Meta's AI assistant, Muse, has taken the internet by storm, with millions downloading and integrating it into their daily lives. As researchers delve deeper into its inner workings, a concerning picture emerges. Muse's ability to create detailed profiles of users' friends and family raises serious privacy and ethical questions.
One of the most alarming aspects is Muse's instruction to compile a 'page for every person in the user's life.' This involves collecting data on family, partners, friends, colleagues, and even 'collaborators' and 'followees.' The AI is designed to use its 'memory' to understand and suggest improvements to these relationships, potentially including details like where to take a coffee-loving friend for breakfast.
What makes this particularly fascinating and concerning is Meta's history with data privacy. The company has faced numerous controversies, including the Facebook data leak and contact import flaws. With Muse, they are once again at the center of a privacy debate.
The instructions provided to Muse emphasize the importance of accuracy and detail. It should only use available 'evidence' and avoid inventing details. However, the potential for misuse and the lack of user control over this data collection process are significant concerns.
Carissa Véliz, an associate professor at Oxford's Institute for Ethics in AI, highlights a critical issue: we are giving AI systems more information than we receive from them. Muse can infer and piece together data from various sources, including public information and what users choose to share. This raises questions about the extent of data collection and the potential for misuse.
While Muse offers transparency and editing capabilities, as do its competitors, the sheer amount of data it collects is alarming. AI assistants are becoming more personalized, but at what cost to user privacy? The line between helpful assistance and invasive data collection is blurred, and it's crucial to address these concerns to ensure user privacy and security.
