ABOUT
About
AI Atlas is an open, free map of AI knowledge. It tries to serve three readers in one language: people with no technical background, people studying systematically, and practitioners who need to look something up quickly.
- Domains
- 8
- Core concepts
- 48
- Interface languages
- 12
- Figures & charts
- 124
Why this site exists
Material about AI is already abundant; what is missing is structure. Terms are scattered across papers, blogs, courses and news, and the dependencies between them have never been drawn clearly. The first-principles goal here is to draw the line: what to learn first, what next, and what depends on what.
Editorial principles
- Give the precise definition first, then the plain-language analogy. The order cannot be reversed: analogy without definition leaves the reader with nothing once they close the tab.
- Every concept must state its boundaries: what it solves, when it fails, and how it is commonly misused.
- Mathematics is not avoided, but it is placed where it is needed. A formula is a supplement, not a gate.
- No marketing language. Not “revolutionary”, not “disruptive”, not “powerful” — only what it does and what it cannot do.
How it is built
The whole site is statically pre-rendered with no backend. Every chart is drawn by hand-written SVG components with no charting library — because only by owning the coordinate system can a chart stay readable on a narrow screen.
On languages
The interface, domain names, entry titles and summaries are localised into twelve languages. The detailed articles are currently complete in Chinese and English; other languages fall back to English, with a notice at the top of the page. The locale structure is built for expansion — adding a language means adding one JSON file.
Licence & credits
Content is released under CC BY 4.0; code under MIT. Papers and articles linked in “Further reading” remain the property of their authors.