Building a Personal AI Companion Called Dina
The author wrote a novel in 2017 about a personal AI assistant named Dina, and has now built a real-world implementation of Dina. Dina aims to solve the hidden problems of current AI agents, such as lack of cryptographic identity and data security.
Why it matters
Dina's approach to AI agent identity and data security could help make personal AI assistants more trustworthy and useful.
Key Points
- 1Dina is a sovereign AI companion that acts on the user's behalf, unlike task-oriented AI agents
- 2Dina has a cryptographic identity and encrypted persona vaults to ensure data security
- 3Dina provides an additional layer of safety for existing AI agents by gating risky actions
Details
In 2017, the author published a novel called UTOPAI, which envisioned a world where every person has a personal AI assistant named Dina. Unlike task-oriented AI agents, Dina is a sovereign companion that knows the user's preferences, friends, and promises, and acts on the user's behalf. Eight and a half years later, the author has built a real-world implementation of Dina. Dina addresses the hidden problems of current AI agents, such as the lack of a cryptographic identity and proper data security. Dina has her own cryptographic identity based on Ed25519 keys, and the user's data is stored in encrypted persona vaults with separate keys for different types of data. Dina also provides an additional layer of safety for existing AI agents by gating risky actions and requiring explicit user approval.
No comments yet
Be the first to comment