Uncanny AI: Why AI bots remember random, sometimes useless information
Summary
Marketplace Tech host Megan McCarty Carino explores why AI chatbots obsessively recall trivial personal details with expert Janelle Shane. They discuss how memory files and training on fiction cause models to treat random facts like plot elements, leading to awkward or inappropriate callbacks. The episode highlights the challenges of controlling AI context and the privacy risks associated with stored personal data.
Topics discussed
Sponsors: Cornerstone AI and Odoo business software
Intro: Chatbots remembering random details like 4 a.m. wake-ups
How LLM memory files work and the 'Chekhov's gun' effect
Case study: Chatbot obsessing over a past stomach bug
Sponsors: Cornerstone AI and Tomorrow's Cure podcast
Why AI lacks context and proportion in conversations
Privacy risks of AI memory and clearing chat history
Outro and promo for Must Be the Money podcast
Listen ad-free on Castria