Replies: 1 comment
-
Fascinating idea. We're definitely barely scratching the surface of linking models to add context, and better mapping. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
TL;DR; Would predigesting or embedding indexes improve fidelity of AI response? Would these non-note specific embeds improve relevance?
Would there be higher fidelity in AI responses if notes were predigested with timeline, machine learning contexts like Jaccard or Overlap, and natural language processing?
For example, before embedding, a map of notes modified and created in time, as well as the YAML of each note, is created. Then, someone can query, "List a summary of all notes in the last two days."
The YAML
type: Article
andlast modified <= 48 hours
would be embedded with the specific context of timeline and classification. This would give the AI more context about where to look.Another example would be to extract all proper names, places, and things, classify them, and embed that classification. A prompt can ask, "What notes have I taken that would interest people like John Doe?"
AI might respond as it does when asked to write in someone else's voice. This would help a marketer take notes about conversations with specific people and predict other topics worth future discussions.
As a newsletter publisher, I'd create a buyer persona around
John Doe
, link in some real customer conversations, and use geographic associations. "John Doe is a 28-year-old Midwestern male interested in hiking and technology. He is similar to [[Jane Doe]], [[Mike Doe]], and [[Roger Doe]] in outlook on life."All my conversations with those other people, as well as any profiles and preferences, would be associated with
John Doe.
The pre-digestion may reinforce context even if I haven't had a specific conversation with the proxy.I build contextual mapping using Dataview, DB Folder, and Graph Analysis. If those derived tables could be embedded, AI may reply in more detail. Fidelity, to me, is higher context and accuracy in predictions.
It's okay if this happens in the Smart Connect application, but it has to happen on my machine locally and securely. Smart actions aren't as important to me as improving the context between ideas and individuals.
Beta Was this translation helpful? Give feedback.
All reactions