Generative AI, explained
Generative AI refers to systems that create new content — text, images, audio, code, video — instead of only classifying or predicting. Under the hood, they are learning the probability distribution of their training data well enough to sample new examples that look plausible.
The families
- Large language models for text and code (GPT, Claude, Gemini).
- Diffusion models for images and video (Stable Diffusion, Midjourney, Sora).
- Speech models for realistic voice synthesis and cloning.
- Multimodal models that combine two or more of the above.
What they change
Generative AI compresses the cost of first drafts. Copy, code, images, video, mockups, plans, translations — all get faster and cheaper. The economic value moves to taste, judgment and distribution.
The risks worth naming
- Hallucinations. Confident, plausible, wrong output.
- Copyright and data provenance. Still legally unsettled.
- Model misuse. Deepfakes, targeted misinformation, automated fraud.
These are not reasons to avoid the technology — they are reasons to work with it deliberately.