Generative AI
Models that generate new text, images, or other content.
CurrentbeginnerBeginner-friendlyFull course available
Overview
Generative AI covers models that produce new content (text, images, audio, code) rather than only classifying or scoring existing input. Large language models are the generative AI systems this platform focuses on -- generating text token by token, conditioned on a prompt.
- What it is
- AI systems that generate new content, most commonly text (LLMs), rather than only classifying existing input.
- Why it's used
- It enables open-ended tasks -- drafting, summarizing, explaining, coding -- that a fixed classifier can't do.
- Where it fits
- A subset of AI; large language models are the dominant current form of generative AI for text.
Core concepts
- Generation vs. classification
- Token-by-token text generation
- Sampling and randomness in output
Example
The distinction matters for evaluation: a classifier has a single correct answer to check against; generative output requires judging quality along multiple dimensions.
// A classifier answers: "Is this email spam?" (yes/no)
// A generative model answers: "Write a reply to this email."
// (open-ended output, not a fixed label)Common use cases
- Drafting and summarizing text
- Code generation
- Conversational assistants
Project ideas
- Compare three different prompts for the same generative task and note how output quality changes