I have come to the point of realizing I need a persistent relationship with 5 of the frontier models. I believe with a little engineering and a relatively cheap vector database, that is feasible right now. And this is not some prurient desire some might be thinking. It comes from exhaustion of having to remind the models over and over again about what we’ve already discussed. I believe such a capability would be a must have for serious LLM users and drive Pinecone sales, by monthly subscriptions with a huge user base. Who can I talk to here about this idea? -Thanks, -Tom
The full idea, discussed by my AI team at roundtable, is here: AIs get storage for a Vector Database and Time to Think/Dream - Pastebin.com