AI Compared 196 Hindu and Buddhist Gods — and Found Near-Identical Twins
A machine just found two goddesses from different religions at the same coordinates.
A single researcher fed descriptions of 196 Hindu and Vajrayana Buddhist deities into machine learning software. One program compared their bodies — how many arms they have, what they hold, whether they sit or stand. Another used a large language model (the same kind of AI behind chatbots) to turn written descriptions into long lists of numbers, so their theological roles could be compared mathematically. Then he checked which deities from different religions ended up closest to each other.
The most striking match: Chinnamasta, a Hindu goddess, and Chinnamunda, her Buddhist counterpart. Both are depicted holding their own severed heads. In the visual comparison they scored 0.288 apart, and in the meaning comparison just 0.068 apart. On a scale where zero means identical, that is essentially a bullseye. The paper calls it a rare "1:1 transfer" — strong evidence the two traditions share a common root or borrowed from each other. The AI also found what the author calls "iconographic camouflage": deities with very different looks serving quietly identical spiritual functions.
There is a twist, and it is the most useful part. The author models something he calls the "Atin Effect." When a single dramatic detail appears — a veena (a stringed instrument) or a severed head — that one detail overrides everything else in the model, dragging orthodox and Tantric figures into the same cluster. That is both a real observation about how human minds latch onto memorable symbols, and a warning for anyone building AI tools: one very distinctive feature can hijack your whole analysis.
Why does this matter outside a theology department? It is a proof of concept that AI can be pointed at culture itself — art, artifacts, rituals, texts — and find patterns across thousands of items faster than any human team. Museums, translators, and anthropologists could use it. But two honest caveats: the paper is still under peer review and comes from one author, and the meaning comparison ran on AI-generated descriptions, which can carry their own biases. Religious meaning resists arithmetic, and this tool maps resemblance, not truth.
- Machine learning sifted 196 Hindu and Buddhist deities by appearance and by meaning, then measured which ones sat closest together.
- A Hindu goddess and a Buddhist goddess — both shown holding their own severed heads — scored nearly identical on both measures.
- The author open-sourced the method, so historians and museums could reuse it to catalog cultural material at a scale humans cannot match.
Why It Matters
AI is moving from writing text to mapping culture, which could speed up how museums, translators, and scholars study art and religion.