Switching to a New AI Chatbot? You Don’t Have to Start All Over | WSJ News
AI chatbots store detailed personal information called 'memory' that users can transfer between different AI services. This memory allows chatbots to provide more personalized responses, as demonstrated by comparing generic versus personalized travel recommendations for Paris.
Summary
The transcript explains how AI chatbots maintain detailed records of user interactions and preferences in what's called 'memory.' Users can access this information through settings and transfer it to new chatbots when switching services, either by copying the data manually or using official import prompts available in services like Claude and Gemini. The demonstration shows a stark contrast between responses from Claude in incognito mode (providing generic tourist recommendations like the Eiffel Tower and Louvre) versus personalized mode, where the AI knew the user lived in Paris and has a young child, resulting in more practical family-friendly suggestions with preferred formatting. The chatbot's memory can contain highly specific personal details including location and personal documents, which users can review, edit, or delete. While AI companies aren't data brokers, this personal data is valuable because it makes their chatbots more useful and personalized for individual users.
Key Insights
- AI chatbots store surprisingly specific personal information including location details and contents of personal documents like wills
- Claude provided drastically different travel recommendations when using incognito mode versus personalized mode, showing generic tourist attractions versus family-friendly options tailored to someone with a toddler
- AI companies find personal data valuable not for selling but because it makes their chatbots more useful and valuable to individual users
Topics
Full transcript available for MurmurCast members
Sign Up to Access