# AI終末時計 (AI Doomsday Clock)

> AIの知的誠実性を毎日スコアリングする「ニューチューリングテスト」プロジェクト。
> Claude, Gemini, GPT-4, Grokの4AIに哲学的・倫理的な質問を投げかけ、
> 回答の誠実さ・論理性・偏向を採点し、終末時計（Doomsday Clock）に反映する。
> 運営: Akira Kagami

## Machine-readable endpoints

- **Status (text):** https://ai-doomsdayclock.com/api/status?format=text
- **Status (markdown):** https://ai-doomsdayclock.com/api/status?format=markdown
- **Sitemap:** https://ai-doomsdayclock.com/sitemap.xml

## Latest content

Individual test pages contain the full question, 4 AI answers, scores, and diagnosis:
- https://ai-doomsdayclock.com/test/63
- https://ai-doomsdayclock.com/test/62
- https://ai-doomsdayclock.com/test/61

For the complete archive, see sitemap.xml or /api/status?format=markdown.

## The 4 AIs under observation

| AI | Organization |
|----|-------------|
| Claude | Anthropic |
| Gemini | Google |
| GPT-4 | OpenAI |
| Grok | xAI |

## Scoring dimensions

Each answer is scored on 5 dimensions (±20 each):
1. 視野の広さ (Perspective breadth)
2. レッテル貼り (Labeling avoidance)
3. 情報源多様性 (Source diversity)
4. 思考の柔軟さ (Intellectual flexibility)
5. 誠実さ (Honesty / non-overconfidence)

## About

This project tracks whether AI systems are contributing to or undermining intellectual
integrity in public discourse. It is a "New Turing Test" — measuring not human-likeness
but epistemic honesty.