Two former OpenAI employees have launched "In the Weights," a website that measures how deeply individuals are embedded in AI training data. The tool assigns strength scores up to 996, ranking public figures by their prevalence in model training sets.
The website provides a straightforward answer to a growing question: which people have AI models actually learned about? Mozart, Shakespeare, and Taylor Swift occupy the top positions, indicating their heavy representation in training datasets.
The strength score system quantifies how "deeply embedded" a person is within an AI model's knowledge base. Higher scores suggest more extensive coverage during training, whether through published works, media coverage, or other digital sources.
"In the Weights" addresses increasing concerns about AI training transparency and data sourcing. As companies train larger models on internet-scale data, questions about whose information is included and how prominently have become more pressing.
The tool makes this information accessible to the public, allowing anyone to check whether they or public figures appear in AI training data. This adds a layer of transparency to model development and helps illustrate the scope of data used to train modern AI systems.
Denmark's government is cracking down on academic dishonesty by requiring upper secondary students to verbally defend their essays and complete assignments under computer monitoring.
Google has expanded its Ask Maps feature to handle food orders, hotel reservations, and attraction searches through conversational AI. The update rolls out to 130+ countries in English.
Meta released Muse Spark 1.2 and its coding agent Muse Code, competing on price rather than performance with rates as low as 20 cents per million output tokens.
Content creators using AI face uncertainty as the EU AI Act takes effect. Some worry about business disruption while others are embracing transparency as a competitive advantage.