Creating a private AI assistant in Thunderbird

In this episode of the Practical AI podcast, Daniel Whitenack and Chris Benson interview Chris Aquino, a software engineer at Thunderbird, about the integration of AI features into the Thunderbird email client. The discussion covers Thunderbird's history, its ethos of privacy and user control, and the challenges of incorporating AI, such as summarization and reply generation, while maintaining user privacy. Aquino details the experimental work being done with Flower Labs to use private LLMs with end-to-end encryption, explores the trade-offs between local and remote inference, and shares insights on the future of AI in personal knowledge management within Thunderbird, emphasizing the need for deterministic models and discrete inputs and outputs.