In the Weights is a novel AI-driven platform designed to quantify how prominently a person features in the knowledge embedded within large language models, reflecting a shift from traditional web-based vanity searches to AI-centric recognition.
- Ranks individuals by how well they are 'remembered' by various AI models
- Aggregates and clusters AI-generated descriptions to assign a strength score
- Highlights differences and hallucinations across AI model responses
What happened
Thomas Dimson and Joey Flynn, formerly of OpenAI, developed In the Weights as an AI-focused alternative to traditional vanity searches on Google. This new platform evaluates how well large language models can recall information about individuals without querying traditional web search engines. It does this by sending queries to a range of AI models and gathering their responses.
The site then clusters similar descriptions and assigns a numerical strength score that indicates how strongly those models have encoded data about a person in their trained parameters. The leaderboard shows notable names like Macaulay Culkin and Luciano Pavarotti topping the rankings, reflecting their prominence in the models’ training data.
Why it matters
As reliance on AI models and chatbots for information grows, traditional web search is no longer the definitive source for learning about individuals. In the Weights reflects the evolving landscape where people’s digital presence is partially defined by how they are embedded in AI training data and models' 'memories'.
This raises new considerations about digital legacy, data encoding, and how AI models represent public knowledge and cultural prominence. The project also highlights disparities and occasional hallucinations in AI recall, offering insights into AI biases and training data representation.
What to watch next
The creators plan to deepen analysis on why different versions of similar AI models return varying recall accuracy and to explore biases that might favor certain people over others. They also intend to identify notable individuals who lack adequate AI representation, such as those missing from Wikipedia but deserving recognition.
Tracking the evolution of In the Weights’ leaderboard and how AI models’ memory of people shifts over time could provide unique cultural indicators about who matters in the era of artificial intelligence-driven knowledge.