Website "In the Weights" shows whether AI models know who you are
Quick Answer
The website 'In the Weights' created by former OpenAI employees reveals how well AI models can recall individuals from their training data, with a scoring system up to 996.
Quick Take
Notable figures like Mozart, Shakespeare, and Taylor Swift rank highest, indicating the depth of their representation in AI datasets.
Key Points
- Website scores individuals based on AI recall strength, with a maximum of 996.
- Mozart, Shakespeare, and Taylor Swift are the top-ranked individuals.
- The platform highlights the extent of AI training data representation.
- Developed by two former employees of OpenAI.
- Reveals implications for privacy and data usage in AI.
📖 Reader Mode
~1 min readThe site In the Weights reveals which people are "stored" in the weights of large language models. Those "weights" are billions of numerical values where AI models encode their knowledge. If you show up in them, the model considered you relevant enough during training to recall without tools like web search.
The site queries several models to figure out who a specific person is, combines the results, and assigns a strength score. My colleague Maximilian Schreiner and I currently have strength scores of 175 and 262, for example. According to the leaderboard, the maximum strength score is 996, reserved for names like Mozart, Shakespeare, or Taylor Swift.
The site was built by Joey Flynn and Thomas Dimson, both former OpenAI employees. According to the creators, smaller models make it harder to show up in results. So anyone who appears in Meta's Llama, which has a billion parameters, counts as highly relevant. The creators also flag the obvious limits of LLMs, like that models can hallucinate biographical details, typos drag down scores, and common names often produce worse results.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
