In detail
- Created by Joey Flynn and Thomas Dimson, both formerly at OpenAI.
- Site queries multiple models, aggregates results and awards a 'strength score' (examples in article: 175, 262; leaderboard max 996 for very famous names).
- Smaller models make appearing in results harder; appearing in Meta's Llama (~1B params) signals high relevance.
- Creators note limits: models can hallucinate biographies, typos lower scores, common names reduce reliability.
Why it matters
Knowing whether people are memorised by models matters for privacy, reputation and compliance when models were trained on public or leaked data — relevant for hiring, customer data and legal risk.
For you Check whether key people in your organisation could appear in external models; document data sources used for training and assess privacy/reputation exposure with your providers.