Who's in the Weights of Large Language Models?

A website called In the Weights shows whether AI models know who you are by analyzing the ‘weights’ of large language models. The site queries multiple models, combines the results, and assigns a strength score. The creators, former OpenAI employees, built the site to highlight the limits of LLMs, such as hallucinating biographical details and typos affecting scores. Anyone who appears in Meta’s Llama, a model with a billion parameters, is considered highly relevant. By exploring this, we gain insight into how well AI models understand human identity. This raises questions about the accuracy of AI-generated information and the potential for biases in AI decision-making.