disinformation]] and [[Information Warfare]]: fabricated candidate "recordings," synthetic crowds, AI-generated propaganda at scale, and the erosion of audio/video as trustworthy evidence. The deeper danger is the **liar's dividend** — once anything *can* be fake, genuine evidence can be dismissed as fabricated, a gift to bad actors. Non-consensual sexual deepfakes are a documented, gendered harm. **Creative labor and copyright.** Generative models are trained on vast scraped corpora of human writing, art, and music, raising live copyright fights (the NYT v. OpenAI suit, artist lawsuits) and existential anxiety in creative industries — directly relevant to the content-creation economy. The fights over training-data consent and compensation are unresolved. **The information commons.** A flood of cheap synthetic content ("AI slop") degrades search, social media, and the open web — the substrate of public discourse. Evenhandedly: generative AI is also a genuine creative and productivity tool, and much of its use is benign. What makes it politically central is the asymmetry — it lowers the cost of producing deception far faster than it lowers the cost of detecting it. ## See Also - [[Large Language Model]] — generative text - Diffusion Model · Generative Adversarial Network — generative images - [[Deep Fakes]] — the synthetic-media weapon - [[Disinformation]] · [[Information Warfare]] — the abuse surface